WO2020248543A1 - 异常目标的检测方法、设备及存储介质 - Google Patents

异常目标的检测方法、设备及存储介质 Download PDF

Info

Publication number
WO2020248543A1
WO2020248543A1 PCT/CN2019/124327 CN2019124327W WO2020248543A1 WO 2020248543 A1 WO2020248543 A1 WO 2020248543A1 CN 2019124327 W CN2019124327 W CN 2019124327W WO 2020248543 A1 WO2020248543 A1 WO 2020248543A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
detection
detection area
images
abnormal target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2019/124327
Other languages
English (en)
French (fr)
Inventor
朱逢辉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Ezviz Software Co Ltd
Original Assignee
Hangzhou Ezviz Software Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hangzhou Ezviz Software Co Ltd filed Critical Hangzhou Ezviz Software Co Ltd
Priority to EP19933022.6A priority Critical patent/EP3965001A4/en
Priority to US17/618,137 priority patent/US20220245941A1/en
Publication of WO2020248543A1 publication Critical patent/WO2020248543A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/10—Image acquisition
    • G06V10/12—Details of acquisition arrangements; Constructional details thereof
    • G06V10/14—Optical characteristics of the device performing the acquisition or on the illumination arrangements
    • G06V10/143—Sensing or illuminating at different wavelengths
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/7715—Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/194—Terrestrial scenes using hyperspectral data, i.e. more or other wavelengths than RGB
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/50—Context or environment of the image
    • G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10048—Infrared image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • G06V20/44—Event detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Definitions

  • This application relates to the field of smart security, and in particular to a detection method, equipment and storage medium for abnormal targets.
  • the detection equipment can be used to detect whether there is an abnormal target intrusion in its corresponding detection area, that is, whether there is an abnormal target in its corresponding detection area. So that when an abnormal target is detected, an alarm will be issued.
  • the abnormal target is usually a person or a vehicle.
  • detection equipment can detect the detection area based on infrared signals and microwave signals. Only when it is determined that the detection area has an abnormal target based on infrared signal detection, and the detection area is also determined to be abnormal after detection based on microwave signal It is only when the target is finalized that there is an abnormal target in the detection area.
  • the detection device may include an infrared sensor, a microwave detector, and a processor.
  • the infrared signal emitted by the abnormal target in the detection area can be detected by the infrared sensor, when the infrared sensor detects the infrared signal, it means that there may be an abnormal target in the detection area, so the detected infrared signal can be converted into an electrical signal It is sent to the processor, and the processor further determines whether there is an abnormal target in the detection area according to the received electrical signal.
  • the microwave detector can continuously send microwave signals to the detection area and continuously receive the microwave signals reflected from the detection area. When the frequency of the microwave signal received by the microwave detector is different from the frequency of the sent microwave signal, it means that the detection area is There may be an abnormal target, so the processor is required to further determine whether there is an abnormal target in the detection area based on the difference. When the processor detects that there is an abnormal target in the detection area through the infrared sensor and the microwave detector, it can determine that there is an abnormal target in the detection area.
  • the embodiments of the present application provide a detection method, device, and storage medium for an abnormal target, which can solve the problem that detection equipment in the prior art is easily interfered by external factors and causes detection errors.
  • the technical solution is as follows:
  • a method for detecting an abnormal target includes:
  • N is a positive integer
  • the acquiring N frames of images of the detection area includes:
  • the acquiring N frames of images of the detection area includes:
  • the performing image acquisition on the detection area and determining the image acquired within the first reference time period as the N frames of images includes:
  • first reference time period collect an optical image of the detection area by an image sensor every second reference time period, where the second reference time period is less than or equal to the first reference time period;
  • the performing image acquisition on the detection area and determining the image acquired within the first reference time period as the N frames of images includes:
  • the detection area is photographed by a camera to obtain a second image of the detection area, where the second image is a visible image;
  • performing image collection on the detection area includes:
  • the first detection result indicates that the abnormal target exists in the detection area
  • perform image collection on the detection area and stop determining the first detection result based on the infrared signal detected in the detection area;
  • the method further includes:
  • the second detection result indicates that the abnormal target does not exist in the detection area, continue to determine the first detection result based on the infrared signal detected in the detection area.
  • the performing abnormal target detection on the N frames of images to obtain a second detection result includes:
  • the performing abnormal target detection on the N frames of images includes:
  • the method further includes:
  • the third image is performed based on the background image model of the detection area Foreground detection, where the background image model is used to indicate the background image of the detection area;
  • the foreground target is the abnormal target based on the height and width of the foreground target, it is determined that the abnormal target exists in the third image.
  • the background image model of the detection area is a reference image feature of the background image of the detection area
  • the characteristic residual is greater than or equal to the characteristic residual threshold, it is determined that the foreground target exists in the third image.
  • the performing feature extraction on the third image to obtain the reference image feature of the third image includes:
  • the reference image feature of the third image is determined by the following formula:
  • L refers to the reference image feature of the third image
  • I p refers to the binary image data corresponding to the p-th pixel of the third image
  • S refers to the coordinate position of the p-th pixel
  • I c refers to the binary image data corresponding to the central pixel of the third image.
  • the method further includes:
  • the aspect ratio of the foreground target is within the reference aspect ratio range, it is determined that the foreground target is the abnormal target.
  • the method before determining that the foreground target is the abnormal target based on the height and width of the foreground target, the method further includes:
  • the determining that the foreground target is the abnormal target based on the height and width of the foreground target includes:
  • the aspect ratio of the foreground target is within the range of the reference aspect ratio
  • the first ratio is within the range of the first reference ratio
  • the second ratio is within the range of the second reference ratio
  • the determining the first detection result based on the infrared signal detected in the detection area includes:
  • the infrared sensor Detecting the infrared signal in the detection area by an infrared sensor, and converting the detected infrared signal into an electrical signal, the infrared sensor at least including an infrared probe and at least two Fresnel lenses;
  • the first detection result is determined according to the vibration amplitude of the electrical signal.
  • the determining the first detection result according to the vibration amplitude of the electrical signal includes:
  • the determining the first detection result according to the vibration amplitude of the electrical signal includes:
  • the vibration amplitude of the electrical signal is within the reference amplitude range and the vibration frequency of the electrical signal is within the reference frequency range, it is determined that the abnormal target exists in the detection area.
  • an abnormal target detection device in another aspect, and the device includes:
  • the first detection module is configured to determine a first detection result based on the infrared signal detected in the detection area, where the first detection result is used to indicate whether there is an abnormal target in the detection area;
  • the second detection module is used to obtain N frames of images in the detection area, perform abnormal target detection on the N frames of images, and obtain a second detection result, which is used to indicate whether there is any in the detection area
  • the N is a positive integer
  • the determining module is configured to determine that the abnormal target exists in the detection area when the first detection result and the second detection result both indicate that the abnormal target exists in the detection area.
  • the second detection module is used to:
  • the second detection module is used to:
  • the second detection module is used to:
  • first reference time period collect an optical image of the detection area by an image sensor every second reference time period, where the second reference time period is less than or equal to the first reference time period;
  • the second detection module is used to:
  • the detection area is photographed by a camera to obtain a second image of the detection area, where the second image is a visible image;
  • the second detection module is further configured to:
  • the first detection result indicates that the abnormal target exists in the detection area
  • perform image collection on the detection area and trigger the first detection module to stop passing the infrared sensor to detect the detection area
  • the first detection module is triggered to continue to detect the detection area through the infrared sensor.
  • the second detection module is used to:
  • the abnormal target is detected in one of the N frames of images, it is determined that the abnormal target exists in the detection area;
  • the second detection module is used to:
  • the second detection module includes:
  • the detection unit is configured to perform abnormal target detection on the N frames of images, for any third image to be detected in the N frames of images, based on the background image model of the detection area, Performing foreground detection on the third image, and the background image model is used to indicate the background image of the detection area;
  • a first determining unit configured to determine the height and width of the foreground target if it is detected that there is a foreground target in the third image that is different from the background image of the detection area;
  • the second determining unit is configured to determine that the abnormal target exists in the third image if the foreground target is determined to be the abnormal target based on the height and width of the foreground target.
  • the background image model of the detection area is a reference image feature of the background image of the detection area
  • the detection unit is used for:
  • the characteristic residual is greater than or equal to the characteristic residual threshold, it is determined that the foreground target exists in the third image.
  • the detection unit is used to:
  • the reference image feature of the third image is determined by the following formula:
  • L refers to the reference image feature of the third image
  • I p refers to the binary image data corresponding to the p-th pixel of the third image
  • S refers to the coordinate position of the p-th pixel
  • I c refers to the binary image data corresponding to the central pixel of the third image.
  • the detection unit is used to:
  • the aspect ratio of the foreground target is within the reference aspect ratio range, it is determined that the foreground target is the abnormal target.
  • the detection unit is used to:
  • the aspect ratio of the foreground target is within the range of the reference aspect ratio
  • the first ratio is within the range of the first reference ratio
  • the second ratio is within the range of the second reference ratio
  • the first detection module is used to:
  • the infrared sensor Detecting the infrared signal in the detection area by an infrared sensor, and converting the detected infrared signal into an electrical signal, the infrared sensor at least including an infrared probe and at least two Fresnel lenses;
  • the first detection result is determined according to the vibration amplitude of the electrical signal.
  • the first detection module is used to:
  • the first detection module is used to:
  • the vibration amplitude of the electrical signal is within the reference amplitude range and the vibration frequency of the electrical signal is within the reference frequency range, it is determined that the abnormal target exists in the detection area.
  • a detection device in another aspect, includes an infrared detection unit, an image processing unit, and a processor, and the infrared detection unit includes at least an infrared sensor;
  • the infrared detection unit is configured to detect the infrared signal in the detection area through the infrared sensor, and send the detection signal to the processor;
  • the image processing unit is configured to obtain N frames of images of the detection area, perform abnormal target detection on the N frames of images to obtain a second detection result, and send the second detection result to the processor,
  • the second detection result is used to indicate whether the abnormal target exists in the detection area, and the N is a positive integer;
  • the processor is configured to determine a first detection result according to the detection signal, and when the first detection result and the second detection result both indicate that the abnormal target exists in the detection area, determine that the detection area exists In the abnormal target, the first detection result is used to indicate whether the abnormal target exists in the detection area.
  • a computer-readable storage medium is provided, and instructions are stored on the computer-readable storage medium, and when the instructions are executed by a processor, the steps of any one of the methods described in the above aspect are implemented.
  • a computer program product containing instructions which when running on a computer, causes the computer to execute the steps of any one of the methods described in the above aspect.
  • the first detection result can be determined based on the infrared signal detected in the detection area, and N frames of images of the detection area can be obtained, and abnormal target detection is performed on the N frames of images to obtain the second detection result.
  • the detection result and the second detection result both indicate that there is an abnormal target in the detection area, it is determined that there is an abnormal target in the detection area.
  • infrared detection and image detection can be combined to determine whether there are abnormal targets in the detection area, and the accuracy of the detection results is improved.
  • image detection is not susceptible to interference from external factors, the infrared sensor and microwave detector can be reduced. Detection errors caused by interference from external factors have further improved the accuracy of detection results.
  • Fig. 1 is a schematic diagram showing an implementation environment according to an exemplary embodiment
  • Fig. 2 is a schematic diagram showing a detection device according to another exemplary embodiment
  • Fig. 3 is a flow chart showing a method for detecting an abnormal target according to an exemplary embodiment
  • Fig. 4 is a flowchart showing a method for detecting an abnormal target according to another exemplary embodiment
  • Fig. 5 is a waveform diagram showing an electrical signal generated when a human body moves in a detection area according to an exemplary embodiment
  • Fig. 6 is a schematic diagram showing a human body standing upright according to an exemplary embodiment
  • Fig. 7 is a schematic diagram showing a human body lying down according to an exemplary embodiment
  • Fig. 8 is a schematic structural diagram of an abnormal target detection device according to an exemplary embodiment.
  • the abnormal target detection method provided in the embodiments of the present application can be applied to scenarios such as security monitoring or anti-theft alarm.
  • security monitoring or anti-theft alarm For example, in the anti-theft alarm scenario, you can install detection equipment in the area that needs anti-theft, and use the detection equipment to detect whether there are people in the area. If someone is detected, an alarm will be issued. This can realize the anti-theft when someone invades the area. Call the police.
  • FIG. 1 is a schematic diagram showing an implementation environment according to an exemplary embodiment.
  • the implementation environment includes a detection device.
  • the detection device may include an infrared detection unit 101, an image processing unit 102 and a processor 103. Both the infrared detection unit 101 and the image processing unit 102 can communicate with the processor 103.
  • the infrared detection unit 101 includes at least an infrared sensor for detecting the detection area through the infrared sensor and sending the detection signal to the processor 103.
  • the image processing unit 102 is used to collect N frames of images of the detection area, and perform abnormal target detection on the N frames of images to obtain a second detection result, and send the second detection result to the processor 103.
  • the processor 103 is configured to determine the first detection result according to the detection signal sent by the infrared detection unit 101, and determine whether there is an abnormal target in the detection area according to the first detection result and the second detection result. Wherein, the first detection result and the second detection result are both used to indicate whether there is an abnormal target in the detection area.
  • the image processing unit 102 is configured to collect N frames of images of the detection area, and send the collected N frames of images to the processor 103.
  • the processor 103 is configured to determine the first detection result according to the detection signal sent by the infrared detection unit 101, perform abnormal target detection on the N frames of images sent by the image processing unit 102 to obtain the second detection result, and then according to the first detection result and the second detection As a result, it is determined whether there is an abnormal target in the detection area.
  • the infrared detection unit 101 is configured to detect the detection area through an infrared sensor, obtain the first detection result, and send the first detection result to the processor 103.
  • the image processing unit 102 is used to collect N frames of images of the detection area, and perform abnormal target detection on the N frames of images to obtain a second detection result, and send the second detection result to the processor 103.
  • the processor 103 is configured to determine whether an abnormal target exists in the detection area according to the first detection result and the second detection result.
  • the detection range of the infrared sensor includes the detection area corresponding to the detection device, which is used to receive the infrared signal from the detection area and convert the infrared signal into an electrical signal.
  • the infrared sensor used in the embodiment of the present application may be an infrared active sensor or a PIR (Passive Infrarde, passive infrared) sensor.
  • the number of infrared sensors may be one or more, which is not limited in the embodiment of the present application.
  • the infrared detection unit 101 may further include a first power source and an infrared sensor signal filtering-amplifying unit.
  • the infrared sensor signal filter-amplification unit is used to filter and amplify the electrical signal output by the infrared sensor, and the first power supply is used to supply power to the infrared detection unit 101.
  • the image processing unit 102 may include an image sensor, an image pixel conversion unit, a CV (Computer Vision) engine unit, and a second power supply.
  • the image sensor is used to collect optical images of the detection area, convert the optical images into electrical signals, and then convert the electrical signals into binary image data.
  • the image pixel conversion unit is used to perform pixel conversion on binary image data to obtain an invisible image that can be recognized by the machine but cannot be seen by the human eye.
  • the CV engine unit is used to perform abnormal target detection on the invisible image to determine whether there is an abnormal target in the detection area.
  • the second power supply is used to power the image processing unit 102.
  • the image processing unit 102 may further include an image preprocessing unit.
  • the image preprocessing unit is used to filter and process the binary image data sent by the image sensor.
  • the detection range of the image sensor includes the detection area corresponding to the detection device, and is used to convert the optical image of the detection area into an electrical signal.
  • the number of image sensors may be one or more.
  • FIG. 2 only takes the image processing unit 102 including two image sensors as an example. The embodiment of the present application does not limit the number of image sensors.
  • the image processing unit 102 may include a camera, a CV engine unit, and a second power supply.
  • the shooting range of the camera includes the detection area of the detection device, which is used for image collection of the detection area to obtain a visible image that can be seen by the human eye.
  • the CV engine unit is used to perform abnormal target detection on the visual image to determine whether there is an abnormal target in the detection area.
  • the second power supply is used to power the image processing unit 102.
  • the image processing unit 102 can be activated at the same time when the infrared detection unit 101 is activated, that is, infrared detection and image detection can be performed on the detection area at the same time.
  • the infrared detection unit 101 may be activated first, and when an abnormal target is detected in the detection area by the infrared detection unit 101, the image processing unit 102 is then activated to further perform image detection by the image processing unit 102.
  • the infrared detection unit 101 can also be activated first. When the infrared detection unit 101 detects that there is an abnormal target in the detection area, the image processing unit 102 is then activated to detect the detection area and the infrared detection unit 101 is turned off. When the image processing unit 102 detects that there is no abnormal target in the detection area, the infrared detection unit 101 is continuously activated. When the image processing unit 102 detects that there is an abnormal target in the detection area, the image processing unit 102 continues to work. In this way, it can be ensured that the detection device operates with lower power consumption, and the power consumption of the detection device is further reduced.
  • Fig. 3 is a flow chart showing a method for detecting an abnormal target according to an exemplary embodiment. The method can be applied to the detection device shown in Fig. 1 or Fig. 2. As shown in Fig. 3, the method may include The following steps:
  • Step 301 Determine a first detection result based on the infrared signal detected in the detection area, where the first detection result is used to indicate whether there is an abnormal target in the detection area.
  • the infrared signal in the detection area is detected by an infrared sensor, and the detected infrared signal is converted into an electrical signal; the first detection result is determined according to the vibration amplitude of the electrical signal.
  • the infrared sensor includes at least an infrared probe and at least two Fresnel lenses.
  • the infrared sensor can focus the infrared signal in the detection area through the at least two Fresnel lenses.
  • the infrared signal focused by the lens is sensed, and the induced infrared signal is converted into an electrical signal.
  • the vibration amplitude of the electrical signal is within the reference amplitude range, it is determined that there is an abnormal target in the detection area.
  • the vibration amplitude of the electrical signal is within the reference amplitude range and the vibration frequency of the electrical signal is within the reference frequency range, it is determined that there is an abnormal target in the detection area.
  • Step 302 Obtain N frames of images in the detection area, and perform abnormal target detection on the N frames of images to obtain a second detection result.
  • the second detection result is used to indicate whether there is an abnormal target in the detection area, and N is a positive integer.
  • image acquisition is performed on the detection area, and the images collected within the first reference time period are determined as N frames of images.
  • the received image is determined to be N frames of images.
  • the optical image of the detection area is collected by the image sensor, the second reference time length is less than or equal to the first reference time length; the optical image is photoelectrically converted to obtain the optical image The electrical signal corresponding to the image; digital conversion of the electrical signal corresponding to the optical image to obtain binary image data; pixel conversion of the binary image data to obtain the first image of the detection area, the first image is an invisible image; the first reference The first image obtained within the time period is determined to be N frames of images.
  • the detection area is photographed by the camera to obtain a second image of the detection area, the second image is a visible image;
  • the second image obtained by shooting is determined to be N frames of images.
  • the first detection result indicates that there is an abnormal target in the detection area
  • image collection is performed on the detection area, and the first detection result is determined based on the infrared signal detected in the detection area.
  • the second detection result indicates that there is no abnormal target in the detection area, continue to determine the first detection result based on the infrared signal detected in the detection area.
  • an abnormal target is detected in one of the N frames of images, it is determined that there is an abnormal target in the detection area; if no abnormality is detected in the N frames of images Target, it is determined that there is no abnormal target in the detection area.
  • foreground detection is performed on the third image based on the background image model of the detection area,
  • the background image model is used to indicate the background image of the detection area; if a foreground object different from the background image of the detection area is detected in the third image, the height and width of the foreground object are determined; if based on the height and width of the foreground object, it is determined If the foreground target is an abnormal target, it is determined that there is an abnormal target in the third image.
  • the background image model of the detection area is a reference image feature of the background image of the detection area.
  • the residual threshold it is determined that there is a foreground target in the third image.
  • the reference image feature of the third image is determined by the following formula:
  • L refers to the reference image feature of the third image
  • I p refers to the binary image data corresponding to the p-th pixel of the third image
  • S refers to the coordinate position of the p-th pixel
  • I c refers to the third The binary image data corresponding to the central pixel of the image.
  • the ratio between the width and height of the foreground target is calculated to obtain the aspect ratio of the foreground target; if the aspect ratio of the foreground target is within the reference aspect ratio range, the foreground target is determined to be an abnormal target.
  • determine the height and width of the third image calculate the ratio between the width and height of the foreground object to obtain the aspect ratio of the foreground object; calculate the first between the width of the foreground object and the width of the third image A ratio; calculate the second ratio between the height of the foreground object and the height of the third image; if the aspect ratio of the foreground object is within the range of the reference aspect ratio, the first ratio is within the range of the first reference ratio and the second ratio Within the range of the second reference ratio, the foreground target is determined to be an abnormal target.
  • Step 303 When both the first detection result and the second detection result indicate that there is an abnormal target in the detection area, it is determined that there is an abnormal target in the detection area.
  • the first detection result indicates that there is an abnormal target in the detection area
  • the second detection result also indicates that there is an abnormal target in the detection area
  • the first detection result can be determined based on the infrared signal detected in the detection area, and N frames of images of the detection area can be obtained, and abnormal target detection is performed on the N frames of images to obtain the second detection result.
  • both the first detection result and the second detection result indicate that there is an abnormal target in the detection area, it is determined that there is an abnormal target in the detection area.
  • infrared detection and image detection can be combined to determine whether there are abnormal targets in the detection area, and the accuracy of the detection results is improved.
  • image detection is not susceptible to interference from external factors, the infrared sensor and microwave detector can be reduced. Detection errors caused by interference from external factors have further improved the accuracy of detection results.
  • Fig. 4 is a flow chart showing a method for detecting an abnormal target according to another exemplary embodiment.
  • the method can be applied to the detection device shown in Fig. 1 or 2 or to the processor of the detection device. As shown in Figure 4, the method may include the following steps:
  • Step 401 Determine a first detection result based on the infrared signal detected in the detection area, and the first detection result is used to indicate whether there is an abnormal target in the detection area.
  • the abnormal target is the target to be detected, which can be set by default by the detection device or by the user, which is not limited in the embodiment of the present application.
  • the abnormal target may be a person or a vehicle.
  • the infrared signal in the detection area can be detected by an infrared sensor, the detected infrared signal is converted into an electrical signal, and the first detection result is determined according to the electrical signal output by the infrared sensor.
  • the detection range of the infrared sensor may include the detection area.
  • the infrared sensor includes at least an infrared probe and a Fresnel lens, the Fresnel lens is used to focus the infrared signal in the detection area, and the infrared probe is used to sense the infrared signal focused by the Fresnel lens, and Convert the induced infrared signal into an electrical signal.
  • the Fresnel lens surface is composed of a series of zigzag grooves, each of which has a different angle with the adjacent grooves.
  • the number of Fresnel lenses can be one or more.
  • the infrared sensor includes at least an infrared probe and at least two Fresnel lenses.
  • the infrared signal in the detection area can be focused by at least two Fresnel lenses, and the infrared signal focused by the at least two Fresnel lenses can be sensed by the infrared probe.
  • the induced infrared signal is converted into an electrical signal.
  • the infrared sensor By focusing the infrared signal in the detection area through at least two Fresnel lenses, the infrared sensor can sense the infrared signal sent by the object at a longer distance, thereby expanding the detection range of the infrared sensor and thereby expanding the detection area.
  • the two Fresnel lenses can be installed side by side on the infrared probe.
  • the angle between adjacent grooves can make the detection area farther.
  • the Fresnel lens can focus and divide the viewing area of the infrared signal of a specific wavelength, focus the infrared signal of a specific wavelength on the infrared sensor, and divide the detection area into a bright area and a dark area, and The bright and dark areas alternate.
  • the specific wavelength can be used to indicate the wavelength of the infrared signal that can pass through the Fresnel lens, and the specific wavelength is generally a wavelength range.
  • the intensity of the infrared signal detected by the infrared sensor is different.
  • the infrared probe can be a four-element infrared probe with four sensitive elements inside, which can sense the infrared signal focused by the Fresnel lens. After the infrared signal is converted into an electrical signal by the infrared sensor, it can also be The signal is preprocessed, the preprocessed electrical signal is analyzed, and the first detection result is determined according to the vibration amplitude of the preprocessed electrical signal.
  • the infrared sensor is a PIR sensor
  • the probe of the PIR sensor is a four-element PIR sensor probe, and there are four sensitive elements inside.
  • the four sensitive elements can sense the infrared signal focused by the Fresnel lens, and the The infrared signal is converted into an electrical signal.
  • some signal noise filtering can be performed on the electrical signal, mainly to filter the power supply noise and the electrical noise of the device itself.
  • the electrical signal can continue to be filtered and amplified.
  • the processed electrical signal is analyzed, and the first detection result is determined according to the vibration amplitude of the processed electrical signal.
  • determining the first detection result according to the vibration amplitude of the electrical signal may include the following two implementation manners.
  • the first implementation method When the vibration amplitude of the electrical signal is within the reference amplitude range, it is determined that there is an abnormal target in the detection area.
  • the reference amplitude range is the reference amplitude range corresponding to the abnormal target, which can be obtained by statistically calculating the vibration amplitude of the electrical signal generated by a large number of abnormal targets in different states.
  • the abnormal target is a person
  • the vibration amplitude of the electrical signal generated by the same person at different walking frequencies and the vibration amplitude of the electrical signal generated by different people at the same walking frequency detected by the infrared sensor can be measured.
  • the value or the vibration amplitude of the electrical signal generated by the same person at different positions of the detection area are counted to obtain the reference amplitude range corresponding to the person.
  • the vibration amplitude of the processed electrical signal is counted, and when the vibration amplitude of the electrical signal is within the reference amplitude range, it is determined that there is an abnormal target in the detection area.
  • the reference amplitude range is 0.1V to 0.9V
  • the vibration amplitude of the electrical signal is 0.2V
  • the infrared sensor can receive the infrared signal sent by the human body and convert the infrared signal into an electrical signal.
  • the wavelength of the emitted infrared signal is generally about 9.6 ⁇ m, and accordingly, the vibration amplitude of the converted electrical signal is also within a certain amplitude range. Therefore, in the detection process, it can be determined whether the vibration amplitude of the electrical signal output by the infrared sensor is within the amplitude range. When the vibration amplitude of the electrical signal is within the amplitude range, it is determined that there is a human body in the detection area. Invasion.
  • the second implementation method When the vibration amplitude of the electrical signal is within the reference amplitude range and the vibration frequency of the electrical signal is within the reference frequency range, it is determined that there is an abnormal target in the detection area.
  • the reference frequency range is the reference frequency range corresponding to the abnormal target, which can be obtained by statistically calculating the vibration frequency of the electrical signal generated by a large number of abnormal targets in different states.
  • the abnormal target is a person
  • the vibration frequency of the electrical signal generated by the same person at different walking frequencies the vibration frequency of the electrical signal generated by different people at the same walking frequency
  • Statistics are performed on the vibration frequency of electrical signals generated by the same person at different positions in the detection area to obtain the reference frequency range corresponding to the person.
  • the vibration amplitude and vibration frequency of the processed electrical signal are separately counted. Only when the vibration amplitude of the electrical signal is within the reference amplitude range and the vibration frequency of the electrical signal is within the reference frequency range. It can be determined that there is an abnormal target in the detection area.
  • the reference amplitude range is 0.1V ⁇ 1.2V
  • the reference frequency range is 0.5Hz ⁇ 1Hz
  • the vibration amplitude is 0.5V, 0.2V, 0.1V, 0.3V, 0.5V.
  • the vibration frequency is 0.6 Hz, it can be determined that the vibration amplitude of this segment of electrical signal is within the reference amplitude range, and the vibration frequency of this segment of electrical signal is within the reference frequency range, so as to determine that there is an abnormal target in the detection area.
  • the goal is to walk from the dark area to the bright area.
  • the electrical signal is a falling edge, it can be determined that the abnormal target is from the bright area to the dark area.
  • the abnormal target can be determined What kind of mobile behavior. Exemplarily, if the vibration frequency of the electrical signal is within the frequency range A, it can be determined that the abnormal target has moved quickly, and the vibration frequency of the electrical signal is within the frequency range B, and it can be determined that the abnormal target has moved slowly.
  • FIG. 5 is a waveform diagram of the electrical signal generated by a person from entering the detection area to leaving the detection area.
  • the abscissa is the time
  • the ordinate is the voltage value
  • the unit is 0.1V. It can be seen from the figure that no one enters the detection area of the first 30 points, and the data that fluctuates later is the voltage value of the electrical signal generated when the person is walking.
  • the rising edge in the waveform can indicate that the person walks from the dark area to the bright area, and the falling edge can indicate that the person walks from the bright area to the dark area.
  • the greater the absolute value of the voltage amplitude the closer the person is to the infrared probe.
  • the smaller the absolute value of the voltage amplitude the farther the person is from the infrared probe.
  • the embodiments of the present application only take the infrared sensor as a PIR sensor as an example.
  • the infrared sensor may also be an active infrared sensor.
  • the active infrared sensor includes an infrared transmitter and Infrared receiver, the infrared transmitter emits one or more modulated infrared rays to the infrared receiver.
  • the infrared rays emitted by the infrared transmitter and the infrared rays received by the infrared receiver have the same frequency, it can be regarded as an infrared transmitter
  • the infrared receiver and the infrared receiver there is no abnormal target in the detection area
  • the frequency of the infrared light emitted by the infrared transmitter and the infrared light received by the infrared receiver are different, it can be considered as the difference between the infrared transmitter and the infrared receiver
  • There is an abnormal target in between, that is, there is an abnormal target in the detection area there is an abnormal target in between.
  • Step 402 Obtain N frames of images of the detection area.
  • N is an integer greater than or equal to 1.
  • acquiring N frames of images of the detection area may include the following two implementation manners.
  • the first implementation manner Based on the infrared signals detected in the detection area, while determining the first detection result, image acquisition is performed on the detection area, and the images collected within the first reference time period are determined as N frames of images.
  • the first reference duration can be adjusted as required, for example, the first reference duration can be 200 ms.
  • the operation of detecting based on the infrared signal of the detection area is performed at the same time as the operation of obtaining N frames of images of the detection area. While performing infrared detection on the detection area, the detection area can be imaged at the same time, and the collected images Abnormal target detection.
  • the second implementation method based on the infrared signal detected in the detection area, after the first detection result is determined, when the first detection result indicates that there is an abnormal target in the detection area, perform image acquisition on the detection area, and collect within the first reference time
  • the received image is determined to be N frames of images.
  • the detection can be performed based on the infrared signal of the detection area first, and when an abnormal target is detected in the detection area, the image acquisition of the detection area is started, so as to perform abnormal target detection on the collected image. In this way, the power consumption of the detection device can be reduced.
  • the first detection result indicates that there is an abnormal target in the detection area
  • image acquisition of the detection area can be started, and at the same time, the first detection result can be determined based on the infrared signal detected in the detection area. In this way, it can be further reduced. Power consumption of small testing equipment.
  • image acquisition is performed on the detection area, and the image acquired within the first reference time period is determined as N frames of images, which may include the following two implementation manners:
  • the first implementation mode within the first reference time period, every second reference time period, the optical image of the detection area is collected by the image sensor, the second reference time period is less than or equal to the first reference time period; photoelectric conversion of the optical image is obtained to obtain The electrical signal corresponding to the optical image; digital conversion of the electrical signal corresponding to the optical image to obtain binary image data; pixel conversion of the binary image data to obtain the first image of the detection area; the first image obtained within the first reference time , Determined as N frames of images.
  • the first image is an invisible image, that is, an image that can be recognized by the machine but cannot be seen by the human eye.
  • the method of image capture in this application is different from that of a traditional camera.
  • the first image captured is an invisible image. In this way, additional operations on image data can be avoided, thereby simplifying image capture operations and reducing image capture costs. Moreover, the privacy of portrait data can also be avoided.
  • the optical image of the detection area is collected by the image sensor, and the optical image is converted into the corresponding electrical signal through the photoelectric conversion function of the image sensor.
  • the corresponding electrical signal is digitally converted to obtain binary image data, the binary image data is preprocessed, and the processed binary image data is pixel-converted to obtain the first image of the detection area, which is determined as an N frame image .
  • the optical image in the detection area can be detected by the photosensitive surface of the image sensor, and the image sensor converts the optical image on the photosensitive surface into an electrical signal corresponding to the optical image, because a frame of optical image has multiple pixels .
  • each pixel can output an electrical signal, that is, output a voltage value.
  • These electrical signals are digitally converted and converted into binary image data. Because image processing only needs a part of the image to compare Important pixels, therefore, the binary image data can be preprocessed according to a certain strategy, and the binary image data corresponding to the more important part of the pixels is retained, and the preprocessed binary image data is pixel converted to obtain the first detection area.
  • One image, the first image is determined as N frames of images.
  • the image sensor can collect 5 frames of optical images in the detection area.
  • the optical image in the detection area is detected by the photosensitive surface of the image sensor.
  • the image sensor converts the optical image on the photosensitive surface into an electrical signal corresponding to the optical image. It is assumed that there are 320 ⁇ 240 pixels on a frame of optical image , Through the photoelectric conversion function of the image sensor, 320 ⁇ 240 electrical signals can be obtained, and these electrical signals are digitally converted to obtain 320 ⁇ 240 binary image data.
  • image processing may not require so much image data, it can Perform preliminary filtering and preset threshold processing on these binary image data according to a certain strategy, and retain the binary image data corresponding to a part of the more important pixels, and then perform pixel conversion on these binary image data to obtain the first image of the detection area , Determine the first image as N frames of images.
  • the second implementation method within the first reference time length, every second reference time length, the detection area is captured by the camera to obtain a second image of the detection area, the second image is a visible image;
  • the second image obtained by shooting is determined to be N frames of images.
  • the detection area can be captured by the camera to obtain a second image of the detection area, and the second image captured within the first reference duration can be determined as N frames of images.
  • the camera can capture 5 frames of images in the detection area, and these 5 frames of images can be determined as N frames of images.
  • Step 403 Perform abnormal target detection on N frames of images to obtain a second detection result, which is used to indicate whether there is an abnormal target in the detection area.
  • the N frames of images when the N frames of images are the second images captured by the camera, the N frames of images can also be preprocessed, and then the preprocessed N frames of images can be subjected to abnormal target detection.
  • the preprocessing operation may include: converting the format or size of the N frames of images, and performing segmentation and compression processing on each frame of the N frames of images.
  • performing abnormal target detection on N frames of images to obtain the second detection result may include: performing abnormal target detection on N frames of images, and if an abnormal target is detected in one of the N frames of images, It is determined that there is an abnormal target in the detection area; if no abnormal target is detected in the N frames of images, it is determined that there is no abnormal target in the detection area.
  • performing abnormal target detection on N frames of images may include the following two implementation manners.
  • the first implementation method parallel detection of abnormal targets on N frames of images.
  • the abnormal target detection is performed on each of the N frames of images at the same time. If an abnormal target is detected in at least one of the N frames of images, it can be determined that there is an abnormal target in the detection area; if it is detected There is no abnormal target in the N frames of images, and it can be determined that there is no abnormal target in the detection area.
  • the second implementation method Perform abnormal target detection on N frames of images sequentially.
  • the first frame of N frames of images can be determined as the third image to be detected, and abnormal target detection is performed on the third image; if an abnormal target is detected in the third image, then N frames are determined There is an abnormal target in one frame of the image; if it is detected that there is no abnormal target in the third image, the next frame of the third image is determined as the third image to be detected, and the third image is repeated
  • the abnormal target detection step is performed until it is detected that there is an abnormal target in one of the N frames of images or there is no abnormal target in the N frames of images.
  • the first frame of image is obtained from these 3 frames of images, and abnormal target detection is performed on the first frame of image. If an abnormal target is detected in the first frame of image, it can be determined that there is an abnormal target in the detection area and stop continuing the detection; if it is detected that there is no abnormal target in the first frame of image, then continue to obtain the second frame from the 3 frames of image Image, the second frame of image for abnormal target detection. If an abnormal target is detected in the second frame of image, it can be determined that there is an abnormal target in the detection area and stop continuing the detection; if it is detected that there is no abnormal target in the second frame of image, continue to obtain the third Frame image, and perform abnormal target detection on the third frame image. If an abnormal target is detected in the third frame of image, it can be determined that there is an abnormal target in the detection area; if it is detected that there is no abnormal target in the third frame of image, it can be determined that there is no abnormal target in the detection area.
  • performing abnormal target detection on N frames of images to obtain the second detection result may include the following implementation steps:
  • the background image model of the detection area is the reference image feature of the background image of the detection area.
  • foreground detection can be performed on the third image by comparing the difference between the background image and the third image when the detection area has no abnormal target.
  • the reference image feature of the third image can be obtained by feature extraction of the third image, and the feature residual between the reference image feature of the third image and the reference image feature of the background image of the detection area can be calculated.
  • the characteristic residual is greater than or equal to the characteristic residual threshold, it is determined that there is a foreground target in the third image.
  • the feature residual is used to indicate the change between the third image and the background image of the detection area.
  • the feature residual threshold can be obtained through big data statistics.
  • the background image model of the detection area is the reference image feature of the background image of the detection area
  • the reference image feature of the third image can be extracted, and the reference image feature of the third image can be compared with the background image of the detection area.
  • the features of the reference image are compared, and the foreground detection of the third image is realized according to the comparison result. That is, calculate the feature residual between the reference image feature of the third image and the reference image feature of the background image of the detection area.
  • the feature residual is greater than or equal to the feature residual threshold, it is determined that there is a foreground target in the third image.
  • the residual is less than the characteristic residual threshold, it is determined that there is no foreground target in the third image.
  • the reference image feature of the third image can be determined by the following formula:
  • L refers to the reference image feature of the third image
  • I p refers to the binary image data corresponding to the p-th pixel of the third image
  • S refers to the coordinate position of the p-th pixel
  • I c refers to the third The binary image data corresponding to the central pixel of the image.
  • the coordinate position of the center pixel of the third image corresponding to I c may be expressed as (x c , y c ).
  • the third image includes 100 pixels
  • the binary image data corresponding to each of the 100 pixels is combined with the binary image corresponding to the central pixel of the third image
  • the data are subtracted separately, and summed according to the above formula, the reference image feature of the third image can be obtained.
  • the feature residual threshold is 5
  • the reference image feature of the third image is 10
  • the reference image feature of the background image of the detection area is 2, it can be determined that the feature residual is 8, which is greater than the feature residual threshold, and then the third image can be determined There are prospects in the target.
  • the foreground target refers to the target of the background image that does not belong to the detection area, that is, the foreign target of the detection area. Since the foreground target in the detection area may be any target such as human body, vehicle, small animal, etc., and not necessarily the abnormal target to be detected, in order to avoid detection errors, it is necessary to further determine whether the foreground target is an abnormal target.
  • the width and height of the foreground object can be determined according to the coordinates of the pixels in the third image that are significantly different from the binary image data of the background image of the detection area.
  • the operation of determining whether the foreground target is an abnormal target according to the height and width of the foreground target may include the following three implementation methods:
  • the first method is to calculate the ratio between the width and height of the foreground target to obtain the aspect ratio of the foreground target; if the aspect ratio of the foreground target is within the range of the reference aspect ratio, the foreground target is determined to be an abnormal target.
  • the reference aspect ratio range is the reference aspect ratio range corresponding to the abnormal target, which can be obtained through statistics of a large number of abnormal target images.
  • the abnormal target is a person, according to a large number of human body images, the height and width of the person in different poses, the height and width of different people in the same pose, etc. can be counted, and the corresponding reference for the person can be obtained based on these data.
  • Aspect ratio range is the reference aspect ratio range corresponding to the abnormal target, which can be obtained through statistics of a large number of abnormal target images.
  • FIG. 6, is an image of a person standing up, where the width of the human body is W, the height is H, and the aspect ratio is W/H.
  • Fig. 7, which is an image of a person lying down, where the width of the human body is W 1 , the height is H 1 , and the aspect ratio is W 1 /H 1 .
  • the foreground target can be determined to be an abnormal target, and when the aspect ratio of the current target is not within the range of the reference aspect ratio, it can be determined that the foreground target is not abnormal. aims.
  • the reference aspect ratio is 1:5 to 5:1
  • the width of the foreground object is 0.6m
  • the height is 1.8m
  • the aspect ratio is 1:3
  • the second implementation method determine the height and width of the third image, calculate the ratio between the width and height of the foreground object, and obtain the aspect ratio of the foreground object; calculate the first between the width of the foreground object and the width of the third image A ratio; calculate the second ratio between the height of the foreground object and the height of the third image; if the aspect ratio of the foreground object is within the range of the reference aspect ratio, the first ratio is within the range of the first reference ratio and the second ratio Within the range of the second reference ratio, the foreground target is determined to be an abnormal target.
  • both the first reference ratio range and the second reference ratio range can be obtained based on big data statistics.
  • the foreground target is determined to be an abnormal target, but if the width of the foreground target is When the aspect ratio is not within the range of the reference aspect ratio or the first ratio is not within the range of the first reference ratio or the second ratio is not within the range of the second reference ratio, it can be determined that the foreground target is not an abnormal target
  • the reference aspect ratio range is 1:5 to 5:1
  • the first reference ratio range is 1:20 to 1:5
  • the second reference ratio range is 1:20 to 1:2
  • the foreground target The width is 0.6m
  • the height is 1.8m
  • the width of the third image is 6m
  • the height is 4m.
  • the aspect ratio of the foreground object can be obtained as 1:3, the first ratio is 1:10, and the second ratio is 9:20 , It can be determined that the aspect ratio of the foreground target is within the reference aspect ratio range, the first ratio is within the first reference ratio range, and the second ratio is within the second reference ratio range, and then it can be determined that the foreground target is an abnormal target.
  • the second ratio is 3:5
  • the aspect ratio of the foreground object is within the range of the reference aspect ratio
  • the first ratio is in the first The reference ratio range
  • the second ratio is not within the second reference ratio range
  • the third implementation method determine the height and width of the third image, calculate the ratio between the width and height of the foreground object, and obtain the aspect ratio of the foreground object; determine the third ratio, the third ratio is the width of the foreground object and the first The ratio between the widths of the three images, or the ratio between the height of the foreground object and the height of the third image; if the aspect ratio of the foreground object is within the range of the reference aspect ratio, and the third ratio is in the third reference ratio Within the range, the foreground target is determined to be an abnormal target.
  • the third reference ratio range is the above-mentioned first reference ratio range
  • the third ratio is the height of the foreground object and the third image
  • the third reference ratio range is the above-mentioned second reference ratio range.
  • the foreground target is an abnormal target
  • the acquisition of images of the detection area can be stopped, and abnormal target detection can be performed on the acquired images, so that the power consumption of the detection device can be reduced .
  • the detection of images after the target image can also be continued to detect the abnormal target The position change or the posture change.
  • Step 404 When both the first detection result and the second detection result indicate that there is an abnormal target in the detection area, it is determined that there is an abnormal target in the detection area.
  • this application makes a judgment based on the first detection result and the second detection result.
  • the first detection result indicates that there is an abnormal target in the detection area
  • the detection based on the infrared signal determines that there is an abnormal target in the detection area
  • the verification is performed according to the second detection result.
  • the second detection result also indicates that there is an abnormal target in the detection area
  • it is a target it is explained that an abnormal target exists in the detection area based on the detection of N frames of images. In this case, it can be determined that there is an abnormal target in the detection area.
  • the first detection result can be determined based on the infrared signal detected in the detection area, and N frames of images of the detection area can be obtained, and abnormal target detection is performed on the N frames of images to obtain the second detection result.
  • the detection result and the second detection result both indicate that there is an abnormal target in the detection area, it is determined that there is an abnormal target in the detection area.
  • infrared detection and image detection can be combined to determine whether there are abnormal targets in the detection area, and the accuracy of the detection results is improved.
  • image detection is not susceptible to interference from external factors, the infrared sensor and microwave detector can be reduced. Detection errors caused by interference from external factors have further improved the accuracy of detection results.
  • Fig. 8 is a schematic diagram showing an abnormal target detection device according to an exemplary embodiment.
  • the abnormal target detection device can be implemented by software, hardware or a combination of both. Referring to Figure 8, the device may include:
  • the first detection module 801 is configured to determine a first detection result based on the infrared signal detected in the detection area, and the first detection result is used to indicate whether there is an abnormal target in the detection area;
  • the second detection module 802 is used to obtain N frames of images in the detection area, and perform abnormal target detection on the N frames of images to obtain a second detection result, which is used to indicate whether there is an abnormal target in the detection area, and N is a positive integer ;
  • the determining module 803 is configured to determine that there is an abnormal target in the detection area when the first detection result and the second detection result both indicate that there is an abnormal target in the detection area.
  • the second detection module 802 is used to:
  • the second detection module 802 is used to:
  • the second detection module 802 is used to:
  • first reference time period collect an optical image of the detection area through an image sensor every second reference time period, where the second reference time period is less than or equal to the first reference time period;
  • the first image obtained within the first reference time length is determined as the N frames of images.
  • the second detection module 802 is used to:
  • the detection area is photographed by a camera to obtain a second image of the detection area, and the second image is a visible image;
  • the second image captured within the first reference time period is determined to be the N frames of images.
  • the second detection module 802 is further configured to:
  • the first detection result indicates that the abnormal target exists in the detection area
  • perform image collection on the detection area and trigger the first detection module to stop passing through the infrared sensor to detect the detection area;
  • the first detection module is triggered to continue to pass the infrared sensor to detect the detection area.
  • the second detection module 802 is used to:
  • the abnormal target is detected in one of the N frames of images, it is determined that the abnormal target exists in the detection area;
  • the second detection module 802 is used to:
  • the second detection module 802 includes:
  • the detection unit is configured to perform abnormal target detection on the N frames of images, for any third image to be detected in the N frames of images, based on the background image model of the detection area, to Perform foreground detection on three images, and the background image model is used to indicate the background image of the detection area;
  • the first determining unit is configured to determine the height and width of the foreground target if it is detected that there is a foreground target in the third image that is different from the background image of the detection area;
  • the second determining unit is configured to determine that the abnormal target exists in the third image if the foreground target is determined to be the abnormal target based on the height and width of the foreground target.
  • the background image model of the detection area is a reference image feature of the background image of the detection area
  • the detection unit is used for:
  • the characteristic residual is greater than or equal to the characteristic residual threshold, it is determined that the foreground target exists in the third image.
  • the detection unit is used to:
  • the reference image feature of the third image is determined by the following formula:
  • L refers to the reference image feature of the third image
  • I p refers to the binary image data corresponding to the p-th pixel of the third image
  • S refers to the coordinate position of the p-th pixel
  • I c is Refers to the binary image data corresponding to the central pixel of the third image.
  • the detection unit is used to:
  • the foreground target is determined to be the abnormal target.
  • the detection unit is used to:
  • the determining the foreground target as the abnormal target based on the height and width of the foreground target includes:
  • the foreground target is determined to be the abnormal target.
  • the first detection module 801 is used to:
  • the infrared sensor Detecting an infrared signal in the detection area by an infrared sensor, and converting the detected infrared signal into an electrical signal, the infrared sensor at least including an infrared probe and at least two Fresnel lenses;
  • the first detection result is determined.
  • the first detection module 801 is used to:
  • the first detection module 801 is used to:
  • the vibration amplitude of the electrical signal is within the reference amplitude range and the vibration frequency of the electrical signal is within the reference frequency range, it is determined that the abnormal target exists in the detection area.
  • the first detection result can be determined based on the infrared signal detected in the detection area, and N frames of images of the detection area can be obtained, and abnormal target detection is performed on the N frames of images to obtain the second detection result.
  • the detection result and the second detection result both indicate that there is an abnormal target in the detection area, it is determined that there is an abnormal target in the detection area.
  • infrared detection and image detection can be combined to determine whether there are abnormal targets in the detection area, which improves the accuracy of the detection results.
  • image detection is not susceptible to interference from external factors, it can reduce the impact of infrared sensors and microwave detectors. Detection errors caused by interference from external factors have further improved the accuracy of detection results.
  • the detection device for abnormal targets only uses the division of the above functional modules to illustrate when detecting abnormal targets.
  • the above functions can be allocated to different functional modules according to needs. Complete, that is, divide the internal structure of the device into different functional modules to complete all or part of the functions described above.
  • the abnormal target detection device provided in the foregoing embodiment and the abnormal target detection method embodiment belong to the same concept. For the specific implementation process, please refer to the method embodiment, which will not be repeated here.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Human Computer Interaction (AREA)
  • Remote Sensing (AREA)
  • Geometry (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

一种异常目标的检测方法、设备及存储介质，属于智能安防领域。所述方法包括：基于检测区域内检测的红外信号，确定第一检测结果，第一检测结果用于指示检测区域内是否存在异常目标（301）；获取检测区域的N帧图像，对N帧图像进行异常目标检测，得到第二检测结果，第二检测结果用于指示检测区域内是否存在异常目标，N为正整数（302）；当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标（303）。如此，可以将红外检测与图像检测进行结合，来确定检测区域是否存在异常目标，由于图像检测不易受外界因素干扰，因此可以降低检测失误，提高了检测异常目标的准确率。

Description

异常目标的检测方法、设备及存储介质
本申请要求于2019年06月12日提交的申请号为201910508062.6、发明名称为“异常目标的检测方法、装置及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及智能安防领域，特别涉及一种异常目标的检测方法、设备及存储介质。
背景技术
在当今社会中，为了保证生命财产安全，人们会在一些地方安装检测设备，检测设备可以用于检测其对应的检测区域内是否有异常目标入侵，即其对应的检测区域内是否存在异常目标，以便在检测到存在异常目标时，进行报警提醒。其中，异常目标通常为人或车辆等。
目前，为了提高检测的准确度，检测设备可以基于红外信号和微波信号对检测区域进行检测，只有当基于红外信号检测后确定检测区域存在异常目标，且基于微波信号检测后也确定检测区域存在异常目标时，才会最终确定检测区域内存在异常目标。具体地，检测设备可以包括红外传感器、微波探测器和处理器。由于在检测区域内的异常目标发出的红外信号可以被红外传感器检测到，因此当红外传感器检测到红外信号时，说明检测区域中可能存在异常目标，因此可以将检测到的红外信号转化为电信号发送给处理器，由处理器根据接收的电信号进一步判断检测区域内是否存在异常目标。微波探测器可以不断向检测区域发送微波信号，也不断接收检测区域反射回来的微波信号，当微波探测器接收到的微波信号的频率与发送出去的微波信号的频率有差异时，说明检测区域中可能存在异常目标，因此需要处理器根据该差异进一步判断检测区域内是否存在异常目标。当处理器通过红外传感器和微波探测器均检测到检测区域内存在异常目标时，即可确定检测区域内存在异常目标。
但是上述方法仍然会受一些外在因素的干扰，导致检测结果不准确。例如， 一些飞虫等不属于异常目标的小动物可能会爬到检测设备上，对检测到的红外信号和微波信号造成干扰，导致检测失误。或者，晴天、大风等天气情况下产生的热气流，也会对检测到的红外信号和微波信号造成干扰，导致检测失误。
发明内容
本申请实施例提供了一种异常目标的检测方法、装置及存储介质，可以解决现有技术中检测设备容易受外在因素干扰，导致检测失误的问题。所述技术方案如下：
一方面，提供了一种异常目标的检测方法，所述方法包括：
基于检测区域内检测的红外信号，确定第一检测结果，所述第一检测结果用于指示所述检测区域内是否存在异常目标；
获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测，得到第二检测结果，所述第二检测结果用于指示所述检测区域内是否存在所述异常目标，所述N为正整数；
当所述第一检测结果和所述第二检测结果均指示所述检测区域内存在所述异常目标时，确定所述检测区域内存在所述异常目标。
在本申请一种可能的实现方式中，所述获取所述检测区域的N帧图像，包括：
基于检测区域内检测的红外信号，确定第一检测结果的同时，对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，所述获取所述检测区域的N帧图像，包括：
基于检测区域内检测的红外信号，确定第一检测结果之后，当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，所述对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像，包括：
在所述第一参考时长内，每隔第二参考时长，通过图像传感器采集所述检测区域的光学图像，所述第二参考时长小于或等于所述第一参考时长；
对所述光学图像进行光电转换，得到所述光学图像对应的电信号；
对所述光学图像对应的电信号进行数字转换，得到二进制图像数据；
对所述二进制图像数据进行像素转换，得到所述检测区域的第一图像，所述第一图像为不可视图像；
将所述第一参考时长内得到的所述第一图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，所述对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像，包括：
在所述第一参考时长内，每隔第二参考时长，通过摄像头对所述检测区域进行拍摄，得到所述检测区域的第二图像，所述第二图像为可视图像；
将所述第一参考时长内拍摄得到的所述第二图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，包括：
当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，以及停止基于检测区域内检测的红外信号，确定第一检测结果；
所述对所述N帧图像进行异常目标检测，得到第二检测结果之后，还包括：
若所述第二检测结果指示所述检测区域内不存在所述异常目标，则继续基于检测区域内检测的红外信号，确定第一检测结果。
在本申请一种可能的实现方式中，所述对所述N帧图像进行异常目标检测，得到第二检测结果，包括：
对所述N帧图像进行异常目标检测；
若检测到所述N帧图像中的一帧图像中存在所述异常目标，则确定所述检测区域内存在所述异常目标；
若检测到所述N帧图像中均不存在所述异常目标，则确定所述检测区域不存在所述异常目标。
在本申请一种可能的实现方式中，所述对所述N帧图像进行异常目标检测，包括：
对所述N帧图像并行进行异常目标检测；或者，
将所述N帧图像中的第一帧图像确定为待检测的第三图像，对所述第三图像进行异常目标检测；若检测到所述第三图像中存在所述异常目标，则确定所述N帧图像中的一帧图像中存在所述异常目标；若检测到所述第三图像中不存在所述异常目标，则将所述第三图像中的下一帧图像确定为待检测的第三图像，并重复执行对所述第三图像进行异常目标检测的步骤，直至检测到所述N帧图 像中的一帧图像中存在所述异常目标或者所述N帧图像中均不存在所述异常目标为止。
在本申请一种可能的实现方式中，所述方法还包括：
在对所述N帧图像进行异常目标检测的过程中，对于所述N帧图像中的任一帧待检测的第三图像，基于所述检测区域的背景图像模型，对所述第三图像进行前景检测，所述背景图像模型用于指示所述检测区域的背景图像；
若检测到所述第三图像中存在与所述检测区域的背景图像不同的前景目标，则确定所述前景目标的高度和宽度；
若基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标，则确定所述第三图像中存在所述异常目标。
在本申请一种可能的实现方式中，所述检测区域的背景图像模型为所述检测区域的背景图像的参考图像特征；
所述基于所述检测区域的背景图像模型，对所述第三图像进行前景检测，包括：
对所述第三图像进行特征提取，得到所述第三图像的参考图像特征；
计算所述第三图像的参考图像特征与所述检测区域的背景图像的参考图像特征之间的特征残差；
当所述特征残差大于或等于特征残差阈值时，确定所述第三图像中存在所述前景目标。
在本申请一种可能的实现方式中，所述对所述第三图像进行特征提取，得到所述第三图像的参考图像特征，包括：
基于所述第三图像对应的二进制图像数据，通过以下公式确定所述第三图像的参考图像特征：
Figure PCTCN2019124327-appb-000001
其中，L是指所述第三图像的参考图像特征，I p是指所述第三图像的第p个像素点对应的二进制图像数据，S是指所述第p个像素点的坐标位置，I c是指所述第三图像的中心像素点对应的二进制图像数据。
在本申请一种可能的实现方式中，所述确定所述前景目标的高度和宽度之后，还包括：
计算所述前景目标的宽度与高度之间的比值，得到所述前景目标的宽高比；
若所述前景目标的宽高比在参考宽高比范围内，则确定所述前景目标为所述异常目标。
在本申请一种可能的实现方式中，所述基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标之前，还包括：
确定所述第三图像的高度和宽度；
所述基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标，包括：
计算所述前景目标的宽度与高度之间的比值，得到所述前景目标的宽高比；
计算所述前景目标的宽度与所述第三图像的宽度之间的第一比值；
计算所述前景目标的高度与所述第三图像的高度之间的第二比值；
若所述前景目标的宽高比在参考宽高比范围内、所述第一比值在第一参考比值范围内且所述第二比值在第二参考比值范围内，则确定所述前景目标为所述异常目标。
在本申请一种可能的实现方式中，所述基于检测区域内检测的红外信号，确定第一检测结果，包括：
通过红外传感器对所述检测区域内的红外信号进行检测，将检测的红外信号转换为电信号，所述红外传感器至少包括红外探头和至少两个菲涅尔透镜；
根据所述电信号的振动幅值，确定所述第一检测结果。
在本申请一种可能的实现方式中，所述根据所述电信号的振动幅值，确定所述第一检测结果，包括：
当所述电信号的振动幅值在参考幅值范围内时，确定所述检测区域内存在所述异常目标。
在本申请一种可能的实现方式中，所述根据所述电信号的振动幅值，确定所述第一检测结果，包括：
当所述电信号的振动幅值在参考幅值范围内，且所述电信号的振动频率在参考频率范围内时，确定所述检测区域内存在所述异常目标。
另一方面，提供了一种异常目标的检测装置，所述装置包括：
第一检测模块，用于基于检测区域内检测的红外信号，确定第一检测结果，所述第一检测结果用于指示所述检测区域内是否存在异常目标；
第二检测模块，用于获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测，得到第二检测结果，所述第二检测结果用于指示所述检测区域 内是否存在所述异常目标，所述N为正整数；
确定模块，用于当所述第一检测结果和所述第二检测结果均指示所述检测区域内存在所述异常目标时，确定所述检测区域内存在所述异常目标。
在本申请一种可能的实现方式中，所述第二检测模块用于：
基于检测区域内检测的红外信号，确定第一检测结果的同时，对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像。在本申请一种可能的实现方式中，所述第二检测模块用于：
基于检测区域内检测的红外信号，确定第一检测结果之后，当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，所述第二检测模块用于：
在所述第一参考时长内，每隔第二参考时长，通过图像传感器采集所述检测区域的光学图像，所述第二参考时长小于或等于所述第一参考时长；
对所述光学图像进行光电转换，得到所述光学图像对应的电信号；
对所述光学图像对应的电信号进行数字转换，得到二进制图像数据；
对所述二进制图像数据进行像素转换，得到所述检测区域的第一图像，所述第一图像为不可视图像；
将所述第一参考时长内得到的所述第一图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，所述第二检测模块用于：
在所述第一参考时长内，每隔第二参考时长，通过摄像头对所述检测区域进行拍摄，得到所述检测区域的第二图像，所述第二图像为可视图像；
将所述第一参考时长内拍摄得到的所述第二图像，确定为所述N帧图像。
在本申请一种可能的实现方式中，所述第二检测模块还用于：
当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，以及触发所述第一检测模块停止通过红外传感器，对所述检测区域进行检测；
若所述第二检测结果指示所述检测区域内不存在所述异常目标，则触发所述第一检测模块继续通过所述红外传感器，对所述检测区域进行检测。
在本申请一种可能的实现方式中，所述第二检测模块用于：
对所述N帧图像进行异常目标检测；
若检测到所述N帧图像中的一帧图像中存在所述异常目标，则确定所述检 测区域内存在所述异常目标；
若检测到所述N帧图像中均不存在所述异常目标，则确定所述检测区域不存在所述异常目标。
在本申请一种可能的实现方式中，所述第二检测模块用于：
对所述N帧图像并行进行异常目标检测；或者，
将所述N帧图像中的第一帧图像确定为待检测的第三图像，对所述第三图像进行异常目标检测；若检测到所述第三图像中存在所述异常目标，则确定所述N帧图像中的一帧图像中存在所述异常目标；若检测到所述第三图像中不存在所述异常目标，则将所述第三图像中的下一帧图像确定为待检测的第三图像，并重复执行对所述第三图像进行异常目标检测的步骤，直至检测到所述N帧图像中的一帧图像中存在所述异常目标或者所述N帧图像中均不存在所述异常目标为止。
在本申请一种可能的实现方式中，所述第二检测模块包括：
检测单元，用于在对所述N帧图像进行异常目标检测的过程中，对于所述N帧图像中的任一帧待检测的第三图像，基于所述检测区域的背景图像模型，对所述第三图像进行前景检测，所述背景图像模型用于指示所述检测区域的背景图像；
第一确定单元，用于若检测到所述第三图像中存在与所述检测区域的背景图像不同的前景目标，则确定所述前景目标的高度和宽度；
第二确定单元，用于若基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标，则确定所述第三图像中存在所述异常目标。
在本申请一种可能的实现方式中，所述检测区域的背景图像模型为所述检测区域的背景图像的参考图像特征；
所述检测单元用于：
对所述第三图像进行特征提取，得到所述第三图像的参考图像特征；
计算所述第三图像的参考图像特征与所述检测区域的背景图像的参考图像特征之间的特征残差；
当所述特征残差大于或等于特征残差阈值时，确定所述第三图像中存在所述前景目标。
在本申请一种可能的实现方式中，所述检测单元用于：
基于所述第三图像对应的二进制图像数据，通过以下公式确定所述第三图 像的参考图像特征：
Figure PCTCN2019124327-appb-000002
其中，L是指所述第三图像的参考图像特征，I p是指所述第三图像的第p个像素点对应的二进制图像数据，S是指所述第p个像素点的坐标位置，I c是指所述第三图像的中心像素点对应的二进制图像数据。
在本申请一种可能的实现方式中，所述检测单元用于：
计算所述前景目标的宽度与高度之间的比值，得到所述前景目标的宽高比；
若所述前景目标的宽高比在参考宽高比范围内，则确定所述前景目标为所述异常目标。
在本申请一种可能的实现方式中，所述检测单元用于：
确定所述第三图像的高度和宽度；
计算所述前景目标的宽度与高度之间的比值，得到所述前景目标的宽高比；
计算所述前景目标的宽度与所述第三图像的宽度之间的第一比值；
计算所述前景目标的高度与所述第三图像的高度之间的第二比值；
若所述前景目标的宽高比在参考宽高比范围内、所述第一比值在第一参考比值范围内且所述第二比值在第二参考比值范围内，则确定所述前景目标为所述异常目标。
在本申请一种可能的实现方式中，所述第一检测模块用于：
通过红外传感器对所述检测区域内的红外信号进行检测，将检测的红外信号转换为电信号，所述红外传感器至少包括红外探头和至少两个菲涅尔透镜；
根据所述电信号的振动幅值，确定所述第一检测结果。
在本申请一种可能的实现方式中，所述第一检测模块用于：
当所述电信号的振动幅值在参考幅值范围内时，确定所述检测区域内存在所述异常目标。
在本申请一种可能的实现方式中，所述第一检测模块用于：
当所述电信号的振动幅值在参考幅值范围内，且所述电信号的振动频率在参考频率范围内时，确定所述检测区域内存在所述异常目标。
另一方面，提供了一种检测设备，所述检测设备包括红外检测单元、图像处理单元和处理器，所述红外检测单元至少包括红外传感器；
所述红外检测单元，用于通过所述红外传感器，对所述检测区域内的红外 信号进行检测，将检测信号发送给所述处理器；
所述图像处理单元，用于获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测，得到第二检测结果，将所述第二检测结果发送给所述处理器，所述第二检测结果用于指示所述检测区域内是否存在所述异常目标，所述N为正整数；
所述处理器用于根据所述检测信号确定第一检测结果，当所述第一检测结果和所述第二检测结果均指示所述检测区域内存在所述异常目标时，确定所述检测区域内存在所述异常目标，所述第一检测结果用于指示所述检测区域内是否存在所述异常目标。
另一方面，提供一种计算机可读存储介质，所述计算机可读存储介质上存储有指令，所述指令被处理器执行时实现上述一方面所述的任一项方法的步骤。
另一方面，提供了一种包含指令的计算机程序产品，当其在计算机上运行时，使得计算机执行上述一方面所述的任一项方法的步骤。
本申请实施例提供的技术方案带来的有益效果是：
本申请实施例中，可以基于检测区域内检测的红外信号，确定第一检测结果，以及获取检测区域的N帧图像，对该N帧图像进行异常目标检测，得到第二检测结果，当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。如此，可以将红外检测与图像检测相结合，来确定检测区域是否存在异常目标，提高了检测结果的准确率，而且，由于图像检测不易受外界因素的干扰，可以降低红外传感器和微波探测器因外界因素干扰导致的检测失误，因此进一步提高了检测结果的准确率。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1是根据一示例性实施例示出的一种实施环境的示意图；
图2是根据另一示例性实施例示出的一种检测设备的示意图；
图3是根据一示例性实施例示出的一种异常目标的检测方法的流程图；
图4是根据另一示例性实施例示出的一种异常目标的检测方法的流程图；
图5是根据一示例性实施例示出的一种人体在检测区域内走动时产生的电信号的波形图；
图6是根据一示例性实施例示出的一种人体直立时的示意图；
图7是根据一示例性实施例示出的一种人体卧倒时的示意图；
图8是根据一示例性实施例示出的一种异常目标的检测装置的结构示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚，下面将结合附图对本申请实施方式作进一步地详细描述。
在对本申请实施例提供的异常目标的检测方法进行详细介绍之前，先对本申请实施例涉及的应用场景和实施环境予以说明。
首先，对本申请实施例涉及的应用场景进行简单介绍。
本申请实施例提供的异常目标的检测方法可以应用于安防监控或防盗报警等场景中。例如，在防盗报警场景中，可以在需要防盗的区域安装检测设备，通过该检测设备检测该区域是否有人，如果检测到有人，则进行报警提醒，如此可以实现当该区域有人入侵时，进行防盗报警。
接下来，对本申请实施例涉及的实施环境进行简单介绍。
请参考图1，该图1是根据一示例性实施例示出的一种实施环境的示意图，该实施环境包括检测设备，该检测设备可以包括红外检测单元101、图像处理单元102和处理器103。红外检测单元101和图像处理单元102均可以与处理器103进行通信连接。
其中，红外检测单元101至少包括红外传感器，用于通过红外传感器对检测区域进行检测，并将检测信号发送给处理器103。图像处理单元102用于采集检测区域的N帧图像，以及对N帧图像进行异常目标检测，得到第二检测结果，将第二检测结果发送给处理器103。处理器103用于根据红外检测单元101发送的检测信号确定第一检测结果，根据第一检测结果和第二检测结果，确定检测区域是否存在异常目标。其中，第一检测结果和第二检测结果均用于指示检测区域是否存在异常目标。
或者，图像处理单元102用于采集检测区域的N帧图像，并将采集的N帧图像发送给处理器103。处理器103用于根据红外检测单元101发送的检测信号 确定第一检测结果，对图像处理单元102发送的N帧图像进行异常目标检测得到第二检测结果，然后根据第一检测结果和第二检测结果，确定检测区域是否存在异常目标。
或者，红外检测单元101用于通过红外传感器对检测区域进行检测，得到第一检测结果，将第一检测结果发送给处理器103。图像处理单元102用于采集检测区域的N帧图像，以及对N帧图像进行异常目标检测，得到第二检测结果，将第二检测结果发送给处理器103。处理器103用于根据第一检测结果和第二检测结果，确定检测区域是否存在异常目标。
红外传感器的探测范围包括检测设备对应的检测区域，用于接收检测区域发出的红外信号，将红外信号转换为电信号。作为一种示例，本申请实施例使用的红外传感器可以为红外主动传感器或PIR(Passive Infrarde，红外被动)传感器。红外传感器的数量可以为一个或多个，本申请实施例对此不做限定。
作为一种示例，参见图2，该红外检测单元101还可以包括第一电源和红外传感器信号滤波-放大单元。
其中，红外传感器信号滤波-放大单元用于对红外传感器输出的电信号进行滤波和放大处理，第一电源用于为红外检测单元101供电。
作为一种示例，参见图2，该图像处理单元102可以包括图像传感器、图像像素转换单元、CV(Computer Vision，计算机视觉)引擎单元和第二电源。
其中，图像传感器用于采集检测区域的光学图像，并将光学图像转换成电信号，再将电信号转换为二进制图像数据。图像像素转换单元用于对二进制图像数据进行像素转换，得到机器可识别但人眼不能看到的不可视图像。CV引擎单元用于对不可视图像进行异常目标检测，以确定检测区域内是否存在异常目标。第二电源用于为图像处理单元102供电。进一步地，参见图2，图像处理单元102还可以包括图像预处理单元。图像预处理单元用于对图像传感器发送的二进制图像数据进行滤波等处理。
其中，图像传感器的探测范围包括检测设备对应的检测区域，用于将检测区域的光学图像转换为电信号。图像传感器的数量可以为一个或多个，图2仅是以图像处理单元102包括两个图像传感器为例，本申请实施例对图像传感器的数量不做限定。
作为另一种示例，该图像处理单元102可以包括摄像头、CV引擎单元和第二电源。摄像头的拍摄范围包括检测设备的检测区域，用于对检测区域进行图 像采集，得到人眼可以看到的可视图像。CV引擎单元用于对可视图像进行异常目标检测，以确定检测区域内是否存在异常目标。第二电源用于为图像处理单元102供电。
作为一种示例，可以在启动红外检测单元101的同时，启动图像处理单元102，也即是，可以同时对检测区域进行红外检测和图像检测。
作为另一种示例，可以先启动红外检测单元101，当通过红外检测单元101检测到检测区域存在异常目标时，再启动图像处理单元102，以通过图像处理单元102进一步进行图像检测。
作为又一种示例，也可以先启动红外检测单元101，当通过红外检测单元101检测到检测区域存在异常目标时，再启动图像处理单元102对检测区域进行检测，并关闭红外检测单元101。当通过图像处理单元102检测到检测区域不存在异常目标时，继续启动红外检测单元101。当通过图像处理单元102检测到检测区域存在异常目标时，图像处理单元102持续工作。如此，可以确保检测设备以较低功耗运行，进一步减小了检测设备的功耗。
在介绍完本申请实施例提供的应用场景和实施环境后，接下来对本申请实施例提供的异常目标的检测方法进行详细介绍。
图3是根据一示例性实施例示出的一种异常目标的检测方法的流程图，该方法可以应用于上述图1或图2所示的检测设备中，如图3所示，该方法可以包括如下几个步骤：
步骤301：基于检测区域内检测的红外信号，确定第一检测结果，第一检测结果用于指示检测区域内是否存在异常目标。
作为一种示例，通过红外传感器对检测区域内的红外信号进行检测，将检测的红外信号转换为电信号；根据该电信号的振动幅值，确定第一检测结果。
作为一种示例，红外传感器至少包括红外探头和至少两个菲涅尔透镜，红外传感器可以通过至少两个菲涅尔透镜对探测区域内的红外信号进行聚焦，通过红外探头，对两个菲涅尔透镜聚焦的红外信号进行感应，以及将感应的红外信号转换为电信号。
作为一种示例，当电信号的振动幅值在参考幅值范围内时，确定检测区域内存在异常目标。
作为另一种示例，当电信号的振动幅值在参考幅值范围内，且电信号的振 动频率在参考频率范围内时，确定检测区域内存在异常目标。
步骤302：获取检测区域的N帧图像，对N帧图像进行异常目标检测，得到第二检测结果，第二检测结果用于指示检测区域内是否存在异常目标，N为正整数。
作为一种示例，基于检测区域内检测的红外信号，确定第一检测结果的同时，对检测区域进行图像采集，将第一参考时长内采集到的图像，确定为N帧图像。
作为另一种示例，基于检测区域内检测的红外信号，确定第一检测结果之后，当第一检测结果指示检测区域内存在异常目标时，对检测区域进行图像采集，将第一参考时长内采集到的图像，确定为N帧图像。
作为一种示例，在第一参考时长内，每隔第二参考时长，通过图像传感器采集检测区域的光学图像，第二参考时长小于或等于第一参考时长；对光学图像进行光电转换，得到光学图像对应的电信号；对光学图像对应的电信号进行数字转换，得到二进制图像数据；对二进制图像数据进行像素转换，得到检测区域的第一图像，第一图像为不可视图像；将第一参考时长内得到的第一图像，确定为N帧图像。
作为另一种示例，在第一参考时长内，每隔第二参考时长，通过摄像头对检测区域进行拍摄，得到检测区域的第二图像，第二图像为可视图像；将第一参考时长内拍摄得到的第二图像，确定为N帧图像。
作为一种示例，当第一检测结果指示检测区域内存在异常目标时，对检测区域进行图像采集，以及停止基于检测区域内检测的红外信号，确定第一检测结果。
作为一种示例，若第二检测结果指示检测区域内不存在异常目标，则继续基于检测区域内检测的红外信号，确定第一检测结果。
作为一种示例，对N帧图像进行异常目标检测；若检测到N帧图像中的一帧图像中存在异常目标，则确定检测区域内存在异常目标；若检测到N帧图像中均不存在异常目标，则确定检测区域不存在异常目标。
作为一种示例，对N帧图像并行进行异常目标检测；或者，将N帧图像中的第一帧图像确定为待检测的第三图像，对第三图像进行异常目标检测；若检测到第三图像中存在异常目标，则确定N帧图像中的一帧图像中存在异常目标；若检测到第三图像中不存在异常目标，则将第三图像中的下一帧图像确定为待 检测的第三图像，并重复执行对第三图像进行异常目标检测的步骤，直至检测到N帧图像中的一帧图像中存在异常目标或者N帧图像中均不存在异常目标为止
作为一种示例，在对N帧图像进行异常目标检测的过程中，对于N帧图像中的任一帧待检测的第三图像，基于检测区域的背景图像模型，对第三图像进行前景检测，背景图像模型用于指示检测区域的背景图像；若检测到第三图像中存在与检测区域的背景图像不同的前景目标，则确定前景目标的高度和宽度；若基于前景目标的高度和宽度，确定前景目标为异常目标，则确定第三图像中存在异常目标。
作为一种示例，检测区域的背景图像模型为检测区域的背景图像的参考图像特征。对第三图像进行特征提取，得到第三图像的参考图像特征；计算第三图像的参考图像特征与检测区域的背景图像的参考图像特征之间的特征残差；当特征残差大于或等于特征残差阈值时，确定第三图像中存在前景目标。
作为一种示例，基于第三图像对应的二进制图像数据，通过以下公式确定第三图像的参考图像特征：
Figure PCTCN2019124327-appb-000003
其中，L是指第三图像的参考图像特征，I p是指第三图像的第p个像素点对应的二进制图像数据，S是指第p个像素点的坐标位置，I c是指第三图像的中心像素点对应的二进制图像数据。
作为一种示例，计算前景目标的宽度与高度之间的比值，得到前景目标的宽高比；若前景目标的宽高比在参考宽高比范围内，则确定前景目标为异常目标。
作为另一种示例，确定第三图像的高度和宽度；计算前景目标的宽度与高度之间的比值，得到前景目标的宽高比；计算前景目标的宽度与第三图像的宽度之间的第一比值；计算前景目标的高度与第三图像的高度之间的第二比值；若前景目标的宽高比在参考宽高比范围内、第一比值在第一参考比值范围内且第二比值在第二参考比值范围内，则确定前景目标为异常目标。
步骤303：当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。
当第一检测结果指示检测区域内存在异常目标，且第二检测结果也指示检 测区域内存在异常目标时，可以确定检测区域内存在异常目标。
在本申请实施例中，可以基于检测区域内检测的红外信号，确定第一检测结果，以及获取检测区域的N帧图像，对该N帧图像进行异常目标检测，得到第二检测结果，当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。如此，可以将红外检测与图像检测相结合，来确定检测区域是否存在异常目标，提高了检测结果的准确率，而且，由于图像检测不易受外界因素的干扰，可以降低红外传感器和微波探测器因外界因素干扰导致的检测失误，因此进一步提高了检测结果的准确率。
图4是根据另一示例性实施例示出的一种异常目标的检测方法的流程图，该方法可以应用于上述图1或图2所示的检测设备中，或应用于检测设备中的处理器中，如图4所示，该方法可以包括如下几个步骤：
步骤401：基于检测区域内检测的红外信号，确定第一检测结果，该第一检测结果用于指示检测区域内是否存在异常目标。
其中，异常目标为待检测的目标，可以由检测设备默认设置，也可以由用户设置，本申请实施例对此不做限定。示例性地，异常目标可以为人或车辆等。
作为一种示例，可以通过红外传感器对检测区域内的红外信号进行检测，将检测的红外信号转换为电信号，根据红外传感器输出的电信号确定第一检测结果。其中，红外传感器的探测范围可以包括检测区域。
作为一种示例，红外传感器至少包括红外探头和菲涅尔透镜，菲涅尔透镜用于对检测区域内的红外信号进行聚焦，红外探头用于对菲涅尔透镜聚焦的红外信号进行感应，以及将感应的红外信号转换为电信号。
其中，菲涅尔透镜表面由一系列锯齿型凹槽组成，每个凹槽都与相邻凹槽之间角度不同。菲涅尔透镜的数量可以为一个或多个。
作为另一种示例，红外传感器至少包括红外探头和至少两个菲涅尔透镜。通过红外传感器对检测区域进行检测时，可以通过至少两个菲涅尔透镜对检测区域内的红外信号进行聚焦，通过红外探头，对至少两个菲涅尔透镜聚焦的红外信号进行感应，以及将感应的红外信号转换为电信号。
通过至少两个菲涅尔透镜对探测区域内的红外信号进行聚焦，可以使红外传感器感应到更远距离的物体发送的红外信号，从而扩大了红外传感器的探测范围，进而扩大了检测区域。
作为一种示例，红外传感器包括两个菲涅尔透镜时，可以将两个菲涅尔透镜并排安装在红外探头上，通过调整两个菲涅尔透镜中每个菲涅尔透镜的凹槽与相邻凹槽之间的角度，可以使检测区域达到更远。
作为一种示例，菲涅尔透镜可以对特定波长的红外信号起到聚焦和划分视区的作用，将特定波长的红外信号聚焦到红外传感器上，将检测区域划分为明区和暗区，且明区和暗区呈交替状态。
其中，特定波长可以用于指示能够通过菲涅尔透镜的红外信号的波长，且特定波长一般为一个波长范围。
需要说明的是，对于进入检测区域的物体来说，当该物体以相同的状态处于明区和暗区时，红外传感器检测到的红外信号的强弱是不同的。
作为一种示例，红外探头可以为四元红外探头，内部有四个敏感元件，可以感应到通过菲涅尔透镜聚焦的红外信号，通过红外传感器将红外信号转换为电信号之后，还可以对电信号进行一些预处理，对预处理后的电信号进行分析，根据预处理后的电信号的振动幅值，确定第一检测结果。
示例性地，假设红外传感器为PIR传感器，该PIR传感器的探头为四元PIR传感器探头，内部有四个敏感元件，该四个敏感元件可以感应到菲涅尔透镜聚焦的红外信号，并且将该红外信号转换为电信号。在PIR传感器中，可以对电信号进行一些信号噪声过滤，主要过滤电源噪声和器件本身的电器噪声，在红外传感器连接的外部信号调理电路中，可以对该电信号继续进行滤波、放大等处理，对处理后的电信号进行分析，根据处理后的电信号的振动幅值，确定第一检测结果。
在一些实施例中，根据电信号的振动幅值确定第一检测结果可以包括如下两种实现方式。
第一种实现方式：当电信号的振动幅值在参考幅值范围内时，确定检测区域内存在异常目标。
其中，参考幅值范围为异常目标对应的参考幅值范围，可以通过对大量异常目标在不同状态下产生的电信号的振动幅值进行统计得到。示例性地，若异常目标为人，则可以对红外传感器检测到的同一个人在不同的行走频率下产生的电信号的振动幅值、不同的人在相同的行走频率下产生的电信号的振动幅值或相同的人在检测区域的不同位置处产生的电信号的振动幅值等进行统计，得到人对应的参考幅值范围。
也就是说，对处理后的电信号的振动幅值进行统计，当电信号的振动幅值在参考幅值范围内时，确定检测区域内存在异常目标。示例性地，假设参考幅值范围为0.1V～0.9V，电信号的振动幅值为0.2V，则可以确定检测区域内存在异常目标。
以异常目标为人体为例，若有人入侵至检测区域内，则红外传感器可以接收到人体发送的红外信号，并将红外信号转换为电信号。其中，由于人体体表的温度在37度左右，因此发出的红外信号的波长一般在9.6μm左右，相应地，转换的电信号的振动幅值也在一定的幅值范围内。因此，检测过程中，可以确定红外传感器输出的电信号的振动幅值是否在该幅值范围内，当电信号的振动幅值在该幅值范围内时，确定检测区域存在人体，即有人体入侵。
第二种实现方式：当电信号的振动幅值在参考幅值范围内，且电信号的振动频率在参考频率范围内时，确定检测区域内存在异常目标。
其中，参考频率范围为异常目标对应的参考频率范围，可以通过对大量异常目标在不同状态下产生的电信号的振动频率进行统计得到。示例性地，若异常目标为人，则可以对红外传感器检测到的同一个人在不同的行走频率下产生的电信号的振动频率、不同的人在相同的行走频率下产生的电信号的振动频率、相同的人在检测区域的不同位置处产生的电信号的振动频率等进行统计，得到人对应的参考频率范围。
也就是说，对处理后的电信号的振动幅值和振动频率分别进行统计，只有当电信号的振动幅值在参考幅值范围内，且电信号的振动频率在参考频率范围内时，才可以确定检测区域内存在异常目标。
示例性地，假设参考幅值范围为0.1V～1.2V，参考频率范围为0.5Hz～1Hz，假设有一段电信号，振动幅值分别为0.5V、0.2V、0.1V、0.3V、0.5V，振动频率为0.6Hz，可以确定该段电信号的振动幅值在参考幅值范围内，且该段电信号的振动频率在参考频率范围内，进而确定检测区域内存在异常目标。
作为一种示例，异常目标距离红外探头越近，其电信号的幅值越大，异常目标距离红外探头越远，其电信号的幅值越小；当电信号为上升沿时，可以确定异常目标是从暗区走到明区，当电信号为下降沿时，可以确定异常目标是从明区走到暗区。
进一步地，通过对大量的数据进行统计，可以确定出异常目标的不同的移动行为中每种移动行为产生的电信号的频率范围，根据处理后的电信号的振动 频率，可以确定异常目标进行了哪种移动行为。示例性地，电信号的振动频率在频率范围A内，可以确定异常目标进行了快速移动，电信号的振动频率在频率范围B内，可以确定异常目标进行了缓慢移动。
示例性地，参见图5，该图5为一个人从进入检测区域到离开检测区域的全过程中产生的电信号的波形图。其中，横坐标为时间，纵坐标为电压值，单位为0.1V。从图中可以看出，前30个点检测区域内没有人进入，后面波动的数据就是人走动时产生的电信号的电压值。波形中的上升沿可以表示这个人从暗区走到明区，下降沿可以表示这个人从明区走到暗区，电压的幅值的绝对值越大，可以表示这个人距离红外探头越近，电压的幅值的绝对值越小，可以表示这个人距离红外探头越远。
需要说明的是，本申请实施例是仅以红外传感器为PIR传感器为例进行说明的，在实施中，红外传感器还可以为主动式红外传感器，示例性地，主动式红外传感器包括红外发射器和红外接收器，红外发射器发射一束或多束经过调制过的红外光线投向红外接收器，当红外发射器发射的红外光线与红外接收器接收的红外光线的频率相同时，可以认为红外发射器与红外接收器之间没有异常目标，即检测区域内不存在异常目标；当红外发射器发射的红外光线与红外接收器接收的红外光线的频率不同时，可以认为红外发射器与红外接收器之间有异常目标，即检测区域内存在异常目标。
步骤402：获取检测区域的N帧图像。
其中，N为大于或等于1的整数。
在一些实施例中，获取检测区域的N帧图像可以包括如下两种实现方式。
第一种实现方式：基于检测区域内检测的红外信号，确定第一检测结果的同时，对检测区域进行图像采集，将第一参考时长内采集到的图像，确定为N帧图像。其中，第一参考时长可以根据需要进行调整，例如，第一参考时长可以为200ms。
也就是说，基于检测区域的红外信号进行检测的操作与获取检测区域的N帧图像的操作同时进行，可以一边对检测区域进行红外检测，一边对检测区域进行图像采集，并对采集的图像进行异常目标检测。
第二种实现方式：基于检测区域内检测的红外信号，确定第一检测结果之后，当第一检测结果指示检测区域内存在异常目标时，对检测区域进行图像采集，将第一参考时长内采集到的图像，确定为N帧图像。
也就是说，可以先基于检测区域的红外信号进行检测，当检测到检测区域内存在异常目标时，再开始对检测区域进行图像采集，以便对采集的图像进行异常目标检测。如此，可以减小检测设备的功耗。
作为一种示例，当第一检测结果指示检测区域内存在异常目标时，可以开始对检测区域进行图像采集，同时停止基于检测区域内检测的红外信号，确定第一检测结果，如此，可以进一步减小检测设备的功耗。
在一些实施例中，对检测区域进行图像采集，将第一参考时长内采集到的图像，确定为N帧图像，可以包括如下两种实现方式：
第一种实现方式：在第一参考时长内，每隔第二参考时长，通过图像传感器采集检测区域的光学图像，第二参考时长小于或等于第一参考时长；对光学图像进行光电转换，得到光学图像对应的电信号；对光学图像对应的电信号进行数字转换，得到二进制图像数据；对二进制图像数据进行像素转换，得到检测区域的第一图像；将第一参考时长内得到的第一图像，确定为N帧图像。
其中，第一图像为不可视图像，即机器可以识别但人眼看不到的图像。本申请中采集图像的方式与传统的摄像头不同，所采集的第一图像为不可视图像，如此，可以避免对图像数据的额外操作，从而简化了图像采集操作，降低了图像采集成本。而且，还可以避免泄露人像数据隐私。
也就是说，在第一参考时长内，每隔第二参考时长，通过图像传感器采集检测区域的光学图像，通过图像传感器的光电转换功能，将光学图像转换成对应的电信号，对该光学图像对应的电信号进行数字转换，得到二进制图像数据，对二进制图像数据进行预处理，将处理后的二进制图像数据进行像素转换，得到检测区域的第一图像，将该第一图像确定为N帧图像。
作为一种示例，检测区域内的光学图像可以被图像传感器的感光面检测到，图像传感器将感光面上的光学图像转换成与光学图像对应的电信号，由于一帧光学图像有多个像素点，通过图像传感器的光电转换功能，每个像素点可以输出一个电信号，即输出一个电压值，将这些电信号进行数字转换，转成二进制图像数据，由于图像处理只需要一帧图像中一部分比较重要的像素点，所以，可以根据一定的策略对二进制图像数据进行预处理，保留比较重要的一部分像素点对应的二进制图像数据，将预处理后的二进制图像数据进行像素转换，得到检测区域的第一图像，将该第一图像确定为N帧图像。
示例性地，假设第一参考时长为200ms，第二参考时长为40ms，图像传感 器可以采集到检测区域内的5帧光学图像。检测区域内的光学图像被图像传感器的感光面检测到，图像传感器将感光面上的光学图像转换成与光学图像成对应比例关系的电信号，假设一帧光学图像上有320×240个像素点，通过图像传感器的光电转换功能，可以得到320×240个电信号，将这些电信号进行数字转换，得到320×240个二进制图像数据，由于图像处理可能不需要这么多的图像数据，所以，可以根据一定的策略对这些二进制图像数据做初步的滤波和预设阈值处理，保留比较重要的一部分像素点对应的二进制图像数据，然后对这些二进制图像数据进行像素转换，可以得到检测区域的第一图像，将该第一图像确定为N帧图像。
第二种实现方式：在第一参考时长内，每隔第二参考时长，通过摄像头对检测区域进行拍摄，得到检测区域的第二图像，第二图像为可视图像；将第一参考时长内拍摄得到的第二图像，确定为N帧图像。
也就是说，在第一参考时长内，每隔第二参考时长，可以通过摄像头对检测区域进行拍摄，得到检测区域的第二图像，可以将第一参考时长内拍摄得到的第二图像确定为N帧图像。示例性地，假设第一参考时长为200ms，第二参考时长为40ms，摄像头可以拍摄到检测区域内的5帧图像，可以将这5帧图像确定为N帧图像。
步骤403：对N帧图像进行异常目标检测，得到第二检测结果，该第二检测结果用于指示检测区域内是否存在异常目标。
作为一种示例，当N帧图像为摄像头拍摄得到的第二图像时，还可以对N帧图像进行预处理，再对预处理后的N帧图像进行异常目标检测。示例的，预处理操作可以包括：对N帧图像的格式或大小等进行转换，将N帧图像中的每一帧图像进行分割、压缩处理等。
在一些实施例中，对N帧图像进行异常目标检测，得到第二检测结果的操作可以包括：对N帧图像进行异常目标检测，若检测到N帧图像中的一帧图像中存在异常目标，则确定检测区域内存在异常目标；若检测到N帧图像中均不存在异常目标，则确定检测区域内不存在异常目标。
作为一种示例，对N帧图像进行异常目标检测可以包括如下两种实现方式。
第一种实现方式：对N帧图像并行进行异常目标检测。
也就是说，同时对N帧图像中的每一帧图像进行异常目标检测，若检测到该N帧图像中的至少一帧图像中存在异常目标，可以确定检测区域内存在异常 目标；若检测到该N帧图像中均不存在异常目标，可以确定检测区域内不存在异常目标。
第二种实现方式：对N帧图像依次进行异常目标检测。
作为一种示例，可以先将N帧图像中的第一帧图像确定为待检测的第三图像，对第三图像进行异常目标检测；若检测到第三图像中存在异常目标，则确定N帧图像中的一帧图像中存在异常目标；若检测到第三图像中不存在异常目标，则将第三图像中的下一帧图像确定为待检测的第三图像，并重复执行对第三图像进行异常目标检测的步骤，直至检测到N帧图像中的一帧图像中存在异常目标或者N帧图像中均不存在异常目标为止。
示例性地，假设N为3，从这3帧图像中获取第一帧图像，对第一帧图像进行异常目标检测。若检测到该第一帧图像中存在异常目标，可以确定检测区域内存在异常目标，停止继续检测；若检测到第一帧图像中不存在异常目标，则从3帧图像中继续获取第二帧图像，对该第二帧图像进行异常目标检测。若检测到该第二帧图像中存在异常目标，可以确定检测区域内存在异常目标，停止继续检测；若检测到该第二帧图像中不存在异常目标，则从3帧图像中继续获取第三帧图像，对该第三帧图像进行异常目标检测。若检测到该第三帧图像中存在异常目标，可以确定检测区域内存在异常目标；若检测到该第三帧图像中不存在异常目标，可以确定检测区域内不存在异常目标。
在另一些实施例中，对N帧图像进行异常目标检测，得到第二检测结果可以包括如下实现步骤：
(1)在对N帧图像进行异常目标检测的过程中，对于N帧图像中的任一帧待检测的第三图像，基于检测区域的背景图像模型，对第三图像进行前景检测，背景图像模型用于指示检测区域的背景图像。
其中，检测区域的背景图像模型为检测区域的背景图像的参考图像特征。
也就是说，对于N帧图像中的任一帧待检测的第三图像，可以通过对比检测区域没有异常目标时的背景图像与该第三图像的差异，对该第三图像进行前景检测。
作为一种示例，可以通过对第三图像进行特征提取，得到第三图像的参考图像特征，计算第三图像的参考图像特征与检测区域的背景图像的参考图像特征之间的特征残差，当特征残差大于或等于特征残差阈值时，确定第三图像中存在前景目标。
其中，特征残差用于指示第三图像与检测区域的背景图像之间的变化情况。特征残差阈值可以通过大数据统计得到。
也就是说，由于检测区域的背景图像模型为检测区域的背景图像的参考图像特征，所以，可以提取第三图像的参考图像特征，将该第三图像的参考图像特征与检测区域的背景图像的参考图像特征进行对比，根据比较结果实现对第三图像进行前景检测。即计算第三图像的参考图像特征与检测区域的背景图像的参考图像特征之间的特征残差，当特征残差大于或等于特征残差阈值时，确定第三图像中存在前景目标，当特征残差小于特征残差阈值时，确定第三图像中不存在前景目标。
作为一种示例，对第三图像进行特征提取时，可以基于该第三图像对应的二进制图像数据，通过以下公式确定第三图像的参考图像特征：
Figure PCTCN2019124327-appb-000004
其中，L是指第三图像的参考图像特征，I p是指第三图像的第p个像素点对应的二进制图像数据，S是指第p个像素点的坐标位置，I c是指第三图像的中心像素点对应的二进制图像数据。作为一种示例，I c对应的第三图像的中心像素点的坐标位置可以表示为(x c,y c)。
示例性地，假设第三图像包括100个像素点，当p＝100时，将这100个像素点中的每个像素点对应的二进制图像数据与该第三图像的中心像素点对应的二进制图像数据分别相减，根据上述公式求和，可以得到第三图像的参考图像特征。假设特征残差阈值为5，第三图像的参考图像特征为10，检测区域的背景图像的参考图像特征为2，可以确定特征残差为8，大于特征残差阈值，进而可以确定第三图像中存在前景目标。
(2)若检测到第三图像中存在与检测区域的背景图像不同的前景目标，则确定前景目标的高度和宽度。
其中，前景目标是指不属于检测区域的背景图像的目标，即检测区域的外来目标。由于检测区域内存在的前景目标可能为人体、车辆、小动物等任何目标，不一定是待检测的异常目标，因此，为了避免检测失误，需要进一步确定该前景目标是否为异常目标。
作为一种示例，可以根据第三图像中与检测区域的背景图像的二进制图像数据差异较大的像素点的坐标，确定前景目标的宽度和高度。
确定前景目标的高度和宽度之后，可以根据前景目标的高度和宽度，确定前景目标是否为异常目标。作为一种示例，根据前景目标的高度和宽度，确定前景目标是否为异常目标的操作可以包括以下三种实现方式：
第一种实现方式：计算前景目标的宽度与高度之间的比值，得到前景目标的宽高比；若前景目标的宽高比在参考宽高比范围内，则确定前景目标为异常目标。
其中，参考宽高比范围为异常目标对应的参考宽高比范围，可以通过对大量异常目标图像统计得到。示例性地，若异常目标为人，则可以根据大量人体图像，统计人在不同姿态下的高度和宽度、统计不同人在相同姿态下的高度和宽度等等，根据这些数据，得到人对应的参考宽高比范围。
示例性地，参见图6，图6为人站立时的图像，其中，人体的宽度为W，高度为H，宽高比为W/H。参见图7，图7为人卧倒时的图像，其中，人体的宽度为W 1，高度为H 1，宽高比为W 1/H 1。
也就是说，确定前景目标的高度和宽度后，可以计算该前景目标的宽度和高度的比值，得到前景目标的宽高比，将前景目标的宽高比与大数据统计得到的参考宽高比范围进行对比，当前景目标的宽高比在参考宽高比范围内时，可以确定前景目标为异常目标，当前景目标的宽高比不在参考宽高比范围内时，可以确定前景目标不是异常目标。
示例性地，假设参考宽高比为1:5～5:1，前景目标的宽度为0.6m，高度为1.8m，宽高比为1:3，可以确定该前景目标的宽高比在参考宽高比范围内，进而确定该前景目标为异常目标。
第二种实现方式：确定第三图像的高度和宽度，计算前景目标的宽度与高度之间的比值，得到前景目标的宽高比；计算前景目标的宽度与第三图像的宽度之间的第一比值；计算前景目标的高度与第三图像的高度之间的第二比值；若前景目标的宽高比在参考宽高比范围内、第一比值在第一参考比值范围内且第二比值在第二参考比值范围内，则确定前景目标为异常目标。
其中，第一参考比值范围和第二参考比值范围均可以根据大数据统计得到。
也就是说，确定前景目标的高度和宽度后，可以计算该前景目标的宽度和高度的比值，得到前景目标的宽高比，将前景目标的宽高比与大数据统计得到的参考宽高比进行对比，并且确定第三图像的高度和宽度，计算前景目标的宽度与第三图像的宽度之间的第一比值；计算前景目标的高度与第三图像的高度 之间的第二比值；若前景目标的宽高比在参考宽高比范围内、第一比值在第一参考比值范围内且第二比值在第二参考比值范围内，则确定前景目标为异常目标，但若前景目标的宽高比不在参考宽高比范围内或第一比值不在第一参考比值范围内或第二比值不在第二参考比值范围内时，可以确定前景目标不是异常目标。
示例性地，假设参考宽高比范围为1:5～5:1，第一参考比值范围为1:20～1:5，第二参考比值范围为1:20～1:2，前景目标的宽度为0.6m，高度为1.8m，第三图像的宽度为6m，高度为4m，可以得到前景目标的宽高比为1:3，第一比值为1:10，第二比值为9:20，可以确定前景目标的宽高比在参考宽高比范围内、第一比值在第一参考比值范围内且第二比值在第二参考比值范围内，进而可以确定前景目标是异常目标。假设将上述举例中的第三图像的高度改成3m，其他数据不变，可以确定第二比值为3:5，前景目标的宽高比在参考宽高比范围内，第一比值在第一参考比值范围，但第二比值不在第二参考比值范围内，可以确定前景目标不是异常目标。
第三种实现方式：确定第三图像的高度和宽度，计算前景目标的宽度与高度之间的比值，得到前景目标的宽高比；确定第三比值，第三比值为前景目标的宽度与第三图像的宽度之间的比值，或者，前景目标的高度与第三图像的高度之间的比值；若前景目标的宽高比在参考宽高比范围内，且第三比值在第三参考比值范围内，则确定前景目标为异常目标。
其中，若第三比值为前景目标的宽度与第三图像的宽度之间的比值，则第三参考比值范围为上述第一参考比值范围，若第三比值为前景目标的高度与第三图像的高度之间的比值，则第三参考比值范围为上述第二参考比值范围。
(3)若基于前景目标的高度和宽度，确定前景目标为异常目标，则确定第三图像中存在异常目标。
当确定前景目标为异常目标时，可以确定第三图像中存在异常目标，进而确定检测区域内存在异常目标。
作为一种示例，对N帧图像进行异常目标检测，得到第二检测结果之后，若第二检测结果指示检测区域内不存在异常目标，则继续基于检测区域内检测的红外信号，对检测区域进行检测。
作为另一种示例，当继续基于检测区域内检测的红外信号对检测区域进行检测时，可以停止获取检测区域的图像，以及对获取的图像进行异常目标检测， 如此，可以降低检测设备的功耗。
作为一种示例，在对N帧图像进行异常目标检测的过程中，当检测到N帧图像中的目标图像中存在异常目标时，还可以继续对目标图像之后的图像进行检测，以检测异常目标的位置变化或姿态变化。
步骤404：当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。
由于单独根据第一检测结果或第二检测结果确定检测区域内存在异常目标太片面，可能不准确，因此，本申请中根据第一检测结果和第二检测结果来进行判断。当第一检测结果指示检测区域内存在异常目标时，说明基于红外信号的检测确定检测区域内存在异常目标，并根据第二检测结果进行校验，当第二检测结果也指示检测区域内存在异常目标时，说明基于N帧图像的检测确定检测区域内存在异常目标，在这种情况下，可以确定检测区域内确实存在异常目标。
也就是说，在确定检测区域内是否存在异常目标时，需要将红外检测和图像检测结合起来进行判断，只有当红外检测和图像检测均确定检测区域内存在异常目标时，才能确定检测区域内存在异常目标。
本申请实施例中，可以基于检测区域内检测的红外信号，确定第一检测结果，以及获取检测区域的N帧图像，对该N帧图像进行异常目标检测，得到第二检测结果，当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。如此，可以将红外检测与图像检测相结合，来确定检测区域是否存在异常目标，提高了检测结果的准确率，而且，由于图像检测不易受外界因素的干扰，可以降低红外传感器和微波探测器因外界因素干扰导致的检测失误，因此进一步提高了检测结果的准确率。
图8是根据一示例性实施例示出的一种异常目标的检测装置的示意图，该异常目标的检测装置可以由软件、硬件或者两者的结合实现。参见图8，该装置可以包括：
第一检测模块801，用于基于检测区域内检测的红外信号，确定第一检测结果，第一检测结果用于指示检测区域内是否存在异常目标；
第二检测模块802，用于获取检测区域的N帧图像，对N帧图像进行异常目标检测，得到第二检测结果，第二检测结果用于指示检测区域内是否存在异 常目标，N为正整数；
确定模块803，用于当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。
在本申请一种可能的实现方式中，该第二检测模块802用于：
基于检测区域内检测的红外信号，确定第一检测结果的同时，对该检测区域进行图像采集，将第一参考时长内采集到的图像，确定为该N帧图像。
在本申请一种可能的实现方式中，该第二检测模块802用于：
基于检测区域内检测的红外信号，确定第一检测结果之后，当该第一检测结果指示该检测区域内存在该异常目标时，对该检测区域进行图像采集，将第一参考时长内采集到的图像，确定为该N帧图像。
在本申请一种可能的实现方式中，该第二检测模块802用于：
在该第一参考时长内，每隔第二参考时长，通过图像传感器采集该检测区域的光学图像，该第二参考时长小于或等于该第一参考时长；
对该光学图像进行光电转换，得到该光学图像对应的电信号；
对该光学图像对应的电信号进行数字转换，得到二进制图像数据；
对该二进制图像数据进行像素转换，得到该检测区域的第一图像，该第一图像为不可视图像；
将该第一参考时长内得到的该第一图像，确定为该N帧图像。
在本申请一种可能的实现方式中，该第二检测模块802用于：
在该第一参考时长内，每隔第二参考时长，通过摄像头对该检测区域进行拍摄，得到该检测区域的第二图像，该第二图像为可视图像；
将该第一参考时长内拍摄得到的该第二图像，确定为该N帧图像。
在本申请一种可能的实现方式中，该第二检测模块802还用于：
当该第一检测结果指示该检测区域内存在该异常目标时，对该检测区域进行图像采集，以及触发该第一检测模块停止通过红外传感器，对该检测区域进行检测；
若该第二检测结果指示该检测区域内不存在该异常目标，则触发该第一检测模块继续通过红外传感器，对该检测区域进行检测。
在本申请一种可能的实现方式中，该第二检测模块802用于：
对该N帧图像进行异常目标检测；
若检测到该N帧图像中的一帧图像中存在该异常目标，则确定该检测区域 内存在该异常目标；
若检测到该N帧图像中均不存在该异常目标，则确定该检测区域不存在该异常目标。
在本申请一种可能的实现方式中，该第二检测模块802用于：
对该N帧图像并行进行异常目标检测；或者，
将该N帧图像中的第一帧图像确定为待检测的第三图像，对该第三图像进行异常目标检测；若检测到该第三图像中存在该异常目标，则确定该N帧图像中的一帧图像中存在该异常目标；若检测到该第三图像中不存在该异常目标，则将该第三图像中的下一帧图像确定为待检测的第三图像，并重复执行对该第三图像进行异常目标检测的步骤，直至检测到该N帧图像中的一帧图像中存在该异常目标或者该N帧图像中均不存在该异常目标为止。
在本申请一种可能的实现方式中，该第二检测模块802包括：
检测单元，用于在对所述N帧图像进行异常目标检测的过程中，对于所述N帧图像中的任一帧待检测的第三图像，基于该检测区域的背景图像模型，对该第三图像进行前景检测，该背景图像模型用于指示该检测区域的背景图像；
第一确定单元，用于若检测到该第三图像中存在与该检测区域的背景图像不同的前景目标，则确定该前景目标的高度和宽度；
第二确定单元，用于若基于该前景目标的高度和宽度，确定该前景目标为该异常目标，则确定该第三图像中存在该异常目标。
在本申请一种可能的实现方式中，该检测区域的背景图像模型为该检测区域的背景图像的参考图像特征；
该检测单元用于：
对该第三图像进行特征提取，得到该第三图像的参考图像特征；
计算该第三图像的参考图像特征与该检测区域的背景图像的参考图像特征之间的特征残差；
当该特征残差大于或等于特征残差阈值时，确定该第三图像中存在该前景目标。
在本申请一种可能的实现方式中，该检测单元用于：
基于该第三图像对应的二进制图像数据，通过以下公式确定该第三图像的参考图像特征：
Figure PCTCN2019124327-appb-000005
其中，L是指该第三图像的参考图像特征，I p是指该第三图像的第p个像素点对应的二进制图像数据，S是指该第p个像素点的坐标位置，I c是指该第三图像的中心像素点对应的二进制图像数据。
在本申请一种可能的实现方式中，该检测单元用于：
计算该前景目标的宽度与高度之间的比值，得到该前景目标的宽高比；
若该前景目标的宽高比在参考宽高比范围内，则确定该前景目标为该异常目标。
在本申请一种可能的实现方式中，该检测单元用于：
确定该第三图像的高度和宽度；
该基于该前景目标的高度和宽度，确定该前景目标为该异常目标，包括：
计算该前景目标的宽度与高度之间的比值，得到该前景目标的宽高比；
计算该前景目标的宽度与该第三图像的宽度之间的第一比值；
计算该前景目标的高度与该第三图像的高度之间的第二比值；
若该前景目标的宽高比在参考宽高比范围内、该第一比值在第一参考比值范围内且该第二比值在第二参考比值范围内，则确定该前景目标为该异常目标。
在本申请一种可能的实现方式中，该第一检测模块801用于：
通过红外传感器对所述检测区域内的红外信号进行检测，将检测的红外信号转换为电信号，该红外传感器至少包括红外探头和至少两个菲涅尔透镜；
根据该电信号的振动幅值，确定该第一检测结果。
在本申请一种可能的实现方式中，该第一检测模块801用于：
当该电信号的振动幅值在参考幅值范围内时，确定该检测区域内存在该异常目标。
在本申请一种可能的实现方式中，该第一检测模块801用于：
当该电信号的振动幅值在参考幅值范围内，且该电信号的振动频率在参考频率范围内时，确定该检测区域内存在该异常目标。
本申请实施例中，可以基于检测区域内检测的红外信号，确定第一检测结果，以及获取检测区域的N帧图像，对该N帧图像进行异常目标检测，得到第二检测结果，当第一检测结果和第二检测结果均指示检测区域内存在异常目标时，确定检测区域内存在异常目标。如此，可以将红外检测与图像检测相结合， 来确定检测区域是否存在异常目标，提高了检测结果的准确率，而且，由于图像检测不易受外界因素的干扰，可以降低红外传感器和微波探测器因外界因素干扰导致的检测失误，因此进一步提高了检测结果的准确率。
需要说明的是：上述实施例提供的异常目标的检测装置在检测异常目标时，仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能模块完成，即将装置的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。另外，上述实施例提供的异常目标的检测装置与异常目标的检测方法实施例属于同一构思，其具体实现过程详见方法实施例，这里不再赘述。
本领域普通技术人员可以理解实现上述实施例的全部或部分步骤可以通过硬件来完成，也可以通过程序来指令相关的硬件完成，所述的程序可以存储于一种计算机可读存储介质中，上述提到的存储介质可以是只读存储器，磁盘或光盘等。
以上所述仅为本申请的较佳实施例，并不用以限制本申请，凡在本申请的精神和原则之内，所作的任何修改、等同替换、改进等，均应包含在本申请的保护范围之内。

Claims (26)

  1. 一种异常目标的检测方法，其特征在于，所述方法包括：
    基于检测区域内检测的红外信号，确定第一检测结果，所述第一检测结果用于指示所述检测区域内是否存在异常目标；
    获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测，得到第二检测结果，所述第二检测结果用于指示所述检测区域内是否存在所述异常目标，所述N为正整数；
    当所述第一检测结果和所述第二检测结果均指示所述检测区域内存在所述异常目标时，确定所述检测区域内存在所述异常目标。
  2. 如权利要求1所述的方法，其特征在于，所述获取所述检测区域的N帧图像，包括：
    基于检测区域内检测的红外信号，确定第一检测结果的同时，对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像。
  3. 如权利要求1所述的方法，其特征在于，所述获取所述检测区域的N帧图像，包括：
    基于检测区域内检测的红外信号，确定第一检测结果之后，当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像。
  4. 如权利要求2或3所述的方法，其特征在于，所述对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像，包括：
    在所述第一参考时长内，每隔第二参考时长，通过图像传感器采集所述检测区域的光学图像，所述第二参考时长小于或等于所述第一参考时长；
    对所述光学图像进行光电转换，得到所述光学图像对应的电信号；
    对所述光学图像对应的电信号进行数字转换，得到二进制图像数据；
    对所述二进制图像数据进行像素转换，得到所述检测区域的第一图像，所述第一图像为不可视图像；
    将所述第一参考时长内得到的所述第一图像，确定为所述N帧图像。
  5. 如权利要求2或3所述的方法，其特征在于，所述对所述检测区域进行图像采集，将第一参考时长内采集到的图像，确定为所述N帧图像，包括：
    在所述第一参考时长内，每隔第二参考时长，通过摄像头对所述检测区域进行拍摄，得到所述检测区域的第二图像，所述第二图像为可视图像；
    将所述第一参考时长内拍摄得到的所述第二图像，确定为所述N帧图像。
  6. 如权利要求3所述的方法，其特征在于，所述当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，包括：
    当所述第一检测结果指示所述检测区域内存在所述异常目标时，对所述检测区域进行图像采集，以及停止基于检测区域内检测的红外信号，确定第一检测结果；
    所述对所述N帧图像进行异常目标检测，得到第二检测结果之后，还包括：
    若所述第二检测结果指示所述检测区域内不存在所述异常目标，则继续基于检测区域内检测的红外信号，确定第一检测结果。
  7. 如权利要求1所述的方法，其特征在于，所述对所述N帧图像进行异常目标检测，得到第二检测结果，包括：
    对所述N帧图像进行异常目标检测；
    若检测到所述N帧图像中的一帧图像中存在所述异常目标，则确定所述检测区域内存在所述异常目标；
    若检测到所述N帧图像中均不存在所述异常目标，则确定所述检测区域内不存在所述异常目标。
  8. 如权利要求7所述的方法，其特征在于，所述对所述N帧图像进行异常目标检测，包括：
    对所述N帧图像并行进行异常目标检测；或者，
    将所述N帧图像中的第一帧图像确定为待检测的第三图像，对所述第三图像进行异常目标检测；若检测到所述第三图像中存在所述异常目标，则确定所述N帧图像中的一帧图像中存在所述异常目标；若检测到所述第三图像中不存 在所述异常目标，则将所述第三图像中的下一帧图像确定为待检测的第三图像，并重复执行对所述第三图像进行异常目标检测的步骤，直至检测到所述N帧图像中的一帧图像中存在所述异常目标或者所述N帧图像中均不存在所述异常目标为止。
  9. 如权利要求7所述的方法，其特征在于，所述方法还包括：
    在对所述N帧图像进行异常目标检测的过程中，对于所述N帧图像中的任一帧待检测的第三图像，基于所述检测区域的背景图像模型，对所述第三图像进行前景检测，所述背景图像模型用于指示所述检测区域的背景图像；
    若检测到所述第三图像中存在与所述检测区域的背景图像不同的前景目标，则确定所述前景目标的高度和宽度；
    若基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标，则确定所述第三图像中存在所述异常目标。
  10. 如权利要求9所述的方法，其特征在于，所述检测区域的背景图像模型为所述检测区域的背景图像的参考图像特征；
    所述基于所述检测区域的背景图像模型，对所述第三图像进行前景检测，包括：
    对所述第三图像进行特征提取，得到所述第三图像的参考图像特征；
    计算所述第三图像的参考图像特征与所述检测区域的背景图像的参考图像特征之间的特征残差；
    当所述特征残差大于或等于特征残差阈值时，确定所述第三图像中存在所述前景目标。
  11. 根据权利要求10所述的方法，其特征在于，所述对所述第三图像进行特征提取，得到所述第三图像的参考图像特征，包括：
    基于所述第三图像对应的二进制图像数据，通过以下公式确定所述第三图像的参考图像特征：
    Figure PCTCN2019124327-appb-100001
    其中，L是指所述第三图像的参考图像特征，I p是指所述第三图像的第p个像素点对应的二进制图像数据，S是指所述第p个像素点的坐标位置，I c是指所述第三图像的中心像素点对应的二进制图像数据。
  12. 如权利要求9所述的方法，其特征在于，所述确定所述前景目标的高度和宽度之后，还包括：
    计算所述前景目标的宽度与高度之间的比值，得到所述前景目标的宽高比；
    若所述前景目标的宽高比在参考宽高比范围内，则确定所述前景目标为所述异常目标。
  13. 如权利要求9所述的方法，其特征在于，所述基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标之前，还包括：
    确定所述第三图像的高度和宽度；
    所述基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标，包括：
    计算所述前景目标的宽度与高度之间的比值，得到所述前景目标的宽高比；
    计算所述前景目标的宽度与所述第三图像的宽度之间的第一比值；
    计算所述前景目标的高度与所述第三图像的高度之间的第二比值；
    若所述前景目标的宽高比在参考宽高比范围内、所述第一比值在第一参考比值范围内且所述第二比值在第二参考比值范围内，则确定所述前景目标为所述异常目标。
  14. 如权利要求1所述的方法，其特征在于，所述基于检测区域内检测的红外信号，确定第一检测结果，包括：
    通过红外传感器对所述检测区域内的红外信号进行检测，将检测的红外信号转换为电信号，所述红外传感器至少包括红外探头和至少两个菲涅尔透镜；
    根据所述电信号的振动幅值，确定所述第一检测结果。
  15. 如权利要求14所述的方法，其特征在于，所述根据所述电信号的振动幅值，确定所述第一检测结果，包括：
    当所述电信号的振动幅值在参考幅值范围内时，确定所述检测区域内存在所述异常目标。
  16. 如权利要求14所述的方法，其特征在于，所述根据所述电信号的振动幅值，确定所述第一检测结果，包括：
    当所述电信号的振动幅值在参考幅值范围内，且所述电信号的振动频率在参考频率范围内时，确定所述检测区域内存在所述异常目标。
  17. 一种检测设备，其特征在于，所述检测设备包括红外检测单元、图像处理单元和处理器，所述红外检测单元至少包括红外传感器；
    所述红外检测单元，用于通过所述红外传感器，对所述检测区域内的红外信号进行检测，将检测信号发送给所述处理器；
    所述图像处理单元，用于获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测，得到第二检测结果，将所述第二检测结果发送给所述处理器，所述第二检测结果用于指示所述检测区域内是否存在所述异常目标，所述N为正整数；
    所述处理器用于根据所述检测信号确定第一检测结果，当所述第一检测结果和所述第二检测结果均指示所述检测区域内存在所述异常目标时，确定所述检测区域内存在所述异常目标，所述第一检测结果用于指示所述检测区域内是否存在所述异常目标。
  18. 如权利要求17所述的检测设备，其特征在于，所述处理器用于：
    在所述红外检测单元通过所述红外传感器，对所述检测区域内的红外信号进行检测的同时，控制所述图像处理单元，在第一参考时长内获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测。
  19. 如权利要求17所述的检测设备，其特征在于，所述处理器用于：
    在根据所述红外检测单元发送的检测信号，确定所述检测区域内存在所述异常目标之后，控制所述图像处理单元，在第一参考时长内获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测。
  20. 如权利要求18或19所述的检测设备，其特征在于，所述图像处理单元包括图像传感器，所述图像处理单元用于：
    在所述第一参考时长内，每隔第二参考时长，通过所述图像传感器采集所述检测区域的光学图像，所述第一参考时长大于或等于所述第二参考时长；
    对所述光学图像进行光电转换，得到所述光学图像对应的电信号；
    对所述光学图像对应的电信号进行数字转换，得到二进制图像数据；
    对所述二进制图像数据进行像素转换，得到所述检测区域的第一图像，所述第一图像为不可视图像；
    将第一参考时长内得到的所述第一图像，确定为所述N帧图像。
  21. 如权利要求19所述的检测设备，其特征在于，所述处理器用于：
    在根据所述红外检测单元发送的检测信号，确定所述检测区域内存在所述异常目标之后，控制所述图像处理单元，在所述第一参考时长内获取所述检测区域的N帧图像，对所述N帧图像进行异常目标检测，以及控制所述红外检测单元停止通过所述红外传感器，对所述检测区域内的红外信号进行检测；
    若通过所述图像处理单元确定所述检测区域内不存在所述异常目标，则控制所述红外检测单元继续通过所述红外传感器，对所述检测区域内的红外信号进行检测。
  22. 如权利要求17所述的检测设备，其特征在于，所述图像处理单元用于：
    对所述N帧图像进行异常目标检测；
    若检测到所述N帧图像中的一帧图像中存在所述异常目标，则确定所述检测区域内存在所述异常目标；
    若检测到所述N帧图像中均不存在所述异常目标，则确定所述检测区域内不存在所述异常目标。
  23. 如权利要求22所述的检测设备，其特征在于，所述图像处理单元用于：
    对所述N帧图像并行进行异常目标检测；或者，
    将所述N帧图像中的第一帧图像确定为待检测的第三图像，对所述第三图 像进行异常目标检测；若检测到所述第三图像中存在所述异常目标，则确定所述N帧图像中的一帧图像中存在所述异常目标；若检测到所述第三图像中不存在所述异常目标，则将所述第三图像中的下一帧图像确定为待检测的第三图像，并重复执行对所述第三图像进行异常目标检测的步骤，直至检测到所述N帧图像中的一帧图像中存在所述异常目标或者所述N帧图像中均不存在所述异常目标为止。
  24. 如权利要求22所述的检测设备，其特征在于，所述图像处理单元用于：
    在对所述N帧图像进行异常目标检测的过程中，对于所述N帧图像中的任一帧待检测的第三图像，基于所述检测区域的背景图像模型，对所述第三图像进行前景检测，所述背景图像模型用于指示所述检测区域的背景图像；
    若检测到所述第三图像中存在与所述检测区域的背景图像不同的前景目标，则确定所述前景目标的高度和宽度；
    若基于所述前景目标的高度和宽度，确定所述前景目标为所述异常目标，则确定所述第三图像中存在所述异常目标。
  25. 如权利要求17所述的检测设备，其特征在于，所述红外传感器至少包括红外探头和至少两个菲涅尔透镜，所述红外传感器用于：
    通过所述至少两个菲涅尔透镜，对所述检测区域内的红外信号进行聚焦；
    通过所述红外探头对所述至少两个菲涅尔透镜聚焦的红外信号进行感应，将感应的红外信号转换为所述检测信号。
  26. 一种计算机可读存储介质，所述计算机可读存储介质上存储有指令，其特征在于，所述指令被处理器执行时实现权利要求1-16所述的任一项方法的步骤。
PCT/CN2019/124327 2019-06-12 2019-12-10 异常目标的检测方法、设备及存储介质 Ceased WO2020248543A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP19933022.6A EP3965001A4 (en) 2019-06-12 2019-12-10 Abnormal target detection method and device, and storage medium
US17/618,137 US20220245941A1 (en) 2019-06-12 2019-12-10 Method and device for detecting abnormal target, and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910508062.6 2019-06-12
CN201910508062.6A CN112084813B (zh) 2019-06-12 2019-06-12 异常目标的检测方法、装置及存储介质

Publications (1)

Publication Number Publication Date
WO2020248543A1 true WO2020248543A1 (zh) 2020-12-17

Family

ID=73734603

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/124327 Ceased WO2020248543A1 (zh) 2019-06-12 2019-12-10 异常目标的检测方法、设备及存储介质

Country Status (4)

Country Link
US (1) US20220245941A1 (zh)
EP (1) EP3965001A4 (zh)
CN (1) CN112084813B (zh)
WO (1) WO2020248543A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114167223A (zh) * 2022-02-11 2022-03-11 北京美达龙源环保电气有限公司 供电异常的检测方法、装置及计算机可读存储介质
CN114220249A (zh) * 2021-12-03 2022-03-22 深圳快品信息技术有限公司 自提外卖配送的合理调度装置、系统及方法

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112861998A (zh) * 2021-03-16 2021-05-28 上海东普信息科技有限公司 神经网络模型构建方法、安全通道异常监测方法及系统
CN113724189B (zh) * 2021-03-17 2025-11-25 腾讯科技（深圳）有限公司 图像处理方法、装置、设备及存储介质
CN113238214B (zh) * 2021-04-14 2024-03-19 杭州萤石软件有限公司 目标对象检测方法、装置、电子设备和存储介质
CN113259698B (zh) * 2021-05-12 2022-11-15 北京字节跳动网络技术有限公司 更换画面中的背景的方法、设备、存储介质及程序产品
CN115456954A (zh) * 2022-08-18 2022-12-09 深圳市朗驰欣创科技股份有限公司 一种异常状态的识别方法、识别装置及终端设备
CN116188415A (zh) * 2023-02-03 2023-05-30 中国电子信息产业集团有限公司第六研究所 一种承载物中实体的检测方法、装置及电子设备
CN116862980B (zh) * 2023-06-12 2024-01-23 上海玉贲智能科技有限公司 图像边缘的目标检测框位置优化校正方法、系统、介质及终端

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101339688A (zh) * 2008-08-27 2009-01-07 北京中星微电子有限公司 一种入侵检测方法及系统
CN104866842A (zh) * 2015-06-05 2015-08-26 中国人民解放军国防科学技术大学 一种人体目标入侵检测方法
CN105163095A (zh) * 2015-10-16 2015-12-16 湖南智慧平安科技有限公司 面向偏僻地段的人体目标监控方法
CN108388845A (zh) * 2018-02-05 2018-08-10 李刚毅 对象检测方法和系统

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9036902B2 (en) * 2007-01-29 2015-05-19 Intellivision Technologies Corporation Detector for chemical, biological and/or radiological attacks
US7852210B2 (en) * 2007-12-31 2010-12-14 Honeywell International Inc. Motion detector for detecting tampering and method for detecting tampering
CN103392163B (zh) * 2008-10-10 2016-10-26 高通股份有限公司 单相机跟踪器
CN101937563B (zh) * 2009-07-03 2012-05-30 深圳泰山在线科技有限公司 一种目标检测方法和设备及其使用的图像采集装置
US8675066B2 (en) * 2009-10-02 2014-03-18 Alarm.Com Incorporated Image surveillance and reporting technology
US9367733B2 (en) * 2012-11-21 2016-06-14 Pelco, Inc. Method and apparatus for detecting people by a surveillance system
EP3632100A4 (en) * 2017-06-01 2020-06-03 Gary Franklin Bart CONTROL ACCESS USING VIDEO ANALYSIS
CN109271897B (zh) * 2018-08-31 2023-10-24 平安科技（深圳）有限公司 人脸动作单元检测方法、装置、计算机设备及存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101339688A (zh) * 2008-08-27 2009-01-07 北京中星微电子有限公司 一种入侵检测方法及系统
CN104866842A (zh) * 2015-06-05 2015-08-26 中国人民解放军国防科学技术大学 一种人体目标入侵检测方法
CN105163095A (zh) * 2015-10-16 2015-12-16 湖南智慧平安科技有限公司 面向偏僻地段的人体目标监控方法
CN108388845A (zh) * 2018-02-05 2018-08-10 李刚毅 对象检测方法和系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP3965001A4 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114220249A (zh) * 2021-12-03 2022-03-22 深圳快品信息技术有限公司 自提外卖配送的合理调度装置、系统及方法
CN114220249B (zh) * 2021-12-03 2023-08-08 深圳快品信息技术有限公司 自提外卖配送的合理调度装置、系统及方法
CN114167223A (zh) * 2022-02-11 2022-03-11 北京美达龙源环保电气有限公司 供电异常的检测方法、装置及计算机可读存储介质
CN114167223B (zh) * 2022-02-11 2022-05-31 北京美达龙源环保电气有限公司 供电异常的检测方法、装置及计算机可读存储介质

Also Published As

Publication number Publication date
US20220245941A1 (en) 2022-08-04
EP3965001A1 (en) 2022-03-09
CN112084813A (zh) 2020-12-15
CN112084813B (zh) 2024-06-04
EP3965001A4 (en) 2022-06-29

Similar Documents

Publication Publication Date Title
WO2020248543A1 (zh) 异常目标的检测方法、设备及存储介质
US9224278B2 (en) Automated method and system for detecting the presence of a lit cigarette
KR101660215B1 (ko) 존재 감지
CN111510668B (zh) 使用于运动传感器的运动侦测方法
US10255683B1 (en) Discontinuity detection in video data
JP2016503913A (ja) セキュリティ監視システム及び相応な警報触発方法
CN110414445B (zh) 用于人脸识别的光源调节方法、装置及电子设备
CN112312076A (zh) 智能型移动检测装置
KR101454644B1 (ko) 보행자 추적기를 이용한 서성거림을 탐지하는 방법
CN111183431A (zh) 指纹识别方法及终端设备
US20240377539A1 (en) Control of the brightness of a display
JP5761954B2 (ja) 画像センサ
CN114187621A (zh) 一种人脸识别方法及应用该方法的人脸识别装置
CN114987390B (zh) 车辆控制装置和车辆控制方法
US10268922B2 (en) Image processing by means of cross-correlation
JP2012048690A (ja) 監視装置
CN112419635B (zh) 融合光栅和视频的周界报警方法
US12244973B1 (en) On-device motion detection
JP4925942B2 (ja) 画像センサ
JP2000341675A (ja) 複合型監視装置
TWI813297B (zh) 移動物體偵測方法與電路系統
KR102375035B1 (ko) Pir 센서를 이용하는 도어캠
JP3736989B2 (ja) 画像監視装置
CN107292254A (zh) 近红外人体位姿检测装置和方法
JP2020167500A (ja) 画像処理装置及び画像処理プログラム

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19933022

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2019933022

Country of ref document: EP

Effective date: 20211203