WO2020183711A1 - 画像処理装置及び3次元計測システム - Google Patents
画像処理装置及び3次元計測システム Download PDFInfo
- Publication number
- WO2020183711A1 WO2020183711A1 PCT/JP2019/010590 JP2019010590W WO2020183711A1 WO 2020183711 A1 WO2020183711 A1 WO 2020183711A1 JP 2019010590 W JP2019010590 W JP 2019010590W WO 2020183711 A1 WO2020183711 A1 WO 2020183711A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- parallax
- predicted
- map
- pair
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/521—Depth or shape recovery from laser ranging, e.g. using interferometry; from the projection of structured light
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
- G06T7/593—Depth or shape recovery from multiple images from stereo images
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
- G01B11/25—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object
- G01B11/2545—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures by projecting a pattern, e.g. one or more lines, moiré fringes on the object with one projection direction and several detection directions, e.g. stereo
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- 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
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- 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
- G06T2207/10016—Video; Image sequence
- G06T2207/10021—Stereoscopic video; Stereoscopic image sequence
-
- 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
- G06T2207/10141—Special mode during image acquisition
- G06T2207/10144—Varying exposure
-
- 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
- G06T2207/10141—Special mode during image acquisition
- G06T2207/10152—Varying illumination
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20221—Image fusion; Image merging
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30244—Camera pose
Definitions
- the present invention relates to three-dimensional measurement using an image.
- various methods for performing three-dimensional measurement of an object are known, and they are roughly classified into a method using the straightness of light and a method using the speed of light, focusing on the properties of light.
- methods that use the straightness of light include methods that are classified into either active measurement (active measurement) or passive measurement (passive measurement), and methods that use the speed of light include methods that use the speed of light.
- Methods classified as active measurement (active measurement) are included.
- Non-Patent Document 1 as a specific example of the spatially coded pattern projection method, which is an example of the active measurement method, a spatially encoded (coded) pattern illumination is projected onto an object, and the pattern is projected. A method of acquiring a three-dimensional shape by analyzing an image of a captured object is described.
- FIG. 14 shows the principle of stereo matching.
- stereo matching for example, two cameras arranged on the left and right photograph the object O at the same time to obtain two images. One is the reference image I1 and the other is the comparison image I2, and the pixel (reference point P1) in the reference image I1 and the pixel (corresponding point P2) having the closest image feature are searched along the epipolar line E in the comparison image I2. Then, the difference in coordinates (parallax) between the reference point P1 and the corresponding point P2 is obtained. Since the geometrical position of each camera is known, the distance D (depth) in the depth direction can be calculated from the parallax by the principle of triangulation, and the three-dimensional shape of the object O can be restored.
- the difference in reflection characteristics on the surface of the object and environmental changes such as lighting tend to cause variations and deterioration in measurement accuracy.
- the illumination light is specularly reflected on the surface of a metal part and halation occurs, or if the dark area gradation is crushed by an object with low reflectance such as rubber, the amount of information required for distance estimation is insufficient and the accuracy is remarkable. There are cases where it drops or becomes unmeasurable.
- the measurement accuracy is affected when the amount of light of the illumination is insufficient, when the object is shaded by another object and the object is not sufficiently illuminated, or when the reflected light from the other object hits the object (so-called mutual reflection). Can exert.
- Patent Documents 1 and 2 disclose the idea of performing stereo matching a plurality of times on the same object and synthesizing the results in order to improve the measurement accuracy.
- the conventional stereo matching may be repeated a plurality of times to reduce the variation, the effect cannot be expected in the case where the amount of image information is insufficient.
- the present invention has been made in view of the above circumstances, and an object of the present invention is to provide a technique for improving accuracy and robustness in measurement by stereo matching.
- One aspect of the present invention is an image processing device that generates a depth map, which is data in which distance information is associated with the coordinates of each pixel, by stereo matching using an image pair.
- Disparity prediction that predicts the disparity between the first image and the second image by an image acquisition means that acquires an image pair consisting of a first image and a second image taken from different viewpoints and a method different from stereo matching.
- the setting means for setting the search range of the corresponding point in stereo matching based on the predicted parallax, and the set search range.
- a disparity map generation means that searches for points and generates a disparity map that is data in which disparity information is associated with the coordinates of each pixel based on the search result, and a plurality of disparity maps generated from each of a plurality of image pairs. It is characterized by having a parallax map synthesizing means for generating a composite parallax map by synthesizing the above, and a depth map generating means for converting the parallax information of the composite parallax map into distance information and generating the depth map.
- An image processing apparatus is provided.
- the search range of the corresponding point is limited based on the predicted parallax.
- the time required for the corresponding point search can be significantly shortened as compared with the conventional general stereo matching, but also the search can be narrowed down to a range in which the corresponding point is likely to exist.
- the accuracy and reliability of the corresponding point search can be improved.
- the predicted parallax generated by a method different from stereo matching the effect of complementing the part that is difficult to measure by stereo matching can be expected. Then, since the parallax maps obtained from the plurality of image pairs are combined, the variation in measurement can be reduced. Therefore, as a whole, highly accurate and highly reliable measurement results can be stably obtained.
- the parallax predicting means may generate a plurality of predicted parallax, and the setting means may use a composite predicted parallax obtained by synthesizing the plurality of predicted parallax to set a search range of the plurality of image pairs. By generating (measuring) the predicted parallax a plurality of times in this way, the accuracy and robustness can be further improved.
- the parallax predicting means may generate a plurality of predicted parallax, and the setting means may change the predicted parallax used for setting the search range for each image pair. By generating (measuring) the predicted parallax a plurality of times in this way, the accuracy and robustness can be further improved.
- the parallax predicting means When the parallax predicting means generates the predicted parallax from an image taken by a camera, the plurality of predicted parallax are generated from different images taken by changing the camera and / or shooting conditions. It may be a new one. Even if there are objects with different reflection characteristics in the field of view, or if there is concern about the shadow of the object or mutual reflection, if you shoot with different cameras and / or shooting conditions, you can start with at least one of the images. The possibility of extracting parallax information increases.
- the image acquisition means acquires a first image pair and a second image pair, and the disparity prediction means generates a first predicted disparity from the first image of the first image pair.
- a second predicted shift is generated from the second image of the second image pair, and the setting means combines the first predicted shift and the second predicted shift to obtain a composite predicted shift. It may be used for setting the search range of the first image pair and the second image pair.
- the image acquisition means acquires a first image pair and a second image pair, and the parallax prediction means generates a first predicted parallax from the first image of the first image pair.
- a second predicted parallax is generated from the second image of the second image pair, and the setting means obtains one of the first predicted parallax and the second predicted parallax. It may be used for setting the search range of the first image pair, and the other predicted parallax may be used for setting the search range of the second image pair.
- the viewpoints of the first image and the second image are different, it is possible to obtain highly accurate predicted parallax by predicting the parallax from both images. Moreover, even if the information of one image is missing, there is a high possibility that the parallax can be predicted using the other image. Therefore, the search range of the corresponding points can be set more appropriately, and the accuracy and robustness can be further improved. Further, by sharing the image between the generation of the predicted parallax and the stereo matching, the number of times of imaging and image transfer can be reduced, so that the efficiency and speed of the entire process can be improved. Further, since the same camera can be used, there is an advantage that the device configuration can be simplified and miniaturized.
- the plurality of image pairs may include two or more image pairs shot under different shooting conditions.
- the two or more image pairs taken under the different shooting conditions may include image pairs taken under different exposure conditions and / or lighting conditions.
- the parallax prediction means may predict the parallax based on the distance information obtained by the spatially coded pattern projection method as the method different from the stereo matching. This is because the spatially coded pattern projection method can obtain distance information in a much shorter processing time than stereo matching when an image sensor having the same resolution as stereo matching is used. Although the spatial resolution of the distance measurement of the spatially coded pattern projection method is lower than that of the stereo matching method, it can be said that it is necessary and sufficient for the purpose of predicting parallax.
- One aspect of the present invention provides a three-dimensional measurement system characterized by having a sensor unit having at least two cameras and an image processing device that generates a depth map using an image captured from the sensor unit. To do.
- the present invention may be regarded as an image processing device having at least a part of the above means, or may be regarded as a three-dimensional measurement system having a sensor unit and an image processing device. Further, the present invention may be regarded as an image processing including at least a part of the above processing, a three-dimensional measurement method, a distance measuring method, a control method of an image processing device, or the like, or a program for realizing such a method.
- the program can also be regarded as a recording medium for recording non-temporarily.
- the present invention can be constructed by combining each of the above means and treatments with each other as much as possible.
- FIG. 1 is a diagram schematically showing a configuration example of a three-dimensional measurement system, which is one of the application examples of the present invention.
- FIG. 2 is a diagram schematically showing an outline of functions and processing of a three-dimensional measurement system.
- FIG. 3 is a functional block diagram of the three-dimensional measurement system according to the first embodiment.
- FIG. 4 is a flow chart showing the flow of the measurement process of the first embodiment.
- FIG. 5 is a timing chart of the first embodiment.
- FIG. 6 is a flow chart showing the flow of the measurement process of the second embodiment.
- FIG. 7 is a timing chart of the second embodiment.
- FIG. 8 is a diagram showing a configuration example of the sensor unit of the third embodiment.
- FIG. 9 is a flow chart showing the flow of the measurement process of the fourth embodiment.
- FIG. 1 is a diagram schematically showing a configuration example of a three-dimensional measurement system, which is one of the application examples of the present invention.
- FIG. 2 is a diagram schematically showing
- FIG. 10 is a timing chart of the fourth embodiment.
- FIG. 11 is a flow chart showing the flow of the measurement process of the fifth embodiment.
- FIG. 12 is a timing chart of the fifth embodiment.
- 13A to 13B are diagrams showing an example of controlling the brightness of the pattern floodlight section.
- FIG. 14 is a diagram illustrating the principle of stereo matching.
- FIG. 1 is a diagram schematically showing a configuration example of a three-dimensional measurement system, which is one of the application examples of the present invention.
- the three-dimensional measurement system 1 is a system for measuring the three-dimensional shape of the object 12 by image sensing, and is roughly composed of a sensor unit 10 and an image processing device 11.
- the sensor unit 10 includes at least a camera (also called an image sensor or an image pickup device), and may include other sensors as needed.
- the output of the sensor unit 10 is taken into the image processing device 11.
- the image processing device 11 is a device that performs various processes using the data captured from the sensor unit 10.
- the processing of the image processing device 11 may include, for example, distance measurement (distance measurement), three-dimensional shape recognition, object recognition, scene recognition, and the like.
- the processing result of the image processing device 11 is output to an output device such as a display, transferred to the outside, and used for inspection, control of other devices, and the like.
- an output device such as a display
- Such a three-dimensional measurement system 1 is applied to a wide range of fields, including, for example, computer vision, robot vision, and machine vision.
- the configuration shown in FIG. 1 is just an example, and the hardware configuration may be appropriately designed according to the application of the three-dimensional measurement system 1.
- the sensor unit 10 and the image processing device 11 may be wirelessly connected, or the sensor unit 10 and the image processing device 11 may be configured as an integrated device.
- the sensor unit 10 and the image processing device 11 may be connected via a wide area network such as a LAN or the Internet.
- a plurality of sensor units 10 may be provided for one image processing device 11, or conversely, the output of one sensor unit 10 may be provided to the plurality of image processing devices 11.
- the viewpoint of the sensor unit 10 may be movable by attaching the sensor unit 10 to a robot or a moving body.
- FIG. 2 is a diagram schematically showing an outline of the functions and processing of the three-dimensional measurement system 1.
- the three-dimensional measurement system 1 includes two measurement systems, a first measurement system 21 and a second measurement system 22, as measurement systems for measuring the distance of the object 12.
- the functions and processes of the measurement systems 21 and 22 are realized by the sensor unit 10 and the image processing device 11 in cooperation with each other.
- the first measurement system 21 measures the depth distance (depth) to the object 12 by stereo matching (also called stereo vision, stereo camera method, etc.). Since stereo matching enables measurement with high spatial resolution, the distance information generated by the first measurement system 21 is used as the final output in the system 1.
- the second measurement system 22 also measures the distance of the object 12, but the distance information obtained by the second measurement system 22 roughly predicts the parallax observed by the first measurement system 21. It is used as an auxiliary for the purpose of narrowing the search range in stereo matching. As the second measurement system 22, any measurement system may be used as long as the distance measurement is performed by a method different from stereo matching.
- active measurement methods that use the straightness of light include, for example, a spatial coded pattern projection method based on triangulation, a time coded pattern projection method, a moire topography method (contour line method), and illuminance.
- a spatial coded pattern projection method based on triangulation a time coded pattern projection method
- a moire topography method a moire topography method
- illuminance examples include a difference stereo method (illumination direction / Photometric Stereo), an illuminance difference method based on coaxial distance measurement, a laser cofocal method, a white cofocal method, an optical interference method, and the like.
- a passive measurement method using the straightness of light for example, a visual volume crossing method (Shape from silhouette), a factorization method (factorization), a Depth from Motion (Structure from Motion) method, a Depth from Shading method, etc. , Depth from focusing method, Depth from defocus method, Depth from zoom method, etc., which are based on coaxial ranging.
- an optical time difference (TOF) measurement method based on simultaneous distance measurement for example, an optical phase difference (TOF) measurement method, and radio waves, sound waves, and millimeter waves ( TOF) method and the like can be mentioned.
- TOF optical time difference
- TOF optical phase difference
- TOF radio waves, sound waves, and millimeter waves
- the second measurement system 22 any of the above methods may be adopted as the second measurement system 22.
- the purpose of the second measurement system 22 is to roughly predict the parallax, the measurement accuracy and spatial resolution may be lower than those of stereo matching. Therefore, a high-speed method having a shorter measurement time than stereo matching should be used. Is preferable.
- the spatial coding pattern projection method is used because of the advantage that the measurement time is short and the advantage that the sensor and the image can be shared with the first measurement system 21.
- the first measurement system 21 acquires a stereo image pair consisting of two images (referred to as a first image and a second image) from the sensor unit 10. These two images are taken of the object 12 from different viewpoints (line-of-sight directions) so as to cause parallax with respect to the object 12.
- the sensor unit 10 includes a plurality of cameras
- the first image and the second image may be captured by the two cameras at the same time.
- the first image and the second image may be acquired by a single camera by continuously shooting while moving the camera.
- the second measurement system 22 measures the distance of the object 12, predicts the parallax between the first image and the second image based on the obtained distance information, and uses the predicted parallax as a reference parallax map. Output as.
- the parallax map generated by the second measurement system 22 is referred to as a “reference parallax map” in order to distinguish it from the parallax map generated by the stereo matching of the first measurement system 21. Since the reference parallax map is used as an auxiliary for narrowing down the search range in the stereo matching of the first measurement system 21, the spatial resolution may be lower (coarse) than that of the first image and the second image. Absent.
- the reference parallax map may be generated on the image processing device 11 side based on the image obtained from the sensor unit 10 or other sensing data, or when the sensor unit 10 itself has a distance measuring function (TOF method).
- a reference parallax map may be generated on the sensor unit 10 side (such as an image sensor).
- the first measurement system 21 sets the search range of the corresponding points in stereo matching using the reference parallax map acquired from the second measurement system 22.
- the search range of the corresponding point may be set so as to include the error range. For example, when the value of the predicted parallax is d [pixels] and the error is ⁇ derr [pixels], the search range may be set as d-derr-c to d + derr + c. c is a margin.
- the search range may be set individually for all the pixels of the first image, or if the change in local parallax in the image is not large, the first image is divided into a plurality of areas. The search range may be set for each area.
- the first measurement system 21 searches for the corresponding point of each pixel between the first image and the second image from the set search range. For example, when the first image is a reference image and the second image is a comparison image, the pixel in the second image having the closest image feature to the pixel (reference point) in the first image is selected as the corresponding point, and the reference point is selected. The difference between the coordinates of the corresponding point and the corresponding point is obtained as the parallax at the reference point.
- a search for corresponding points is performed for all the pixels in the first image, and a parallax map is generated from the search results.
- the parallax map is data in which parallax information is associated with the coordinates of each pixel.
- the first measurement system 21 performs the above-mentioned processes (1) to (4) twice or more to obtain a plurality of parallax maps. Then, the first measurement system 21 generates a composite parallax map by synthesizing a plurality of parallax maps.
- the first measurement system 21 uses the principle of triangulation to convert the parallax information of the composite parallax map into distance information (depth) to generate a depth map.
- the search range of the corresponding point is limited based on the predicted parallax.
- the time required for the corresponding point search can be significantly shortened as compared with the conventional general stereo matching, but also the search can be narrowed down to a range in which the corresponding point is likely to exist.
- the accuracy and reliability of the corresponding point search can be improved.
- the predicted parallax generated by a method different from stereo matching the effect of complementing the part that is difficult to measure by stereo matching can be expected. Then, since the parallax maps obtained from the plurality of image pairs are combined, the variation in measurement can be reduced. Therefore, as a whole, highly accurate and highly reliable measurement results can be stably obtained.
- FIG. 3 is a functional block diagram of the three-dimensional measurement system 1.
- the sensor unit 10 includes a first camera 101, a second camera 102, a pattern floodlight unit 103, an illumination unit 104, an image transfer unit 105, and a drive control unit 106.
- the first camera 101 and the second camera 102 are a pair of cameras constituting a so-called stereo camera, and are arranged at a predetermined distance. By simultaneously shooting with the two cameras 101 and 102, an image pair shot from different viewpoints can be obtained (the image of the first camera 101 is called the first image, and the image of the second camera 102 is called the second image). ).
- the two cameras 101 and 102 may be arranged so that their optical axes intersect each other and the horizontal lines (or vertical lines) are on the same plane. By adopting such an arrangement, the epipolar line becomes parallel to the horizontal line (or vertical line) of the image, so that the corresponding point in stereo matching may be searched from within the horizontal line (or vertical line) at the same position. This is because the search process can be simplified.
- a monochrome camera or a color camera may be used as the cameras 101 and 102.
- the pattern floodlight unit 103 is a device for projecting the pattern illumination used in the distance measurement of the space-coded pattern projection method onto the object 12, and is also called a projector.
- the pattern floodlight unit 103 includes, for example, a light source unit, a light guide lens, a pattern generation unit, a projection lens, and the like.
- a light source unit an LED, a laser, a VCSEL (Vertical cavity Surface-emitting Laser), or the like can be used.
- the light guide lens is an optical element for guiding light from the light source unit to the pattern generation unit, and a lens, a glass rod, or the like can be used.
- the pattern generator is a member or device that generates a coded pattern, and is a photomask, a diffractive optical element (for example, DOE (Diffractive Optical Element)), an optical modulation element (for example, DLP (Digital Light Processing), LCD (for example). LiquidCrystalDisplay), LCoS (LiquidCrystalonSilicon), MEMS (MicroElectroMechanicalSystems), etc. can be used.
- a projection lens is an optical element that magnifies and projects a generated pattern.
- the illumination unit 104 is uniform illumination used for capturing a general visible light image.
- white LED lighting is used.
- the illumination may be in the same wavelength band as the active projection.
- the image transfer unit 105 transfers the data of the first image taken by the first camera 101 and the data of the second image taken by the second camera 102 to the image processing device 11.
- the image transfer unit 105 may transfer the first image and the second image as separate image data, or join the first image and the second image to generate a side-by-side image and transfer it as a single image data. You may. Further, the image transfer unit 105 may connect a plurality of images (for example, a plurality of images having different exposure times) taken under different shooting conditions to generate a side-by-side image and transfer the images as a single image data.
- the drive control unit 106 is a unit that controls the first camera 101, the second camera 102, the pattern floodlight unit 103, and the illumination unit 104. The image transfer unit 105 and the drive control unit 106 may be provided on the image processing device 11 side instead of the sensor unit 10 side.
- the image processing device 11 includes an image acquisition unit 110, a pattern decoding unit 111, a parallax prediction unit 112, a preprocessing unit 113, a search range setting unit 115, a corresponding point search unit 116, a parallax map synthesis unit 114, and a parallax map post-processing unit 117. , Has a depth map generator 118.
- the image acquisition unit 110 has a function of capturing necessary image data from the sensor unit 10.
- the image acquisition unit 110 sends an image pair composed of the first image and the second image to the pattern decoding unit 111 and the preprocessing unit 113.
- the pattern decoding unit 111 has a function of acquiring distance information from the first image or the second image by the spatially coded pattern projection method.
- the spatial resolution is determined depending on the size of the unit pattern used. For example, when a unit pattern of 5 pixels ⁇ 5 pixels is used, the spatial resolution of the distance information is 1/25 of the input image.
- the parallax prediction unit 112 has a function of predicting the parallax between the first image and the second image based on the distance information obtained by the pattern decoding unit 111 and outputting a reference parallax map.
- the preprocessing unit 113 has a function of performing necessary preprocessing on the first image and the second image.
- the search range setting unit 115 has a function of setting the search range of the corresponding point based on the predicted parallax.
- the corresponding point search unit 116 has a function of searching for a corresponding point between the first image and the second image and generating a parallax map based on the search result.
- the parallax map synthesizing unit 114 has a function of generating a composite parallax map by synthesizing a plurality of parallax maps generated from each of the plurality of image pairs.
- the parallax map post-processing unit 117 has a function of performing necessary post-processing on the composite parallax map.
- the depth map generation unit 118 has a function of converting the parallax information of the composite parallax map into the distance information and generating the depth map.
- the image processing device 11 is composed of, for example, a computer including a CPU (processor), a RAM (memory), a non-volatile storage device (hard disk, SSD, etc.), an input device, an output device, and the like.
- the CPU expands the program stored in the non-volatile storage device into the RAM and executes the program to realize the various functions described above.
- the configuration of the image processing device 11 is not limited to this, and all or part of the above-mentioned functions may be realized by a dedicated circuit such as FPGA or ASIC, or realized by cloud computing or distributed computing. You may.
- the first camera 101, the pattern floodlight unit 103, the image transfer unit 105, the image acquisition unit 110, the drive control unit 106, the pattern decoding unit 111, and the parallax prediction unit 112 make up the second measurement system 22 in FIG.
- the first measurement system 21 in FIG. 2 is configured by the map composition unit 114, the parallax map post-processing unit 117, and the depth map generation unit 118.
- FIG. 4 is a flow chart showing a flow of processing executed by the image processing device 11.
- FIG. 5 is a timing chart.
- the first measurement is performed using the start signal from the drive control unit 106 as a trigger.
- the pattern projection unit 103 lights up, and a predetermined pattern illumination is projected onto the object 12.
- the first camera 101 and the second camera 102 simultaneously perform image capture, and the image transfer unit 105 transfers the first image and the second image.
- the image acquisition unit 110 acquires a first image pair composed of a first image and a second image.
- the image acquisition unit 110 sends the first image to the pattern decoding unit 111, and sends the first image and the second image to the preprocessing unit 113.
- the second measurement is performed using the start signal from the drive control unit 106 as a trigger.
- the pattern projection unit 103 lights up, and a predetermined pattern illumination is projected onto the object 12.
- the first camera 101 and the second camera 102 simultaneously perform image capture, and the image transfer unit 105 transfers the first image and the second image.
- the image acquisition unit 110 acquires a second image pair composed of the first image and the second image.
- the image acquisition unit 110 sends the second image to the pattern decoding unit 111, and sends the first image and the second image to the preprocessing unit 113.
- step S402 image processing for the first image pair obtained in the first measurement is started.
- the preprocessing unit 113 performs parallelization processing (rectification) on the first image and the second image.
- the parallelization process is a process of geometrically transforming one or both images so that the corresponding points between the two images are on the same horizontal line (or vertical line) in the image. Since the epipolar line becomes parallel to the horizontal line (or vertical line) of the image by the parallelization process, the process of searching for the corresponding point in the subsequent stage becomes easy. If the parallelism of the image captured from the sensor unit 10 is sufficiently high, the parallelization process may be omitted.
- step S403 the preprocessing unit 113 calculates the hash feature amount for each pixel of the parallelized first image and the second image, and replaces the value of each pixel with the hash feature amount.
- the hash feature amount represents the luminance feature of the local region centered on the pixel of interest, and here, the hash feature amount consisting of a bit string of eight elements is used.
- step S404 the pattern decoding unit 111 analyzes the first image and decodes the pattern to acquire distance information in the depth direction at a plurality of points on the first image.
- processing for the first image may be started from the time when the transfer of the first image is completed ( ⁇ in FIG. 5) to shorten the overall processing time.
- step S412 the preprocessing unit 113 performs parallelization processing (rectification) on the first image and the second image.
- the preprocessing unit 113 calculates the hash feature amount for each pixel of the parallelized first image and the second image, and replaces the value of each pixel with the hash feature amount.
- step S414 the pattern decoding unit 111 analyzes the second image and decodes the pattern to acquire the distance information based on the second image. Then, in step S415, the parallax prediction unit 112 calculates the predicted parallax based on the distance information.
- the predicted parallax obtained in the first time (step S405) and the predicted parallax obtained in the second time (step S415) should be the same, but in reality they are not exactly the same. This is because the first and second times use images taken by different cameras (viewpoints), so there is a difference in the appearance (that is, image information) of the images. Therefore, it is possible that there is a difference in the predicted parallax value between the first and second times, or that one succeeds in predicting the parallax but the other fails. Therefore, in step S420, the parallax prediction unit 112 synthesizes the first predicted parallax and the second predicted parallax to obtain the combined predicted parallax.
- the synthesis method is not particularly limited, but for example, when the predicted parallax is obtained in both the first and second times, the average value thereof is set as the composite predicted parallax, and the predicted parallax is obtained only in either the first time or the second time. If is obtained, the value may be used as it is as the composite predicted parallax.
- the search range setting unit 115 sets the search range of the corresponding points for each of the first image pair and the second image pair based on the composite predicted parallax.
- the size of the search range is determined in consideration of the prediction error. For example, when the prediction error is ⁇ 10 pixels, it is considered sufficient to set about ⁇ 20 pixels centered on the predicted parallax in the search range even if the margin is included. If the horizontal line has 640 pixels and the search range can be narrowed down to ⁇ 20 pixels (that is, 40 pixels), the search process can be simply reduced to 1/16 compared to searching the entire horizontal line. Can be done.
- step S422 the corresponding point search unit 116 searches for the corresponding point between the first image pairs and obtains the parallax of each pixel.
- the corresponding point search unit 116 generates parallax data in which parallax information is associated with points (pixel coordinates) that have succeeded in detecting the corresponding points. This information is the parallax map 1.
- step S423 the corresponding point search unit 116 searches for the corresponding point between the second image pairs and generates the parallax map 2.
- the parallax map compositing unit 114 synthesizes the parallax map 1 obtained from the first image pair and the parallax map 2 obtained from the second image pair to generate a composite parallax map.
- the composition method is not particularly limited, but for example, when parallax is obtained in both the parallax map 1 and the parallax map 2, the average value thereof is used as the composite parallax, and only one of the parallax map 1 and the parallax map 2 is used. When the parallax is obtained, the value may be used as it is as the composite parallax.
- step S425 the parallax map post-processing unit 117 corrects the composite parallax map. Since the parallax map estimated by the corresponding point search includes erroneous measurement points and measurement omissions, the erroneous measurement points are corrected and the measurement omissions are complemented based on the parallax information of the surrounding pixels. Either of the processes of steps S424 and S425 may be performed first.
- the depth map generation unit 118 converts the parallax information of each pixel of the composite parallax map into three-dimensional information (distance information in the depth direction) to generate a depth map. This depth map (three-dimensional point cloud data) is used, for example, for shape recognition and object recognition of the object 12.
- the time required for the corresponding point search can be significantly shortened as compared with the conventional general stereo matching, but also the accuracy of the corresponding point search and the accuracy of the corresponding point search can be achieved. Reliability can be improved.
- the predicted parallax generated by the spatially coded pattern projection method an effect of complementing a part that is difficult to measure by stereo matching can be expected. Then, since the parallax maps obtained from the plurality of image pairs are combined, the variation in measurement can be reduced. Therefore, as a whole, highly accurate and highly reliable measurement results can be stably obtained.
- the search range of the corresponding points can be set more appropriately, and the accuracy and robustness can be further improved.
- the number of times of imaging and image transfer can be reduced, so that the efficiency and speed of the entire process can be improved. Further, since the same camera can be used, there is an advantage that the device configuration can be simplified and miniaturized.
- FIG. 6 is a flow chart showing a flow of processing executed by the image processing device 11.
- FIG. 7 is a timing chart.
- the predicted parallax obtained from the first image and the predicted parallax obtained from the second image are combined, and the search range is set based on the combined predicted parallax, whereas in the second embodiment, individual predictions are made.
- the search range is set based on the parallax (steps S600, S610). Since the other processing is the same as that of the first embodiment, the same reference numerals as those in FIG. 4 are added and the description thereof will be omitted.
- the same effects as those of the first embodiment can be obtained by the configuration and processing of the present embodiment.
- the processing of the present embodiment does not synthesize the predicted parallax, as shown in FIG. 7, as soon as the image processing 1 is completed (that is, when the predicted parallax for the first image pair is obtained).
- the parallax map 1 generation process can be started. Therefore, the total processing time can be shortened as compared with the first embodiment.
- FIG. 8 schematically shows the configuration of the sensor unit 10 of the three-dimensional measurement system according to the third embodiment.
- the sensor unit 10 of the present embodiment has a structure in which four cameras 81, 82, 83, and 84 are arranged around the pattern floodlight unit 80.
- the four cameras 81 to 84 simultaneously shoot with the pattern illumination projected from the pattern floodlight unit 80 onto the object 12, and four images having different viewpoints are captured.
- An image pair selected from these four images is used for stereo matching. There are six combinations of image pairs, and any pair may be selected, or two or more pairs may be selected. For example, an image pair whose horizontal line or vertical line is parallel to the epipolar line may be preferentially selected, or an image pair having less loss of image information may be preferentially selected. Since the other configurations and processes are the same as those in the above-described embodiment, the description thereof will be omitted.
- FIG. 8 is an example, and the arrangement of the pattern floodlight and the camera, the number of the pattern floodlight, the number of cameras, and the like can be arbitrarily designed.
- FIG. 9 is a flow chart showing a flow of processing executed by the image processing device 11.
- FIG. 10 is a timing chart.
- the first measurement and the second measurement are performed under the same shooting conditions, whereas in the fourth embodiment, the exposure time is changed between the first measurement and the second measurement.
- the image processing device 11 performs stereo matching by the image pair (steps S900, S901) of the first image and the second image captured in the exposure time 1, and also captures the image in the exposure time 2. Stereo matching is performed by the image pair (steps S910 and S911) of the first image and the second image, and the two obtained disparity maps 1 and 2 are combined to output the final measurement result.
- the parallax map 1 obtained from the image pair having the exposure time 1 and the parallax map 2 obtained from the image pair having the exposure time 2 the difference in reflection characteristics, the lighting condition, and the like can be obtained.
- robust three-dimensional measurement can be realized.
- two images, a first image taken at the exposure time 1 and a first image taken at the exposure time 2 were used for predicting the parallax, but the same as in the first embodiment, the first image was used.
- Two of the 1st image and the 2nd image may be used for predicting the misalignment.
- the configuration in which three or more cameras described in the third embodiment are provided may be applied to the present embodiment.
- FIG. 11 is a flow chart showing a flow of processing executed by the image processing device 11.
- FIG. 12 is a timing chart.
- the search range is set based on the combined predicted parallax
- the search range is set based on the individual predicted parallax (steps S1100 and S1110). Since the other processing is the same as that of the fourth embodiment, the same reference numerals as those in FIG. 9 are added and the description thereof will be omitted.
- the same effects as those of the fourth embodiment can be obtained by the configuration and processing of the present embodiment.
- the processing of the present embodiment does not synthesize the predicted parallax, as shown in FIG. 12, as soon as the image processing 1 is completed (that is, when the predicted parallax for the first image pair is obtained).
- the parallax map 1 generation process can be started. Therefore, the total processing time can be shortened as compared with the fourth embodiment.
- the configuration in which three or more cameras described in the third embodiment are provided may be applied to the present embodiment.
- the exposure conditions of the camera are changed, but the same processing can be performed by changing the lighting conditions instead. That is, in the first measurement, the pattern floodlight unit 103 is set to the first brightness for shooting, and in the second measurement, the pattern floodlight unit 103 is changed to the second brightness for shooting. At this time, if the setting is set as "first brightness ⁇ second brightness" and the exposure conditions of the camera are the same in the first measurement and the second measurement, the first measurement will be a mirror object or the like.
- An image suitable for obtaining 3D information (parallax information) of a bright-colored object is obtained, and in the second measurement, 3D information (parallax information) of a low-reflectance object or a shaded part is obtained.
- An image suitable for is obtained. Therefore, by performing stereo matching on each image pair, synthesizing the obtained parallax map, and generating the final measurement result, robust three-dimensional measurement can be performed for differences in reflection characteristics and lighting conditions. realizable.
- a method of controlling the brightness of the pattern floodlight unit 103 there are a method of controlling the lighting time, a method of controlling the duty ratio, a method of controlling the lighting intensity, and the like, and any method may be used.
- 13A to 13C show an example of a drive signal of the pattern floodlight unit 103.
- the above-described embodiment is merely an example of a configuration example of the present invention.
- the present invention is not limited to the above-mentioned specific form, and various modifications can be made within the scope of its technical idea.
- the space-coded pattern projection method is illustrated, but any method may be adopted as the distance measuring method of the second measurement system as long as it is a method other than stereo matching.
- the hash feature amount is used for stereo matching, but another method may be used for evaluating the similarity of the corresponding points.
- an evaluation index of similarity there is a similarity calculation method of pixels of left and right images by SAD (Sum of Absolute Difference), SSD (Sum of Squared Difference), NC (Normalized Correlation) and the like. Further, in the above embodiment, the image of the camera shared by the generation of the reference depth map (prediction of parallax) and the stereo matching is used, but different images of the cameras for three-dimensional measurement may be used.
- An image processing device (11) that generates a depth map that is data in which distance information is associated with the coordinates of each pixel by stereo matching using an image pair.
- An image acquisition means (110) for acquiring an image pair consisting of a first image and a second image taken from different viewpoints, and Parallax prediction means (112) that predicts the parallax between the first image and the second image by a method different from stereo matching, and
- a setting means (115) for setting a search range for corresponding points in stereo matching based on the predicted parallax, and The corresponding point of each pixel between the first image and the second image is searched only in the set search range, and based on the search result, the data in which the parallax information is associated with the coordinates of each pixel is used.
- Parallax map generation means (116) that generates a certain parallax map
- a parallax map synthesizing means (114) that generates a composite parallax map by synthesizing a plurality of parallax maps generated from each of a plurality of image pairs
- Depth map generation means (118) that converts the parallax information of the composite parallax map into distance information and generates the depth map
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Optics & Photonics (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Measurement Of Optical Distance (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
Description
異なる視点から撮影された第1画像及び第2画像からなる画像ペアを取得する画像取得手段と、ステレオマッチングとは異なる方式により前記第1画像と前記第2画像の間の視差を予測する視差予測手段と、前記予測視差に基づき、ステレオマッチングにおける対応点の探索範囲を設定する設定手段と、前記設定された探索範囲に限定して前記第1画像と前記第2画像の間の各画素の対応点を探索し、その探索結果に基づき、各画素の座標に視差情報が関連付けられたデータである視差マップを生成する視差マップ生成手段と、複数の画像ペアのそれぞれから生成された複数の視差マップを合成することにより合成視差マップを生成する視差マップ合成手段と、前記合成視差マップの視差情報を距離情報に変換し、前記デプスマップを生成するデプスマップ生成手段と、を有することを特徴とする画像処理装置を提供する。
図1は、本発明の適用例の一つである3次元計測システムの構成例を模式的に示す図である。3次元計測システム1は、画像センシングによって対象物12の3次元形状を計測するためのシステムであり、概略、センサユニット10と画像処理装置11から構成される。センサユニット10は、少なくともカメラ(イメージセンサや撮像装置とも呼ばれる)を備えており、必要に応じて他のセンサを備える場合もある。センサユニット10の出力は画像処理装置11に取り込まれる。画像処理装置11は、センサユニット10から取り込まれたデータを用いて各種の処理を行うデバイスである。画像処理装置11の処理としては、例えば、距離計測(測距)、3次元形状認識、物体認識、シーン認識などが含まれてもよい。画像処理装置11の処理結果は、例えば、ディスプレイなどの出力装置に出力されたり、外部に転送されて検査や他の装置の制御等に利用される。このような3次元計測システム1は、例えば、コンピュータビジョン、ロボットビジョン、マシンビジョンをはじめとして、幅広い分野に適用される。
図3を参照して、第1実施形態に係る3次元計測システム1の構成例について説明する。図3は、3次元計測システム1の機能ブロック図である。
センサユニット10は、第1カメラ101、第2カメラ102、パターン投光部103、照明部104、画像転送部105、駆動制御部106を有する。
画像処理装置11は、画像取得部110、パターン復号部111、視差予測部112、前処理部113、探索範囲設定部115、対応点探索部116、視差マップ合成部114、視差マップ後処理部117、デプスマップ生成部118を有する。
図4及び図5を参照して、第1実施形態の計測処理の流れを説明する。図4は、画像処理装置11により実行される処理の流れを示すフロー図である。図5は、タイミングチャートである。
図6及び図7を参照して、第2実施形態の計測処理の流れを説明する。図6は、画像処理装置11により実行される処理の流れを示すフロー図である。図7は、タイミングチャートである。第1実施形態では、第1画像から求めた予測視差と第2画像から求めた予測視差を合成し、合成予測視差に基づき探索範囲を設定したのに対し、第2実施形態では、個別の予測視差に基づき探索範囲を設定する(ステップS600、S610)。それ以外の処理は第1実施形態と同様であるため、図4と同一の符号を付して説明を省略する。
図8に第3実施形態に係る3次元計測システムのセンサユニット10の構成を模式的に示す。本実施形態のセンサユニット10は、パターン投光部80を中心にして4つのカメラ81、82、83、84が配置された構造を有している。
図9及び図10を参照して、第4実施形態の計測処理の流れを説明する。図9は、画像処理装置11により実行される処理の流れを示すフロー図である。図10は、タイミングチャートである。第1実施形態では、同一の撮影条件で1回目の計測と2回目の計測を行ったのに対し、第4実施形態では、1回目の計測と2回目の計測とで露光時間を変える。
図11及び図12を参照して、第5実施形態の計測処理の流れを説明する。図11は、画像処理装置11により実行される処理の流れを示すフロー図である。図12は、タイミングチャートである。第4実施形態では、合成予測視差に基づき探索範囲を設定したのに対し、第5実施形態では、個別の予測視差に基づき探索範囲を設定する(ステップS1100、S1110)。それ以外の処理は第4実施形態と同様であるため、図9と同一の符号を付して説明を省略する。
第4及び第5実施形態では、カメラの露光条件を変化させたが、代わりに照明条件を変化させることによっても、同様の処理を行うことが可能である。すなわち、1回目の計測ではパターン投光部103を第1の明るさに設定して撮影を行い、2回目の計測ではパターン投光部103を第2の明るさに変更して撮影を行う。このとき、「第1の明るさ<第2の明るさ」のように設定し、1回目の計測と2回目の計測でカメラの露光条件を同一にすれば、1回目の計測では鏡面物体や明るい色の物体の3次元情報(視差情報)を得るのに適した画像が得られ、2回目の計測では低反射率の物体や陰になっている部分の3次元情報(視差情報)を得るのに適した画像が得られる。したがって、それぞれの画像ペアでステレオマッチングを行い、得られた視差マップを合成し、最終的な計測結果を生成することにより、反射特性の違いや照明の状況などに対してロバストな3次元計測が実現できる。
上記実施形態は、本発明の構成例を例示的に説明するものに過ぎない。本発明は上記の具体的な形態には限定されることはなく、その技術的思想の範囲内で種々の変形が可能である。例えば上記実施形態では、空間コード化パターン投影方式を例示したが、第2の計測系の測距方式はステレオマッチング以外の方式であれば如何なる方式を採用してもよい。また、上記実施形態では、ステレオマッチングにハッシュ特徴量を利用したが、対応点の類似度評価には他の手法を用いてもよい。例えば、類似度の評価指標としてはSAD(Sum of Absolute Difference)、SSD(Sum of Squared Difference)、NC(Normalized Correlation)などによる左右画像の画素の類似度計算法がある。また、上記実施形態では、参考デプスマップの生成(視差の予測)とステレオマッチングとで共有するカメラの画像を用いたが、それぞれ異なる三次元計測用のカメラの画像を用いてもよい。
(1) 画像ペアを用いたステレオマッチングにより、各画素の座標に距離情報が関連付けられたデータであるデプスマップを生成する画像処理装置(11)であって、
異なる視点から撮影された第1画像及び第2画像からなる画像ペアを取得する画像取得手段(110)と、
ステレオマッチングとは異なる方式により前記第1画像と前記第2画像の間の視差を予測する視差予測手段(112)と、
前記予測視差に基づき、ステレオマッチングにおける対応点の探索範囲を設定する設定手段(115)と、
前記設定された探索範囲に限定して前記第1画像と前記第2画像の間の各画素の対応点を探索し、その探索結果に基づき、各画素の座標に視差情報が関連付けられたデータである視差マップを生成する視差マップ生成手段(116)と、
複数の画像ペアのそれぞれから生成された複数の視差マップを合成することにより合成視差マップを生成する視差マップ合成手段(114)と、
前記合成視差マップの視差情報を距離情報に変換し、前記デプスマップを生成するデプスマップ生成手段(118)と、
を有することを特徴とする画像処理装置(11)。
10:センサユニット
11:画像処理装置
12:対象物
21:第1の計測系
22:第2の計測系
Claims (12)
- 画像ペアを用いたステレオマッチングにより、各画素の座標に距離情報が関連付けられたデータであるデプスマップを生成する画像処理装置であって、
異なる視点から撮影された第1画像及び第2画像からなる画像ペアを取得する画像取得手段と、
ステレオマッチングとは異なる方式により前記第1画像と前記第2画像の間の視差を予測する視差予測手段と、
前記予測視差に基づき、ステレオマッチングにおける対応点の探索範囲を設定する設定手段と、
前記設定された探索範囲に限定して前記第1画像と前記第2画像の間の各画素の対応点を探索し、その探索結果に基づき、各画素の座標に視差情報が関連付けられたデータである視差マップを生成する視差マップ生成手段と、
複数の画像ペアのそれぞれから生成された複数の視差マップを合成することにより合成視差マップを生成する視差マップ合成手段と、
前記合成視差マップの視差情報を距離情報に変換し、前記デプスマップを生成するデプスマップ生成手段と、
を有することを特徴とする画像処理装置。 - 前記視差予測手段は、複数の予測視差を生成し、
前記設定手段は、前記複数の予測視差を合成した合成予測視差を、前記複数の画像ペアの探索範囲の設定に用いる
ことを特徴とする請求項1に記載の画像処理装置。 - 前記視差予測手段は、複数の予測視差を生成し、
前記設定手段は、探索範囲の設定に用いる予測視差を、画像ペアごとに変える
ことを特徴とする請求項1に記載の画像処理装置。 - 前記視差予測手段は、カメラで撮影された画像から前記予測視差の生成を行うものであり、
前記複数の予測視差は、カメラ及び/又は撮影条件を変えて撮影された、異なる画像から生成されたものである
ことを特徴とする請求項2又は3に記載の画像処理装置。 - 前記画像取得手段が、第1の画像ペアと第2の画像ペアを取得し、
前記視差予測手段が、前記第1の画像ペアのうちの第1画像から第1の予測視差を生成するとともに、前記第2の画像ペアのうちの第2画像から第2の予測視差を生成し、
前記設定手段が、前記第1の予測視差と前記第2の予測視差を合成した合成予測視差を、前記第1の画像ペア及び前記第2の画像ペアの探索範囲の設定に用いる
ことを特徴とする請求項2に記載の画像処理装置。 - 前記画像取得手段が、第1の画像ペアと第2の画像ペアを取得し、
前記視差予測手段が、前記第1の画像ペアのうちの第1画像から第1の予測視差を生成するとともに、前記第2の画像ペアのうちの第2画像から第2の予測視差を生成し、
前記設定手段が、前記第1の予測視差と前記第2の予測視差のうちの一方の予測視差を前記第1の画像ペアの探索範囲の設定に用い、他方の予測視差を前記第2の画像ペアの探索範囲の設定に用いる
ことを特徴とする請求項3に記載の画像処理装置。 - 前記複数の画像ペアは、異なる撮影条件で撮影された2以上の画像ペアを含む
ことを特徴とする請求項1~6のいずれか1項に記載の画像処理装置。 - 前記異なる撮影条件で撮影された2以上の画像ペアは、露光条件及び/又は照明条件を変えて撮影された画像ペアを含む
ことを特徴とする請求項7に記載の画像処理装置。 - 前記視差予測手段は、空間コード化パターン投影方式により得られた距離情報に基づいて、視差を予測する
ことを特徴とする請求項1~8のいずれか1項に記載の画像処理装置。 - 少なくとも2つのカメラを有するセンサユニットと、
前記センサユニットから取り込まれる画像を用いてデプスマップを生成する請求項1~9のいずれか1項に記載の画像処理装置と、
を有することを特徴とする3次元計測システム。 - コンピュータを、請求項1~9のいずれか1項に記載の画像処理装置の各手段として機能させるためのプログラム。
- 画像ペアを用いたステレオマッチングにより、各画素の座標に距離情報が関連付けられたデータであるデプスマップを生成する画像処理方法であって、
異なる視点から撮影された第1画像及び第2画像からなる画像ペアを複数取得するステップと、
前記複数の画像ペアのそれぞれについて、ステレオマッチングとは異なる方式により前記第1画像と前記第2画像の間の視差を予測するステップと、
前記予測視差に基づき、ステレオマッチングにおける対応点の探索範囲を設定するステップと、
前記複数の画像ペアのそれぞれについて、前記設定された探索範囲に限定して前記第1画像と前記第2画像の間の各画素の対応点を探索し、その探索結果に基づき、各画素の座標に視差情報が関連付けられたデータである視差マップを生成するステップと、
複数の画像ペアのそれぞれから生成された複数の視差マップを合成することにより合成視差マップを生成するステップと、
前記合成視差マップの視差情報を距離情報に変換し、前記デプスマップを生成するステップと、
を有することを特徴とする画像処理方法。
Priority Applications (5)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/JP2019/010590 WO2020183711A1 (ja) | 2019-03-14 | 2019-03-14 | 画像処理装置及び3次元計測システム |
| EP19919470.5A EP3832601B1 (en) | 2019-03-14 | 2019-03-14 | Image processing device and three-dimensional measuring system |
| CN201980054085.2A CN112602118B (zh) | 2019-03-14 | 2019-03-14 | 图像处理装置和三维测量系统 |
| JP2021505461A JP7078173B2 (ja) | 2019-03-14 | 2019-03-14 | 画像処理装置及び3次元計測システム |
| US17/272,919 US11803982B2 (en) | 2019-03-14 | 2019-03-14 | Image processing device and three-dimensional measuring system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/JP2019/010590 WO2020183711A1 (ja) | 2019-03-14 | 2019-03-14 | 画像処理装置及び3次元計測システム |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2020183711A1 true WO2020183711A1 (ja) | 2020-09-17 |
Family
ID=72426687
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2019/010590 Ceased WO2020183711A1 (ja) | 2019-03-14 | 2019-03-14 | 画像処理装置及び3次元計測システム |
Country Status (5)
| Country | Link |
|---|---|
| US (1) | US11803982B2 (ja) |
| EP (1) | EP3832601B1 (ja) |
| JP (1) | JP7078173B2 (ja) |
| CN (1) | CN112602118B (ja) |
| WO (1) | WO2020183711A1 (ja) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022091578A1 (ja) * | 2020-10-26 | 2022-05-05 | オムロン株式会社 | 制御装置、ロボット、制御方法、プログラム |
| EP4310784A1 (en) | 2022-07-21 | 2024-01-24 | Canon Kabushiki Kaisha | Image processing apparatus, image processing method, and program |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11989896B2 (en) * | 2019-11-27 | 2024-05-21 | Trinamix Gmbh | Depth measurement through display |
| WO2021177163A1 (ja) * | 2020-03-05 | 2021-09-10 | ファナック株式会社 | 複数のカメラにて撮像された画像に基づいて物体の表面の位置情報を生成する三次元測定装置 |
| JP7398749B2 (ja) * | 2021-12-06 | 2023-12-15 | 国立大学法人東北大学 | 3次元形状計測方法及び3次元形状計測装置 |
| EP4332499A1 (en) * | 2022-08-30 | 2024-03-06 | Canon Kabushiki Kaisha | Three-dimensional measuring apparatus, three-dimensional measuring method, storage medium, system, and method for manufacturing an article |
| CN117367331B (zh) * | 2023-12-04 | 2024-03-12 | 山西阳光三极科技股份有限公司 | 一种矿区地表形变的雷达监测方法、装置以及电子设备 |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH05303629A (ja) | 1991-11-29 | 1993-11-16 | Nec Corp | 形状合成方法 |
| JP2011013706A (ja) * | 2009-06-30 | 2011-01-20 | Hitachi Ltd | ステレオ画像処理装置およびステレオ画像処理方法 |
| JP2012248221A (ja) | 2012-09-07 | 2012-12-13 | Casio Comput Co Ltd | 三次元モデリング装置、三次元モデリング方法、ならびに、プログラム |
| US20140153816A1 (en) * | 2012-11-30 | 2014-06-05 | Adobe Systems Incorporated | Depth Map Stereo Correspondence Techniques |
| US20150178936A1 (en) * | 2013-12-20 | 2015-06-25 | Thomson Licensing | Method and apparatus for performing depth estimation |
| JP2017045283A (ja) * | 2015-08-26 | 2017-03-02 | 株式会社ソニー・インタラクティブエンタテインメント | 情報処理装置および情報処理方法 |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2418314A (en) * | 2004-09-16 | 2006-03-22 | Sharp Kk | A system for combining multiple disparity maps |
| US7561731B2 (en) * | 2004-12-27 | 2009-07-14 | Trw Automotive U.S. Llc | Method and apparatus for enhancing the dynamic range of a stereo vision system |
| JP5870510B2 (ja) * | 2010-09-14 | 2016-03-01 | 株式会社リコー | ステレオカメラ装置、校正方法およびプログラム |
| JP5158223B2 (ja) | 2011-04-06 | 2013-03-06 | カシオ計算機株式会社 | 三次元モデリング装置、三次元モデリング方法、ならびに、プログラム |
| JP2012257198A (ja) * | 2011-05-17 | 2012-12-27 | Canon Inc | 立体画像符号化装置、その方法、および立体画像符号化装置を有する撮像装置 |
| US9762881B2 (en) * | 2011-11-03 | 2017-09-12 | Texas Instruments Incorporated | Reducing disparity and depth ambiguity in three-dimensional (3D) images |
| CN103868460B (zh) * | 2014-03-13 | 2016-10-05 | 桂林电子科技大学 | 基于视差优化算法的双目立体视觉自动测量方法 |
| JP6805534B2 (ja) * | 2015-07-02 | 2020-12-23 | 株式会社リコー | 視差画像生成装置、視差画像生成方法及び視差画像生成プログラム、物体認識装置、機器制御システム |
| KR101690645B1 (ko) * | 2015-09-21 | 2016-12-29 | 경북대학교 산학협력단 | 다단계 시차영상 분할이 적용된 시차탐색범위 추정 방법 및 이를 이용한 스테레오 영상 정합장치 |
| JP6899673B2 (ja) * | 2017-03-15 | 2021-07-07 | 日立Astemo株式会社 | 物体距離検出装置 |
| EP3832600B1 (en) * | 2019-03-14 | 2025-09-17 | OMRON Corporation | Image processing device and three-dimensional measuring system |
-
2019
- 2019-03-14 US US17/272,919 patent/US11803982B2/en active Active
- 2019-03-14 EP EP19919470.5A patent/EP3832601B1/en active Active
- 2019-03-14 WO PCT/JP2019/010590 patent/WO2020183711A1/ja not_active Ceased
- 2019-03-14 CN CN201980054085.2A patent/CN112602118B/zh active Active
- 2019-03-14 JP JP2021505461A patent/JP7078173B2/ja active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH05303629A (ja) | 1991-11-29 | 1993-11-16 | Nec Corp | 形状合成方法 |
| JP2011013706A (ja) * | 2009-06-30 | 2011-01-20 | Hitachi Ltd | ステレオ画像処理装置およびステレオ画像処理方法 |
| JP2012248221A (ja) | 2012-09-07 | 2012-12-13 | Casio Comput Co Ltd | 三次元モデリング装置、三次元モデリング方法、ならびに、プログラム |
| US20140153816A1 (en) * | 2012-11-30 | 2014-06-05 | Adobe Systems Incorporated | Depth Map Stereo Correspondence Techniques |
| US20150178936A1 (en) * | 2013-12-20 | 2015-06-25 | Thomson Licensing | Method and apparatus for performing depth estimation |
| JP2017045283A (ja) * | 2015-08-26 | 2017-03-02 | 株式会社ソニー・インタラクティブエンタテインメント | 情報処理装置および情報処理方法 |
Non-Patent Citations (1)
| Title |
|---|
| P. VUYLSTEKEA. OOSTERLINCK: "Range Image Acquisition with a Single Binary-Encoded Light Pattern", IEEE PAMI, vol. 12, no. 2, 1990, pages 148 - 164 |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022091578A1 (ja) * | 2020-10-26 | 2022-05-05 | オムロン株式会社 | 制御装置、ロボット、制御方法、プログラム |
| JP7512839B2 (ja) | 2020-10-26 | 2024-07-09 | オムロン株式会社 | 制御装置、ロボット、制御方法、プログラム |
| US12377547B2 (en) | 2020-10-26 | 2025-08-05 | Omron Corporation | Control device, robot, control method, and program |
| EP4310784A1 (en) | 2022-07-21 | 2024-01-24 | Canon Kabushiki Kaisha | Image processing apparatus, image processing method, and program |
| US12620111B2 (en) | 2022-07-21 | 2026-05-05 | Canon Kabushiki Kaisha | Image processing apparatus, image processing method, and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| US20220398760A1 (en) | 2022-12-15 |
| CN112602118A (zh) | 2021-04-02 |
| EP3832601B1 (en) | 2025-09-03 |
| US11803982B2 (en) | 2023-10-31 |
| EP3832601A1 (en) | 2021-06-09 |
| CN112602118B (zh) | 2024-06-07 |
| JP7078173B2 (ja) | 2022-05-31 |
| EP3832601A4 (en) | 2022-03-16 |
| JPWO2020183711A1 (ja) | 2021-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP7078173B2 (ja) | 画像処理装置及び3次元計測システム | |
| JP7078172B2 (ja) | 画像処理装置及び3次元計測システム | |
| US9392262B2 (en) | System and method for 3D reconstruction using multiple multi-channel cameras | |
| JP7168077B2 (ja) | 3次元計測システム及び3次元計測方法 | |
| JP7184203B2 (ja) | 画像処理装置、3次元計測システム、画像処理方法 | |
| WO2019230813A1 (ja) | 三次元再構成方法および三次元再構成装置 | |
| JP7607229B2 (ja) | 三次元変位計測方法及び三次元変位計測装置 | |
| EP3688407B1 (en) | Light projection systems | |
| JP2016217833A (ja) | 画像処理システム及び画像処理方法 | |
| JP2026034619A (ja) | 撮像装置および撮像処理方法 | |
| JP2021018219A (ja) | 3次元形状計測システム、3次元形状計測方法、及び3次元形状計測プログラム | |
| JPH11183142A (ja) | 三次元画像撮像方法及び三次元画像撮像装置 | |
| JP7581638B2 (ja) | 情報処理装置および情報処理方法 | |
| JP2016223980A (ja) | 情報処理装置、情報処理方法及びプログラム | |
| WO2023084790A1 (ja) | 3次元計測システム、並びにその制御方法及び制御プログラム |
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: 19919470 Country of ref document: EP Kind code of ref document: A1 |
|
| ENP | Entry into the national phase |
Ref document number: 2021505461 Country of ref document: JP Kind code of ref document: A |
|
| ENP | Entry into the national phase |
Ref document number: 2019919470 Country of ref document: EP Effective date: 20210304 |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |
|
| WWG | Wipo information: grant in national office |
Ref document number: 2019919470 Country of ref document: EP |