WO2024084675A1 - 判定装置 - Google Patents
判定装置 Download PDFInfo
- Publication number
- WO2024084675A1 WO2024084675A1 PCT/JP2022/039227 JP2022039227W WO2024084675A1 WO 2024084675 A1 WO2024084675 A1 WO 2024084675A1 JP 2022039227 W JP2022039227 W JP 2022039227W WO 2024084675 A1 WO2024084675 A1 WO 2024084675A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- judgment
- pass
- fail
- good
- judged
- 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
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B19/00—Program-control systems
- G05B19/02—Program-control systems electric
- G05B19/418—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM]
- G05B19/41875—Total factory control, i.e. centrally controlling a plurality of machines, e.g. direct or distributed numerical control [DNC], flexible manufacturing systems [FMS], integrated manufacturing systems [IMS] or computer integrated manufacturing [CIM] characterised by quality surveillance of production
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/93—Detection standards; Calibrating baseline adjustment, drift correction
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0006—Industrial image inspection using a design-rule based approach
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/001—Industrial image inspection using an image reference approach
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/60—Extraction of image or video features relating to illumination properties, e.g. using a reflectance or lighting model
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N2201/00—Features of devices classified in G01N21/00
- G01N2201/10—Scanning
- G01N2201/104—Mechano-optical scan, i.e. object and beam moving
- G01N2201/1042—X, Y scan, i.e. object moving in X, beam in Y
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B2219/00—Program-control systems
- G05B2219/30—Nc systems
- G05B2219/32—Operator till task planning
- G05B2219/32201—Build statistical model of past normal proces, compare with actual process
Definitions
- An embodiment of the present invention relates to technology for determining whether an object is good or bad.
- the aim is to provide a determination device that can improve determination accuracy.
- the judgment device of the embodiment judges pass/fail using monitoring data of the judgment object.
- the judgment device has a storage unit that stores pass/fail judgment feature information generated using feature amounts extracted from each of the previously collected monitoring data of judgment objects belonging to a pass/fail judgment, a judgment unit that extracts the feature amounts from the monitoring data of each judgment object input in sequence and performs pass/fail judgments on multiple judgment objects based on the pass/fail feature information, and a criteria control unit that updates the pass/fail judgment feature information using the feature amounts of the judgment object judged as pass when the judgment object is judged as pass in the successive pass/fail judgments.
- the judgment unit performs pass/fail judgments on the judgment objects subsequent to the one judged as pass based on the updated pass/fail feature information.
- FIG. 1 is a functional block diagram of a determination system including a determination device according to a first embodiment.
- 5A to 5C are explanatory diagrams illustrating a pass/fail judgment process performed on image data obtained by photographing an object to be judged in the first embodiment.
- 11 is an explanatory diagram of a quality determination process to which the quality determination function of the first embodiment is applied;
- FIG. 1 is an explanatory diagram of a conventional quality determination process.
- FIG. 11 is a diagram showing a flow of a process for generating good judgment feature information according to the first embodiment.
- 11 is a diagram showing a flow of pass/fail determination processing including update processing of pass/fail determination feature information in the first embodiment
- 13A to 13C are diagrams for explaining a first modified example of the update process of the good judgment feature information in the first embodiment.
- 13A to 13C are diagrams for explaining a second modified example of the update process of the good judgment feature information in the first embodiment.
- First Embodiment 1 to 8 are diagrams for explaining the determination system of the first embodiment.
- FIG. 1 is a functional block diagram of a judgment system including a judgment device 100 according to this embodiment.
- the judgment system includes the judgment device 100 and a judgment criterion generating device 200, and the judgment device 100 judges the pass/fail of an object to be judged.
- the judgment system of this embodiment can be used, for example, for visual inspection of an object.
- the monitoring device K is a photographing device that photographs an object, and the monitoring data is photographed image data.
- the judgment system of this embodiment will be explained using the visual inspection of an object as an example.
- the photographing device K photographs the item A to be inspected (evaluated) and outputs the image data to the judgment device 100.
- the judgment device 100 stores the acquired image data in the storage device 130.
- the photographing device K can acquire still or moving images of the item A and output them to the judgment device 100. In the case of moving images, a still image cut out from the moving image is used as the image data of the item A.
- the direction in which the item A is photographed is appropriately set so that image data suitable for the inspection is obtained. For example, multiple items A are transported sequentially by the transport device H.
- the photographing device K is fixed and photographs the multiple items A in sequence as they are transported, and the image data of each of the multiple items is input sequentially to the judgment device 100.
- the judgment device 100 is configured to judge the quality of a plurality of different objects that are conveyed continuously in time series, and the quality judgment is performed continuously.
- the quality judgment result of the judgment device 100 is output to the display device D.
- the judgment device 100 includes a judgment unit 110, a criterion control unit 120, and a storage unit 130.
- the judgment criterion generating device 200 includes a judgment characteristic information generating unit 210 and a storage unit 220.
- FIG. 2 is an explanatory diagram of the pass/fail judgment process performed using image data of an object to be judged.
- the pass/fail judgment of item A is judged based on pass/fail characteristic information.
- feature amounts are extracted from the image data of items that are judged to be passable, out of the image data of multiple different items A that have been collected in advance, and pass/fail characteristic information is generated based on the extracted feature amounts; image data of items that are judged to be failing is not used.
- the average pixel value (average brightness value) is calculated at the same pixel position in the image data of multiple items that are judged to be good.
- the pixel value at each pixel position in the image data is a feature (feature information)
- the good judgment feature information is information that includes the average pixel value (average brightness value) at each pixel position.
- a range within which this average pixel value can be judged as good is set; for example, the ranges of average pixel value + predetermined value (first threshold) x1 and average pixel value - predetermined value (second threshold) x2 are set for each pixel position.
- first threshold predetermined value
- second threshold predetermined value
- upper and lower limits based on the average pixel value are set for each pixel position.
- the predetermined values x1 and x2 may be different values for each pixel position, or may be the same value for each of multiple pixel positions.
- the predetermined values x1 and x2 can be set based on the distribution range (variance or deviation) of pixel values of each image data of an item that is judged to be good.
- the good/bad judgment process it is judged whether the pixel value (feature value) of each pixel position in the image data of the item A to be judged belongs to the good/bad judgment range set by the first and second thresholds (whether it is below the upper limit or above the lower limit), and the good/bad judgment result is output.
- the good judgment characteristic information is, as an example, statistical information (statistical good judgment characteristic information) set for each pixel position by the average pixel value of each image data that is judged to be good and the distribution range of pixel values of each image data of items that are judged to be good relative to the average pixel value.
- the process of generating good judgment feature information is performed by the judgment criterion generating device 200 (judgment feature information generating section 210), and multiple image data for generating good judgment feature information is stored in the storage section 220.
- the good judgment feature information generated by the judgment criterion generating device 200 is provided to the judgment device 100 as initial good judgment feature information and stored in the storage device 130.
- the judgment unit 110 of the judgment device 100 extracts features from the image data of the first object A1.
- the features are pixel values (brightness values) at each pixel position of the image data.
- the judgment unit 110 compares the features extracted from the image data with the pass/fail judgment feature information to make a pass/fail judgment. For example, a pass/fail judgment process is performed to judge whether the pixel value of the image data is within the range of the pass/fail judgment feature information for each pixel position.
- the determination device 100 stores the pass/fail determination result of the item A1 as a log in the memory unit 130 and outputs it to the display device D.
- the display device D displays the pass/fail determination result output from the determination device 100 so that the user can visually confirm it.
- the judgment process is performed for the subsequent item A2.
- the initial pass/fail judgment characteristic information used for item A1 is not used for the pass/fail judgment process for item A2, and updated pass/fail judgment characteristic information that reflects the characteristic amounts of item A1, which was judged to be good, in the initial pass/fail judgment characteristic information is used.
- the average value between the pixel value of the image data of item A1 and the average pixel value of the initial good judgment characteristic information is calculated, and the good judgment characteristic information is updated to have the calculated average pixel value for each pixel position (good judgment characteristic information that takes into account the characteristic values of item A1).
- the update process of the good judgment characteristic information is performed by the reference control unit 120, and the updated good judgment characteristic information is stored in the memory unit 130.
- the judgment unit 110 extracts the features of the image data of item A2, reads the updated pass/fail judgment feature information from the storage unit 130, and performs pass/fail judgment processing. Furthermore, if the pass/fail judgment result for item A2 is "pass", the standard control unit 120 calculates the average value of the pixel values of the image data of item A2 and the average pixel value of the updated pass/fail judgment feature information, and updates the pass/fail judgment feature information to include the features of item A2.
- the pass/fail judgment feature information at this point is pass/fail judgment feature information that includes the features of both item A1 and item A2, and is applied to the pass/fail judgment processing for item A3.
- the predetermined values x1 and x2 constituting the good judgment characteristic information can be configured to carry over the values set in the initial good judgment characteristic information as is, and not change before and after the update. Also, the predetermined values x1 and x2 may be configured to change according to the amount of fluctuation in the average pixel value after the characteristic values of the articles judged to be good are added to the average pixel value.
- the judgment device 100 extracts features from the image data of each item that is input in sequence, and continuously judges the quality of multiple items based on the quality judgment characteristic information.
- the judgment device 100 is equipped with a standard control unit 120, and when item A1 is judged to be good in the successive quality judgments, the quality judgment characteristic information, which is the judgment standard, is updated using the feature values of item A1 that was judged to be good.
- the quality judgment of the next item A2 traveling on the conveying device H is controlled to be made based on the updated quality judgment characteristic information (quality judgment characteristic information that is different from the quality judgment characteristic information used in the immediately preceding quality judgment).
- the quality of the items that will be judged as good or bad will be judged based on the quality judgment characteristic information that has been updated according to the characteristics of items that have been judged as good or bad in the past.
- the process of updating the pass/fail judgment characteristic information is controlled so that it reflects the characteristic amounts of items with a pass/fail judgment result of "pass", but does not reflect the characteristic amounts of items with a pass/fail judgment result of "fail”. For example, if the pass/fail judgment result of item A2 is "fail", the process of updating the pass/fail judgment characteristic information is not performed, and the pass/fail judgment process of item A3 is performed using the updated pass/fail judgment characteristic information that takes into account only the characteristic amounts of item A1. Therefore, if the pass/fail judgment results continue to be "fail" in the pass/fail judgment processes that are performed sequentially, the pass/fail judgment characteristic information during that time is not updated and does not change.
- FIG 3 is an explanatory diagram of the pass/fail judgment process that applies the pass/fail judgment function of this embodiment.
- Figure 4 is an explanatory diagram of the conventional pass/fail judgment process.
- the initial good judgment feature information is composed of a group of judgment targets that belong to a good judgment, indicated by white circles.
- the good judgment feature information has a good judgment range (the range of image data features that can be judged as a good judgment) set, so for example, in the example of Figure 3, the average value of the good judgment feature information becomes the reference axis for a good judgment, and a width (predetermined values x1, x2) within which a judgment can be made as a good judgment is set relative to the reference axis.
- the good judgment feature information can be expressed as a range of feature quantities with a predetermined width, with the average value as the reference axis, and in this embodiment, the reference axis of the good judgment feature information is shifted in real time by the update process of the good judgment feature information.
- the good judgment feature information is represented by gray rectangular blocks. Note that one rectangular block shows a group of multiple judgment targets (white circles) that have been judged to be good, but this is a simplified representation of the group of judgment targets that belong to the rectangular block.
- the pass/fail judgment feature information is updated according to the feature value of the previous judgment target, the reference axis fluctuates in real time over time as shown by the thick dotted line, and the pass/fail judgment feature information (pass/fail judgment range) shifts to track the feature value of the judgment target that was judged to be good, and is applied to the next pass/fail judgment process.
- the conventional example shown in Figure 4 is a method for expanding the range of pass/fail judgment feature information.
- This conventional method performs pass/fail judgment for a certain period of time based on the initial pass/fail judgment feature information, and extracts judgment targets that are judged as "fail" during that period because they fall outside the range of the initial pass/fail judgment feature information, but are actually judged as pass/fail, i.e., judgment targets that are overdetected.
- the initial good judgment feature information is updated so that the overdetected judgment target is judged as good, a pass/fail judgment is made for a certain period of time based on the updated good judgment feature information, the overdetected judgment target is extracted during that period, and the good judgment feature information is updated. This process is repeated.
- the first is that because pass/fail judgment processing is performed for a certain period of time based on the same pass/fail feature information, if the environment changes during that time, overdetection will occur frequently. For example, if the amount of light hitting the object to be judged changes due to the season or deterioration of equipment over time, the feature amount extracted from the image data will change even for objects that belong to the pass/fail judgment category. For this reason, it is vulnerable to environmental changes during pass/fail judgment processing (during inspection), and there is an issue of overdetection occurring frequently in response to environmental changes, resulting in reduced judgment accuracy.
- the second is to update the good judgment feature information based on the feature values of the monitoring data of the overdetected judgment target, but as shown in Figure 4, the reference axis of the good judgment feature information after the update is the same as the reference axis of the good judgment feature information before the update, and the range of good judgments is expanded so that overdetected judgment targets are judged as good.
- the reference axis for good judgments does not change, and the same reference axis remains after the update, expanding the area where good judgments are made, resulting in an increase in oversights.
- overdetection is when a good product is judged to be defective, and the conventional method described above expands the range of pass judgments, making it possible to suppress overdetection.
- the wider the range of pass judgments the more likely a product is to be judged as good, and oversights increase.
- Oversights are when a defective product is judged to be good, and narrowing the range of pass judgments results in more overdetections, while widening the range of pass judgments results in more oversights, with the two being inversely proportional.
- the characteristic quantities of the judgment object are updated by reflecting them in the pass/fail judgment characteristic information, and the updated pass/fail judgment characteristic information is applied to the subsequent judgment object. For example, if the weather improves and the sun shines, changing the illuminance on the judgment object being inspected, this change in the environment is reflected in the characteristic quantities of the judgment object that was judged as pass. Therefore, the pass/fail judgment of the subsequent judgment object is performed based on the updated pass/fail judgment characteristic information that corresponds to the change in the environment, making it possible to suppress overdetection due to changes in the environment.
- FIG. 5 is a diagram showing the flow of the process for generating pass/fail judgment characteristic information in this embodiment.
- FIG. 6 is a diagram showing the flow of the pass/fail judgment process including the process of updating the pass/fail judgment characteristic information.
- the storage unit 220 of the judgment criterion generating device 200 stores the monitoring data of the judgment target belonging to the good judgment among the monitoring data of the judgment targets collected in advance (S201). Then, the judgment characteristic information generating unit 210 extracts the feature amount necessary for judging the good or bad of the judgment target from each monitoring data (for example, the luminance value of each pixel position in the case of image data) (S202). The judgment characteristic information generating unit 210 calculates the average value of each extracted feature amount, and generates the good judgment characteristic information centered on the average value of each feature amount (S203). At this time, the good judgment characteristic information can set the range in which the judgment can be judged as good for the average value.
- the judgment characteristic information generating unit 210 stores the generated good judgment characteristic information in the storage unit 220 as the initial good judgment characteristic information (S204).
- the initial good judgment characteristic information is provided to the judgment device 100 at any time.
- the pass/fail judgment process of this embodiment is a continuous process in chronological order for each of the multiple judgment targets (see FIG. 2).
- the judgment device 100 accepts input of monitoring data, and executes pass/fail judgment processing with the input of the monitoring data as a trigger.
- the judgment device 100 extracts features from the input monitoring data (S102). Then, it compares the pass/fail judgment feature information (initial pass/fail judgment feature information) with the feature information of the judgment target to make a pass/fail judgment (S103). If the pass/fail judgment result is "pass” (YES in S104), the judgment device 100 performs an update process of the pass/fail judgment feature information (S105), and if the pass/fail judgment result is "fail", the judgment device 100 does not perform an update process of the pass/fail judgment feature information. The judgment device 100 outputs the pass/fail judgment result to the display device D (S106). The pass/fail judgment result may also be stored in the storage unit 130 as a log.
- the determination device 100 of this embodiment determines whether or not the next monitoring data for the determination target has been input in chronological order (S107), and if the next monitoring data has been input (YES in S107), returns to step S102 and performs pass/fail determination processing (S103). At this time, if the result of the previous pass/fail determination is "good", the pass/fail determination characteristic information has been updated, so the determination device 100 performs pass/fail determination processing based on the updated pass/fail determination characteristic information in step S103.
- step S107 If it is determined in step S107 that no further monitoring data has been input (NO in S107), the presence or absence of an end instruction to end the pass/fail determination process is checked, and if an end instruction has been received (YES in S108), the pass/fail determination process is terminated, and if no end instruction has been received (YES in S108), the process waits until the next monitoring data is input.
- the pass/fail judgment characteristic information can be reset and the initial pass/fail judgment characteristic information can be applied, or the pass/fail judgment characteristic information most recently updated in the previous pass/fail judgment process (the most recent updated pass/fail judgment characteristic information) can be applied.
- the image data input to the judgment device 100 can be subjected to various image processing before being compared with the good judgment feature information.
- the image data can be subjected to smoothing processing such as noise removal (Median filter, average value filter, etc.) and the pixel values at each pixel position can be extracted to extract the feature values of the image data.
- the good judgment feature information is "statistical good judgment image data" based on each feature value of each of the multiple image data items that belong to a good judgment. Therefore, the above-mentioned predetermined image processing is also applied when creating the good judgment feature information, and the feature values are compared in a state where the same image processing has been applied.
- the pass/fail judgment process described above judges the result to be "pass” if the feature value of the object to be judged does not exceed the range specified in the pass/fail judgment feature information, but the judgment process can be further subdivided.
- the pass/fail judgment process can be configured in two stages: whether the object to be judged has an abnormality, and if so, whether the abnormality exceeds a predetermined tolerance range.
- the good judgment feature information can include first good judgment feature information for determining whether or not the judgment target has an abnormality, and second good judgment feature information for determining whether or not the abnormality is within a predetermined acceptable range.
- the judgment criterion generating device 200 generates the first good judgment feature information and the second good judgment feature information from the features of the judgment target that belong to a good judgment.
- the pixel value of each pixel position of the image data is checked to determine whether or not there are any pixels that exceed the threshold (first judgment process).
- the good judgment characteristic information used at this time corresponds to the first good judgment characteristic information, and is the average pixel value + predetermined value (first threshold) x1 and the average pixel value - predetermined value (second threshold) x2 of each pixel position composed of the image data group belonging to the good judgment described above.
- the pass/fail judgment characteristic information used at this time corresponds to the second pass/fail judgment characteristic information.
- the judgment unit 110 compares each of the multiple pixel values in the image data with the first pass/fail judgment characteristic information and extracts abnormal pixels that deviate from the threshold, so that it is possible to grasp the extracted group of abnormal pixels, i.e., the abnormal area.
- a second judgment process is performed to judge whether the abnormal area identified through the first pass/fail judgment process is within the range of pass/fail judgment using the characteristics of the abnormal area, such as the area of the abnormal area, the shape of the abnormal area, and the distribution of pixel values in the abnormal area.
- the information on the characteristics of the abnormal area at this time becomes second pass/fail judgment characteristic information. For example, if the area is used as the characteristic of the abnormal area, a threshold value for the area is set in advance. The area is represented by the number of pixels contained in the abnormal area. If the area of the abnormal area does not exceed the threshold, the judgment device 100 judges item A to be "pass”; if it does exceed the threshold, the judgment device 100 judges item A to be "fail.”
- the aspect ratio is used as the shape.
- the aspect ratio is the ratio of the length in the long axis direction to the length in the short axis direction.
- the long axis direction is the direction parallel to the longest line segment obtained by connecting any two points on the outer edge of the particle.
- the short axis direction is perpendicular to the long axis direction.
- the determination device 100 determines the article as "bad” when the aspect ratio exceeds the threshold, and determines the article as "good” when the aspect ratio does not exceed the threshold.
- the distribution for example, the ratio of the number of pixels below a lower threshold to the number of pixels in the abnormal region, or the ratio of the number of pixels above an upper threshold to the number of pixels in the abnormal region can be used.
- a threshold is set for each of these ratios. For example, if each of these ratios does not exceed a threshold, the determination device 100 determines the item as "good,” and if either one of these ratios exceeds the corresponding threshold, the determination device 100 determines the item as "no good.”
- step S103 of FIG. 6 the judgment unit 110 judges the product as "good” when the first judgment process judges that there is no abnormality, or when the first judgment process judges that there is an abnormality but the second good judgment characteristic information is satisfied in the second judgment process.
- the criteria control unit 120 can update at least the first good judgment characteristic information using the characteristic amount of the object judged as good, and can control so that the updated first good judgment characteristic information and the second good judgment characteristic information can be applied when judging the next object to be good or bad.
- FIG. 7 is a diagram for explaining a first modified example of the process for updating the pass/fail judgment characteristic information.
- the pass/fail judgment characteristic information is updated by taking into account not only the characteristic values of the judgment object A-1 that was judged to be pass/fail just before, but also the characteristic values of the judgment objects A-2, A-3, A-4, ..., A-N that were judged to be pass/fail in the past before the judgment object A-1.
- the number of judgment objects judged to be pass/fail in the past to be taken into account is arbitrary.
- time information may be used to trigger the judgment result of pass/fail being judged to be "pass” to extract the characteristic values of past judgment objects that were judged to be pass/fail up to a certain time ago (several tens of seconds ago, several minutes ago, etc.) and update the pass/fail judgment characteristic information.
- the good judgment feature information can be updated so that it includes only the feature values of objects judged as good, without taking into account (excluding) the feature values of objects judged as "bad” that are included in between.
- the criteria control unit 120 can update the pass/fail judgment characteristic information by adding and averaging the characteristic amounts of the multiple objects A-2, A-3, A-4, ..., A-N judged as good before the object A-1 that was judged as good to the pass/fail judgment characteristic information used for the pass/fail judgment.
- the judgment unit 110 can then make a pass/fail judgment on the subsequent object A based on the updated pass/fail judgment characteristic information in which the average value has been shifted by the characteristic amounts of the multiple objects A-1, A-2, A-3, A-4, ..., A-N that were judged as good in the past.
- FIG. 8 is a diagram for explaining a second modified example of the process of updating good judgment feature information.
- the second modified example is an embodiment in which weight values are applied to the first modified example.
- the criteria control unit 120 applies weighting values ⁇ 1, ⁇ 2, ⁇ 3, ⁇ 4, ... ⁇ n to the features of the multiple objects A-1, A-2, A-3, A-4, ... A-N that were judged to be good before the object A-1 that was judged to be good, and adds and averages the features to which these weighting values have been applied, thereby updating the good judgment feature information.
- the weights can be assigned arbitrarily so that the sum of the weights is "1". For example, the largest value can be assigned to weight value ⁇ 1 of the immediately preceding judgment target A-1, and the weights can be assigned so that the weights decrease each time one goes back in time.
- the update process of the good judgment characteristic information may be performed each time a good judgment is made, or may be configured to be performed when a predetermined number of good judgments are made in succession.
- the frequency of the update process can be set arbitrarily so as to be able to follow changes in the environment, and the update process of the good judgment characteristic information is performed when a good judgment is made.
- the appearance inspection of an object has been taken as an example to describe pass/fail judgment using a photographed image of the object as monitoring data, but for example, sound data measured by a sensor device or a sensor value can be applied as the monitoring data for the object.
- sound data collected by a non-destructive acoustic inspection can be applied as monitoring data to configure a judgment device 100 that judges whether the internal condition of an object is pass/fail (presence or absence of cracks, etc.).
- the pass/fail judgment characteristic information can be generated, for example, based on the wavelength range of each sound collected from a judgment target that belongs to a pass/fail judgment.
- the device can be configured as a judgment device 100 that collects height values of the cross-sectional shape using a distance sensor and judges whether the condition of the surface shape of the article is good or bad.
- the good judgment characteristic information can be generated, for example, based on each sensor value collected from a judgment target that belongs to a good judgment.
- the judgment device 100 can use 3D image data (stereoscopic image data) such as 3D ultrasound image data as monitoring data.
- 3D ultrasound image acquisition device 3D echo (3D ultrasound) inspection device
- the good judgment characteristic information can be generated based on 3D image data collected from judgment targets that belong to a good judgment.
- the judgment object can be, for example, an item that is manufactured in a continuous flow (rolled steel material, steel material such as H-shaped steel or I-shaped steel, or long members such as bar-shaped rebar).
- the judgment object is divided in the longitudinal direction and photographed continuously. Then, each successive image data obtained by dividing and photographing is input as monitoring data, and a judgment device 100 can be configured to judge the quality of the long item for each longitudinal portion.
- the judgment device 100 and the judgment criterion generating device 200 may be configured as a single device.
- the judgment device 100 can also be configured as a cloud-based service provision form. In other words, it is configured to transmit monitoring data (e.g., image data) output from the monitoring device K to the judgment device 100 over an IP network. This makes it possible to perform pass/fail judgment processing on the cloud side and transmit the pass/fail judgment result to the display device D, a monitoring terminal, monitoring equipment, etc. over the IP network.
- Both the judgment device 100 and the judgment criterion generating device 200 can be configured as a cloud-based service provision form.
- the judgment device 100 and judgment criterion generating device 200 are computer devices equipped with the calculation function, storage function, communication function, etc. of a server device, etc.
- the hardware configuration may include a memory (main storage device), operation input means such as a mouse, keyboard, touch panel, etc., output means such as a printer, auxiliary storage device (hard disk, etc.), etc.
- each function of the present invention can be realized by a program, and a computer program prepared in advance to realize each function is stored in an auxiliary storage device, and a control unit such as a CPU reads the program stored in the auxiliary storage device into a main storage device, and the control unit executes the program read into the main storage device, thereby causing the computer to operate the functions of each unit of the present invention.
- each function of device 100 can be configured as an individual device, or a computer system can be configured by connecting multiple devices directly or via a network.
- the above program can also be provided to a computer in a state in which it is recorded on a computer-readable recording medium.
- computer-readable recording media include optical discs such as CD-ROMs, phase-change optical discs such as DVD-ROMs, magneto-optical discs such as MO (Magnet Optical) and MD (Mini Disk), magnetic discs such as floppy (registered trademark) disks and removable hard disks, and memory cards such as Compact Flash (registered trademark), Smart Media, SD memory cards, and memory sticks.
- hardware devices such as integrated circuits (IC chips, etc.) specially designed and configured for the purposes of the present invention are also included as recording media.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- General Engineering & Computer Science (AREA)
- Manufacturing & Machinery (AREA)
- Automation & Control Theory (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Analysis (AREA)
Abstract
Description
図1から図8は、第1実施形態の判定システムを説明するための図である。
110 判定部
120 基準制御部
130 記憶部
200 判定基準生成装置
210 判定特徴情報生成部
220 記憶部
A 判定対象
D 表示装置
H 搬送装置
K 監視装置
Claims (8)
- 判定対象の監視データを用いて良否を判定する判定装置であって、
予め収集された良判定に属する判定対象の監視データそれぞれから抽出される特徴量を用いて生成される、良判定特徴情報を記憶する記憶部と、
順次入力される各判定対象の監視データから前記特徴量を抽出し、前記良判定特徴情報に基づいて複数の判定対象に対する良否判定を連続して行う判定部と、
連続して行われる前記良否判定において判定対象が良判定された場合に、良判定された判定対象の前記特徴量を用いて前記良判定特徴情報を更新する基準制御部と、を有し、
前記判定部は、更新された前記良判定特徴情報に基づいて、当該良判定された後に続く判定対象に対する良否判定を行うことを特徴とする判定装置。 - 前記良判定特徴情報は、良判定された判定対象の監視データそれぞれから抽出される特徴量の平均値を中心とした情報であり、
前記基準制御部は、連続して行われる前記良否判定において判定対象が良判定された場合に、当該良判定に用いられた前記良判定特徴情報の平均値に、当該良判定された判定対象の前記特徴量を加算して平均し、前記良判定特徴情報を更新し、
前記判定部は、直前に良判定された判定対象の前記特徴量によって前記平均値がシフトした更新後の前記良判定特徴情報に基づいて、次の判定対象に対する良否判定を行うことを特徴とする請求項1に記載の判定装置。 - 前記良判定特徴情報は、良判定された判定対象の監視データそれぞれから抽出される特徴量の平均値を中心とした情報であり、
前記基準制御部は、連続して行われる前記良否判定において判定対象が良判定された場合に、当該良判定に用いられた前記良判定特徴情報に、当該良判定された判定対象以前の複数の良判定された判定対象の前記各特徴量を加算して平均し、良判定特徴情報を更新し、
前記判定部は、過去に良判定された複数の判定対象の前記各特徴量によって前記平均値がシフトした更新後の前記良判定特徴情報に基づいて、後続の判定対象に対する良否判定を行うことを特徴とする請求項1に記載の判定装置。 - 前記基準制御部は、当該良判定された判定対象以前の複数の良判定された判定対象の前記各特徴量に重み値を適用し、前記重み値が適用された前記各特徴量を加算して平均し、前記良判定特徴情報を更新することを特徴とする請求項3に記載の判定装置。
- 前記監視データは、撮影装置で判定対象を撮影した画像データであり、
前記特徴量は、前記画像データの各画素位置での輝度値であり、
前記良判定特徴情報は、複数の良判定に属する判定対象の各画像データの前記輝度値を各画素位置において平均した平均輝度値を中心とした情報であり、
前記基準制御部は、連続して行われる良否判定において良否判定結果が良判定である判定対象から抽出された各画素位置での輝度値と、良否判定に適用された前記良判定特徴情報の各画素位置の平均輝度値と、を加算平均して前記良判定特徴情報を更新することを特徴とする請求項1から4のいずれか1つに記載の判定装置。 - 前記良判定特徴情報は、判定対象に対する異常の有無を判定するための第1良判定特徴量情報と、当該異常が所定の許容範囲であるか否かを判定するための第2良判定特徴情報と、を含み、
前記判定部は、前記第1良判定特徴情報に基づいて判定対象に異常の有無を判定する第1判定処理と、異常ありと判定された場合に前記第2良判定特徴情報に基づいて前記異常が所定の許容範囲内であるか否かを判定する第2判定処理と、を含む良否判定処理を行い、
前記基準制御部は、連続して行われる前記良否判定において判定対象が良判定された場合に、良判定された判定対象の前記特徴量を用いて前記第1良判定特徴情報を更新することを特徴とする請求項1に記載の判定装置。 - 判定対象の監視データを用いて良否を判定するコンピュータ装置によって実行されるプログラムであって、
予め収集された良判定に属する判定対象の監視データそれぞれから抽出される特徴量を用いて生成される、良判定特徴情報を記憶する第1機能と、
順次入力される各判定対象の監視データから前記特徴量を抽出し、前記良判定特徴情報に基づいて複数の判定対象に対する良否判定を連続して行う第2機能と、
連続して行われる前記良否判定において判定対象が良判定された場合に、良判定された判定対象の前記特徴量を用いて前記良判定特徴情報を更新する第3機能と、を有し、
前記第2機能は、更新された前記良判定特徴情報に基づいて、当該良判定された後に続く判定対象に対する良否判定を行うことを特徴とするプログラム。 - 判定対象の監視データを用いて良否を判定する判定システムであって、
予め収集された良判定に属する判定対象の監視データそれぞれから特徴量を抽出し、抽出された前記特徴量を用いて良判定特徴情報を生成する良否判定特徴情報生成部と、
前記良判定特徴情報を記憶する記憶部と、
順次入力される各監視データから前記特徴量を抽出し、前記良判定特徴情報に基づいて複数の各判定対象の良否判定を連続して行う判定部と、
連続して行われる前記良否判定において判定対象が良判定された場合に、良判定された判定対象の前記特徴量を用いて前記良判定特徴情報を更新する基準制御部と、を有し、
前記判定部は、更新された前記良判定特徴情報に基づいて、当該良判定された後に続く判定対象に対する良否判定を行うことを特徴とする判定システム。
Priority Applications (6)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR1020237028262A KR20240056679A (ko) | 2022-10-21 | 2022-10-21 | 판정 장치 |
| US18/551,284 US20250076211A1 (en) | 2022-10-21 | 2022-10-21 | Judgment apparatus |
| CN202280020850.0A CN118318157A (zh) | 2022-10-21 | 2022-10-21 | 判断装置 |
| PCT/JP2022/039227 WO2024084675A1 (ja) | 2022-10-21 | 2022-10-21 | 判定装置 |
| EP22927621.7A EP4390380A4 (en) | 2022-10-21 | 2022-10-21 | DETERMINATION DEVICE |
| JP2023516747A JP7330417B1 (ja) | 2022-10-21 | 2022-10-21 | 判定装置 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/JP2022/039227 WO2024084675A1 (ja) | 2022-10-21 | 2022-10-21 | 判定装置 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2024084675A1 true WO2024084675A1 (ja) | 2024-04-25 |
Family
ID=87577122
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/JP2022/039227 Ceased WO2024084675A1 (ja) | 2022-10-21 | 2022-10-21 | 判定装置 |
Country Status (6)
| Country | Link |
|---|---|
| US (1) | US20250076211A1 (ja) |
| EP (1) | EP4390380A4 (ja) |
| JP (1) | JP7330417B1 (ja) |
| KR (1) | KR20240056679A (ja) |
| CN (1) | CN118318157A (ja) |
| WO (1) | WO2024084675A1 (ja) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS63225111A (ja) * | 1987-03-13 | 1988-09-20 | Ckd Corp | 視覚検査方法 |
| JPH05332740A (ja) * | 1992-05-28 | 1993-12-14 | Yamatake Honeywell Co Ltd | 回路部品の接続状態検出方法 |
| JPH06143287A (ja) * | 1992-10-30 | 1994-05-24 | Rhythm Watch Co Ltd | 金型監視方法 |
| JPH10308939A (ja) * | 1997-05-08 | 1998-11-17 | Nec Corp | 赤外線監視システム |
| JP2007309729A (ja) * | 2006-05-17 | 2007-11-29 | Matsushita Electric Ind Co Ltd | プレス加工品の外観検査方法とそれに用いる外観検査装置 |
| JP2013224833A (ja) | 2012-04-20 | 2013-10-31 | Keyence Corp | 外観検査装置、外観検査方法及びコンピュータプログラム |
| JP2021014100A (ja) * | 2019-07-16 | 2021-02-12 | Nok株式会社 | 型監視システム |
| JP2022065961A (ja) | 2020-10-16 | 2022-04-28 | 株式会社東芝 | 判定装置、判定システム、判定方法、プログラム、及び記憶媒体 |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2951037B2 (ja) * | 1991-05-31 | 1999-09-20 | 大日本印刷株式会社 | 印刷物の検査方法および装置 |
| JP2005265661A (ja) * | 2004-03-19 | 2005-09-29 | Ovit:Kk | 画像処理方法およびその装置 |
| JP5085187B2 (ja) * | 2007-05-15 | 2012-11-28 | 日本オートマチックマシン株式会社 | 電線端末部の芯線検査装置および検査方法 |
| WO2012172695A1 (ja) * | 2011-06-15 | 2012-12-20 | キリンテクノシステム株式会社 | ガラス壜の検査方法および装置 |
| CN208678394U (zh) * | 2015-06-30 | 2019-04-02 | 株式会社高永科技 | 物品检查装置 |
| KR20180054063A (ko) * | 2016-11-14 | 2018-05-24 | 주식회사 고영테크놀러지 | 검사체에 대한 양부 판정 조건을 조정하는 방법 및 장치 |
| JP7271305B2 (ja) * | 2019-05-16 | 2023-05-11 | 株式会社キーエンス | 画像検査装置及び画像検査装置の設定方法 |
| JP7462429B2 (ja) * | 2020-02-26 | 2024-04-05 | キヤノン株式会社 | 情報処理装置、情報処理方法、及びプログラム |
| CN113758926A (zh) * | 2020-06-03 | 2021-12-07 | 泰连服务有限公司 | 具有视觉检查站的产品组装机 |
| US20230351582A1 (en) * | 2020-10-07 | 2023-11-02 | Crest Solutions Limited | A line clearance system |
| JP7509027B2 (ja) * | 2020-12-22 | 2024-07-02 | トヨタ自動車株式会社 | 車両検査方法および車両検査システム |
| JP7699978B2 (ja) * | 2021-07-01 | 2025-06-30 | 株式会社日立製作所 | 計算機及び外観検査方法 |
| KR20230028050A (ko) * | 2021-08-20 | 2023-02-28 | 현대모비스 주식회사 | 단계적 합부 판정을 기반으로 한 불량 검출방법 및 장치 |
| WO2023053728A1 (ja) * | 2021-09-29 | 2023-04-06 | 富士フイルム株式会社 | 表示処理装置、表示処理方法、及び、表示処理プログラム |
-
2022
- 2022-10-21 CN CN202280020850.0A patent/CN118318157A/zh active Pending
- 2022-10-21 WO PCT/JP2022/039227 patent/WO2024084675A1/ja not_active Ceased
- 2022-10-21 US US18/551,284 patent/US20250076211A1/en active Pending
- 2022-10-21 JP JP2023516747A patent/JP7330417B1/ja active Active
- 2022-10-21 EP EP22927621.7A patent/EP4390380A4/en active Pending
- 2022-10-21 KR KR1020237028262A patent/KR20240056679A/ko active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS63225111A (ja) * | 1987-03-13 | 1988-09-20 | Ckd Corp | 視覚検査方法 |
| JPH05332740A (ja) * | 1992-05-28 | 1993-12-14 | Yamatake Honeywell Co Ltd | 回路部品の接続状態検出方法 |
| JPH06143287A (ja) * | 1992-10-30 | 1994-05-24 | Rhythm Watch Co Ltd | 金型監視方法 |
| JPH10308939A (ja) * | 1997-05-08 | 1998-11-17 | Nec Corp | 赤外線監視システム |
| JP2007309729A (ja) * | 2006-05-17 | 2007-11-29 | Matsushita Electric Ind Co Ltd | プレス加工品の外観検査方法とそれに用いる外観検査装置 |
| JP2013224833A (ja) | 2012-04-20 | 2013-10-31 | Keyence Corp | 外観検査装置、外観検査方法及びコンピュータプログラム |
| JP2021014100A (ja) * | 2019-07-16 | 2021-02-12 | Nok株式会社 | 型監視システム |
| JP2022065961A (ja) | 2020-10-16 | 2022-04-28 | 株式会社東芝 | 判定装置、判定システム、判定方法、プログラム、及び記憶媒体 |
Non-Patent Citations (1)
| Title |
|---|
| See also references of EP4390380A4 |
Also Published As
| Publication number | Publication date |
|---|---|
| US20250076211A1 (en) | 2025-03-06 |
| JPWO2024084675A1 (ja) | 2024-04-25 |
| KR20240056679A (ko) | 2024-04-30 |
| EP4390380A1 (en) | 2024-06-26 |
| EP4390380A4 (en) | 2024-11-20 |
| JP7330417B1 (ja) | 2023-08-21 |
| CN118318157A (zh) | 2024-07-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3904868A1 (en) | Learned model generation method, learned model, surface defect inspection method, steel manufacturing method, pass/fail determination method, grade determination method, surface defect determination program, pass/fail determination program, determination system, and steel manufacturing equipment | |
| CN110047095B (zh) | 基于目标检测的跟踪方法、装置及终端设备 | |
| CN111563893B (zh) | 基于航拍图像的均压环缺陷检测方法、装置、介质和设备 | |
| Gupta et al. | Deep learning model for defect analysis in industry using casting images | |
| CN109598698A (zh) | 用于对多个项进行分类的系统、方法和计算机程序产品 | |
| CN115908344B (zh) | 一种基于MLCT-YOLO的Micro LED芯片缺陷检测方法 | |
| CN111178392B (zh) | 基于深度神经网络的航空发动机孔探图像损伤分割方法 | |
| JP7059889B2 (ja) | 学習装置、画像生成装置、学習方法、及び学習プログラム | |
| JP2001517308A (ja) | 移動するストリップの表面を、その画像を懐疑的ゾーンにセグメント化することにより検査する方法 | |
| Jian et al. | A multiple species railway defects detection method based on sample generation | |
| CN114203571B (zh) | 用于调谐经调制晶片的敏感度及确定用于经调制晶片的工艺窗的系统、方法及非暂时性计算机可读媒体 | |
| WO2024084675A1 (ja) | 判定装置 | |
| CN117669808A (zh) | 一种基于贝叶斯优化的滑坡易发性预测方法 | |
| Xue et al. | Detection of Various Types of Metal Surface Defects Based on Image Processing. | |
| JP3587029B2 (ja) | 画像検査方法および画像検査装置 | |
| CN120707512A (zh) | 一种基于机器视觉的混凝土桥梁损伤检测和剩余寿命预测方法 | |
| CN119785137A (zh) | 用于评估用于机器学习模型的训练数据集的方法 | |
| CN103164853B (zh) | 成像匹配辅助导航的可匹配性评估方法 | |
| US12530761B2 (en) | Feature quantity selection method and device | |
| TW202534621A (zh) | 使用生成對抗網路模型偵測影像邊界的方法及電子裝置 | |
| CN110766036A (zh) | 一种钻石净度定级的智能方法 | |
| CN106796432A (zh) | 确定用于检查复合部件的采样方案的方法和系统 | |
| JP4166247B2 (ja) | 半導体設計データ修正方法 | |
| Ham et al. | Enhancing structural crack detection through a multiscale multilevel mask deep convolutional neural network and line similarity index | |
| RU2796806C2 (ru) | Способ формирования обучаемой модели, обучаемая модель, способ контроля поверхностных дефектов, способ производства стали, способ определения соответствия/несоответствия, способ определения сорта, программа определения поверхностных дефектов, программа определения соответствия/несоответствия, система определения и оборудование для производства стали |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| ENP | Entry into the national phase |
Ref document number: 2023516747 Country of ref document: JP Kind code of ref document: A |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 202280020850.0 Country of ref document: CN |
|
| ENP | Entry into the national phase |
Ref document number: 2022927621 Country of ref document: EP Effective date: 20230901 |
|
| WWE | Wipo information: entry into national phase |
Ref document number: 18551284 Country of ref document: US |
|
| WWP | Wipo information: published in national office |
Ref document number: 18551284 Country of ref document: US |
|
| NENP | Non-entry into the national phase |
Ref country code: DE |