Determining device, determining system, determining method, and storage medium

Abstract

Provided are a determination device, a determination system, a determination method, and a storage medium capable of automatically determining conditions for processing of an image. The determination device determines whether or not an article is acceptable using an image of the article, repeatedly performs a processing group including first to fourth processes, and determines one of a plurality of conditions based on evaluations of the respective conditions. In the first process, any condition is selected from condition data including a plurality of conditions in which a channel to be used and a pre-process are specified. In the second process, the selected condition is applied to a plurality of images, respectively, and a plurality of processed images are obtained. In the third process, based on results of comparing a plurality of pixel units included in the respective processed images with a plurality of threshold values, respectively, it is determined whether or not the article photographed in each processed image is acceptable. In the fourth process, using one or more selected from determination results of whether or not the article is acceptable for each acceptable product image and each unaccepted product image, an evaluation for the selected condition is determined.

Description

Technical Field

[0001] The embodiments of the present invention relate to a determination device, a determination system, a determination method, and a storage medium. Background Technology

[0002] There are techniques for determining whether an item captured in an image is qualified or not. To improve the accuracy of the determination, it is preferable to apply processing suitable for the determination to the image.

[0003] [Existing technical documents]

[0004] [Patent Literature]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2018-189559 Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a determination device, determination system, determination method and storage medium capable of automatically determining the conditions for image processing.

[0007] The determination device in this embodiment uses images of captured items to determine whether the items are qualified or not. The determination device repeatedly performs a processing group including a first process, a second process, a third process, and a fourth process. In the first process, any one of the conditions is selected from condition data containing multiple conditions, wherein the conditions specify the channel used and the preprocessing. In the second process, multiple processed images are obtained by applying the selected condition to multiple qualified product images and multiple unqualified product images, respectively. In the third process, the qualification or unqualification of the items captured in each of the multiple processed images is determined based on the results of comparing multiple pixel units and multiple thresholds contained in each of the multiple processed images. In the fourth process, an evaluation of the selected condition is determined using one or more evaluation results selected from the qualification / unqualification results for the multiple qualified product images and the qualification / unqualification results for the multiple unqualified product images. The determination device determines one of the multiple conditions based on the evaluations of each of the multiple conditions. Attached Figure Description

[0008] Figure 1 This is a schematic diagram showing the structure of the determination system for the implementation method.

[0009] Figure 2 This is a flowchart illustrating the inspection process of the determination device in the implementation method.

[0010] Figure 3 It is a table that displays conditional data.

[0011] Figure 4It is an image used to illustrate the inspection process.

[0012] Figure 5 It is an image used to illustrate the inspection process.

[0013] Figure 6 This is a flowchart illustrating the condition setting process of the determination device in the implementation method.

[0014] Figure 7 This is a schematic diagram illustrating a user interface.

[0015] Figure 8 This is a schematic diagram illustrating a user interface.

[0016] Figure 9 This is a schematic diagram used to illustrate image processing.

[0017] Figure 10 This is a schematic diagram used to illustrate image processing.

[0018] Figure 11 This is a schematic diagram used to illustrate image processing.

[0019] Figure 12 This is a flowchart illustrating the method for adjusting the threshold of the determination device in the implementation method.

[0020] Figure 13 This is a flowchart illustrating another adjustment method of the determination device for the implementation method.

[0021] Figure 14 This is a flowchart illustrating another adjustment method of the determination device for the implementation method.

[0022] Figure 15 It is a schematic diagram representing the hardware structure. Detailed Implementation

[0023] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In this specification and the drawings, elements that are the same as those already described are labeled with the same reference numerals, and detailed descriptions are omitted where appropriate.

[0024] Figure 1 This is a schematic diagram showing the structure of the determination system for the implementation method.

[0025] like Figure 1 As shown, the determination system 1 of the embodiment includes a determination device 10, an input device 11, a display device 12, a storage device 20, and a camera device 30. The determination system 1 of the embodiment can be used for the visual inspection of articles.

[0026] The camera device 30 captures images of the item A to be inspected. The camera device 30 stores the captured images in the storage device 20. The camera device 30 acquires still images of item A. The camera device 30 can also acquire moving images and extract still images from the moving images. The orientation of the camera on item A is appropriately set to obtain images suitable for inspection. For example, multiple items A are sequentially conveyed by the conveyor 40 in a horizontal direction. The camera device 30 is fixed and captures images of each conveyed item A sequentially from above.

[0027] For example, the imaging device 30 acquires an RGB image. The imaging device 30 may also acquire an HSV image or an HSL image. The image contains pixel values ​​of multiple pixels and data for each pixel's channels. In the case of an RGB image, each channel represents color information in terms of R (red), G (green), or B (blue). In the case of an HSV image, each channel represents color information in terms of H (hue), S (saturation), or V (lightness). In the case of an HSL image, each channel represents color information in terms of H (hue), S (saturation), or L (lightness). In addition to color, the image may also include channels for infrared light, ultraviolet light, or distance. Alternatively, the imaging device 30 may include a polarization image sensor to acquire a polarization image. In this case, each channel represents information about the polarization angle (0 degrees, 45 degrees, 90 degrees, 135 degrees, etc.).

[0028] The judging device 10 acquires the image stored in the storage device 20. Alternatively, the image can be sent directly from the camera device 30 to the judging device 10. The judging device 10 determines whether item A in the captured image is qualified or not. The input device 11 is used for the user to input data to the judging device 10. The display device 12 displays the data output from the judging device 10 in a manner that allows the user to visually confirm it.

[0029] Figure 2 This is a flowchart illustrating the inspection process of the determination device in the implementation method.

[0030] The determining device 10 accesses the storage device 20 and obtains an image (step S1). An image contains multiple pixel units arranged in the X and Y directions. For example, a pixel unit consists of one pixel. Alternatively, a pixel unit may consist of two or more adjacent pixels.

[0031] The determination device 10 accesses the storage device 20 to obtain the processing conditions for the image of item A (step S2). The conditions specify the channels in the image used and the preprocessing steps for the image. The preprocessing steps include, for example, at least one selected from edge enhancement, noise removal, position alignment, and grayscale correction. For edge enhancement, for example, one or more selected from Sobel filters, Laplacian filters, dispersion filters, sharpening filters, maximum filters, minimum filters, Laplacian of Gaussian (LOG) filters, and Difference of Gaussian (DoG) filters are used. For noise removal, for example, one or more selected from smoothing filters, Gaussian filters, median filters, bilateral filters, dilation processing, contraction processing, and gamma correction are used. For position alignment, for example, one or more selected from template matching, feature matching, and position-limited correlation methods are used. For grayscale correction, coefficients for the pixel values ​​of each channel are set. Preprocessing may also include inter-image operations. The conditions are set according to each type of item. The determination device 10 determines the image application conditions and obtains the processed image (step S3).

[0032] The determination device 10 accesses the storage device 20 to obtain multiple thresholds (step S4). These multiple thresholds are used to determine whether multiple pixel units contained in the processed image are normal. A threshold is set for each pixel unit. Specifically, an upper limit and a lower limit are set as thresholds. Typically, the pixel value is any integer from 0 to 255. The upper and lower limits are each set as any integer from 0 to 255. If the pixel value is between the lower and upper limits, the pixel unit with that pixel value is determined to be normal. If the pixel value deviates from the lower and upper limits, the pixel unit with that pixel value is determined to be abnormal.

[0033] When a pixel unit consists of a single pixel, the pixel value of that single pixel can be used as the pixel value of that pixel unit. When a pixel unit consists of two or more pixels, the average, weighted average, or median value of the pixel values ​​of the two or more pixels can be used as the pixel value of that pixel unit.

[0034] The determination device 10 compares multiple pixel values ​​in the processed image with multiple thresholds respectively, and extracts abnormal pixel units that deviate from the thresholds (step S5). The determination device 10 infers abnormal regions based on the extracted abnormal pixel units (step S6). For example, in inferring abnormal regions, noise removal, dilation, contraction, and labeling are performed. Through these processes, blocks (particles) of abnormal pixel units are generated and identified. The determination device 10 infers the particles of abnormal pixel units as abnormal regions.

[0035] The determination device 10 determines whether an item captured in the image is qualified or not based on the characteristics of the abnormal region (step S7). As a characteristic, at least one selected from the area of ​​the abnormal region, the shape of the abnormal region, and the distribution of pixel values ​​in the abnormal region is used.

[0036] When using area as a feature, a threshold is preset for the area. The area is represented by the number of pixels contained in the abnormal region. If the area of ​​the abnormal region exceeds the threshold, the determination device 10 determines the item as unqualified.

[0037] As a shape feature, aspect ratio can be used, for example. Aspect ratio is the ratio of the length along the major axis to the length along the minor axis. The major axis is parallel to the longest line segment connecting any two points on the outer edge of the particle. The minor axis is perpendicular to the major axis. When using shape as a feature, a threshold for aspect ratio is preset. The relationship between aspect ratio and the threshold is set according to the item. As an example, when inspecting items with poor quality, there is a tendency for the aspect ratio to be larger in abnormal areas. The judgment device 10 determines the item as unqualified if the aspect ratio exceeds the threshold.

[0038] As a distribution, for example, the ratio of the number of pixel units below the lower threshold to the number of pixel units in the abnormal area (first ratio), and the ratio of the number of pixel units exceeding the upper threshold to the number of pixel units in the abnormal area (second ratio), etc. When using the distribution as a feature, a first threshold for the first ratio and a second threshold for the second ratio are preset respectively. If the first ratio exceeds the first threshold, or if the second ratio exceeds the second threshold, the determination device 10 determines that the item is unqualified.

[0039] The determination device 10 outputs a determination result (step S8). For example, the determination device 10 saves the determination result in the storage device 20. The determination device 10 may also save the image used in the inspection in association with the determination result in the storage device 20. The saved image may be an image acquired by the imaging device 30, or an image processed according to conditions. The extracted abnormal pixel units or the inferred abnormal areas may also be shown in the image. The determination device 10 may also display the determination result and the image on the display device 12.

[0040] Figure 3 It is a table that displays conditional data.

[0041] like Figure 3As shown, the condition data 100 includes multiple conditions 110. Each condition 110 includes a combination of the channel 120 used and preprocessing 130 applied to the image. The conditions used are pre-set for each type of item. The determination device 10 selects one of the multiple conditions based on the item to be inspected and applies it to the image of the item.

[0042] Here, an example is illustrated where channels and preprocessing are stored in a single table. Channels and preprocessing can be stored in separate tables. In this case, the determining device 10 selects conditions based on both channels and preprocessing, and applies these conditions to the image.

[0043] Figure 4 (a)~ Figure 4 (c) and Figure 5 (a)~ Figure 5 (c) is an image used to illustrate the inspection process.

[0044] The following explains the case where a pixel unit consists of a single pixel. Figure 4 (a) represents the processed image IM1 obtained by applying conditions selected from the conditional data to the image of item A1. Figure 4 (b) and Figure 4 (c) represents an image where the threshold values ​​for each pixel are set for the processed image. Specifically, Figure 4 In the image shown in (b), IM2 represents the upper limit of the threshold for each pixel. Figure 4 Image IM3, shown in (c), represents the lower limit of the threshold for each pixel. If any pixel in processed image IM1 is brighter than the pixel at the same coordinate in image IM2, or darker than the pixel at the same coordinate in image IM3, that pixel in processed image IM1 is determined to be abnormal. In this example, item A1 captured in processed image IM1 is determined to be acceptable.

[0045] Figure 5 Image (a) represents the processed image IM4 obtained by applying conditions selected from conditional data to the image of item A2. Item A2 has a defect. Specifically, a portion of item A2 is darker than item A1. The determination device 10 uses multiple thresholds to extract abnormal pixels from multiple pixels. Figure 5 Image (b) shows IM5, which represents the extracted anomalous pixels. Figure 5 In example (b), the particles P1 to P6 of the abnormal pixels are extracted.

[0046] The determination device 10 estimates the abnormal region based on particles P1 to P6. For example, due to noise removal and shrinkage, particles P5 and P6 disappear. Due to expansion, particles P1 to P4 combine with each other. Thus, [the following is obtained] Figure 5Image IM6 is shown in (c). In image IM6, particles P7 containing abnormal pixels are shown. The determination device 10 presumes particles P7 as an abnormal region. Based on the characteristics of the presumed abnormal region, the determination device 10 determines whether item A2 is qualified or not.

[0047] The appearance of items and defects in an image varies considerably depending on the conditions applied. To accurately determine whether an item captured in an image is acceptable, it is preferable to set the applied conditions appropriately.

[0048] Figure 6 This is a flowchart illustrating the condition setting process of the determination device in the implementation method.

[0049] First, the determining device 10 accesses the storage device 20 to obtain multiple images of qualified products and multiple images of unqualified products (step S10). The multiple images of qualified products and multiple images of unqualified products are prepared by the user. Qualified product images are those deemed by the user to represent items of good quality. Unqualified product images are those deemed by the user to represent items of poor quality.

[0050] The determination device 10 selects a condition from the condition data (step S11, first processing). The determination device 10 applies the selected condition to multiple qualified product images and multiple unqualified product images to obtain multiple processed images (step S12, second processing). The multiple processed images include multiple qualified product processed images generated by processing multiple qualified product images, and multiple unqualified product processed images generated by processing multiple unqualified product images.

[0051] The determination device 10 sets multiple thresholds (step S13). For example, the thresholds are set using the average and dispersion of pixel values ​​located at the same coordinate in multiple qualified product processing images. As an example, for each pixel, the lower threshold is set to (μ-3σ), and the upper threshold is set to (μ+3σ). μ is the average value. σ is the standard deviation. For each pixel, the average value and dispersion can also be calculated by including surrounding pixels. Thus, for contour areas where pixel values ​​vary significantly, over-detection of defects can be suppressed, and the accuracy of the qualification / disqualification determination can be improved.

[0052] The determination device 10 determines whether an item is qualified or not based on comparisons of pixel values ​​of multiple pixels and multiple thresholds for each processed image (step S14, third processing). In the determination of qualification or non-qualification, the following steps are performed: Figure 2 Steps S5 to S7 of the flowchart shown. The determination device 10 evaluates the selected conditions using the determination results of whether multiple qualified product images are qualified or not, or the determination results of whether multiple unqualified product images are qualified or not (step S15, fourth process).

[0053] In the evaluation, one or more items selected from the following is / are used: the number V1 of items determined as acceptable for a plurality of acceptable product images, the number V2 of items determined as unacceptable for a plurality of acceptable product images, the number V3 of items determined as acceptable for a plurality of unacceptable product images, and the number V4 of items determined as unacceptable for a plurality of unacceptable product images.

[0054] For example, accuracy rate is used in the evaluation. The accuracy rate is represented by the ratio of the sum of the number V1 and the number V4 to the total of the numbers V1 to V4. In the evaluation, fit rate, recall rate, specificity or F-value may also be used. The fit rate is represented by the ratio of the number V4 to the sum of the number V2 and V4. The recall rate is represented by the ratio of the number V4 to the sum of the number V3 and V4. The specificity is represented by the ratio of the number V1 to the sum of the number V1 and V2. The F-value is represented by (2×V4) / (V2+V3+2×V4). Higher accuracy rate, fit rate, recall rate, specificity and F-value are more preferable. The determination apparatus 10 takes the accuracy rate, fit rate, recall rate, specificity or F-value as an evaluation score representing the condition. A higher score indicates a higher evaluation of the condition.

[0055] Incorrect rate may also be used in the evaluation. The incorrect rate is represented by the ratio of the number V2 to the sum of the number V1 and V2. The incorrect rate may also be represented by the ratio of the number V3 to the sum of the number V3 and V4. A lower incorrect rate is more preferable. The determination apparatus 10 takes the incorrect rate as an evaluation score representing the condition. A lower score indicates a higher evaluation of the condition.

[0056] The determination apparatus 10 determines whether an end condition is satisfied (step S16). For example, setting that all preset conditions are selected by repeating steps S12 to S15 is taken as the end condition. Setting that steps S12 to S15 are repeated for a preset number of times may also be taken as the end condition. A threshold for the score may also be set. In this case, setting that the score reaches the threshold is taken as the end condition.

[0057] If the end condition is not satisfied, the determination apparatus 10 executes steps S12 to S15 again. In step S12, a condition different from the previously selected condition is selected. If the end condition is satisfied, the determination apparatus 10 determines one condition to be used in inspection based on the evaluation of each condition (step S17).

[0058] Thereafter, the inspection process is executed using the determined condition. In the inspection, the image capturing the object to be inspected is subjected to Figure 2 processing of the flowchart shown. Thereby, the acceptability of the item captured in the image is determined. For example, an item determined as acceptable is passed the inspection, and an item determined as unacceptable is failed the inspection.

[0059] The conditions included in the condition data can be preset by the user. The determination device 10 can also automatically select from a candidate set of conditions. For example, the user may prepare a list of selectable channels and a list of selectable preprocessing steps in advance. The determination device 10 selects one or more channels from the list of channels and one or more preprocessing steps from the list of preprocessing steps. The determination device 10 combines the selected channels and preprocessing steps to create conditions. The determination device 10 can select channels and preprocessing steps randomly or according to rules. The determination device 10 repeats the creation of conditions, generating multiple conditions.

[0060] Figure 7 and Figure 8 This is a schematic diagram illustrating the user interface.

[0061] The candidate conditions can also be specified by the user. For example, the decision device 10 will... Figure 7 The window 150 shown is displayed on the display device 12 as a graphical user interface (GUI). Channel candidate 160 and preprocessing candidate 170 are displayed in the window 150. Checkboxes 161 and 171 are displayed adjacent to the channel candidate 160 and preprocessing candidate 170, respectively.

[0062] The user can select candidates for a condition by checking checkboxes 161 or 171 using input device 11. Channel candidate 162 is a grayscale transformation. Regarding channel candidate 162, the coefficients of the pixel values ​​of each channel are automatically searched for when a grayscale transformation occurs. The user can specify the step size of each coefficient in the search by entering a value in input field 163. Additionally, the user can specify the range of filter size by entering a value for filter preprocessing candidate 170 in input field 172. The user can specify the weights when applying the filter by entering a value in input field 173. The user can specify the step size of the weights in the search by entering a value in input field 174.

[0063] like Figure 8 As shown, as a preprocessing candidate 170, the user can also set arbitrary conditions. For example, the user can check a checkbox 175 and input a filter size in the input field 176 of the checked item. For example, if the user checks a checkbox and inputs a filter size, window 180 is displayed. In window 180, a matrix corresponding to the input filter size is displayed. The user can freely set the preprocessing content by inputting weights 181 into each cell of the matrix.

[0064] The determination device 10 saves the candidates of the selected conditions in the storage device 20. The determination device 10 executes... Figure 6 When the flowchart shown is executed, select criteria from the candidates specified by the user.

[0065] Figure 9 (a)~ Figure 9 of (h) Figure 10 (a)~ Figure 10 (e) and Figure 11 (a)~ Figure 11 (f) is a schematic diagram used to illustrate the processing of an image.

[0066] Figure 9 (a) and Figure 9 (e) represents images obtained by photographing a portion of article 200a and a portion of article 200b, respectively. Furthermore, the images obtained by photography are actually RGB images, but... Figure 9 (a) and Figure 9 In (e), the image after grayscale transformation is shown. Items 200a and 200b belong to the same category. Item 200a is a qualified product. Item 200b is a defective product. Figure 9 (a) and Figure 9 As shown in (e), items 200a and 200b include regions 201 to 203. If the items captured in the image are acceptable, regions 201 and 203 are ice green, and region 202 is rose pink. During the inspection process, region 202 is checked for any red tinge.

[0067] Figure 9 (b) Figure 9 (d) and Figure 9 (f) Figure 9 (h) represents the processed image after processing according to the conditions. Specifically, Figure 9 (b) and Figure 9 (f) are respectively from Figure 9 (a) and Figure 9 It was obtained by extracting only the "G" channel from the image (e). Figure 9 (c) and Figure 9 (g) are respectively from Figure 9 (a) and Figure 9 It was obtained by extracting only the "B" channel from the image (e). Figure 9 (d) and Figure 9 (h) are respectively from Figure 9 (a) and Figure 9 It was obtained by extracting only the "R" channel from the image (e).

[0068] from Figure 9 (b) Figure 9 (d) and Figure 9 (f) Figure 9 From (h), we can know that Figure 9The pixel value of region 202 in image (d) is... Figure 9 The difference between pixel values ​​in region 202 of the image (h) is greater than Figure 9 The pixel value of region 202 in image (b) is... Figure 9 The difference between pixel values ​​in region 202 of image (f) is greater than Figure 9 The pixel value of region 202 in image (c) is... Figure 9 The difference between pixel values ​​in region 202 of the image (g). By using an image with only the "R" channel extracted for inspection, the detection of defective products becomes easier.

[0069] Figure 10 (a) represents an image obtained by photographing a portion of article 210. Figure 10 In (a), the RGB image is transformed to grayscale. For example... Figure 10 As shown in (a), in article 210, area 211 may have dirt 212 and rust 213 attached. For example, area 211 is gray in color. Dirt 212 is black in color. Rust 213 is red in color. Regarding article 210, regardless of the attachment of dirt 212, articles without rust 213 should be judged as acceptable. Articles with rust 213 should be judged as unacceptable.

[0070] Figure 10 (b) Figure 10 (e) represents the processed image after processing according to the conditions. Figure 10 (b) is from Figure 10 It was obtained by extracting only the "R" channel from the image (a). Figure 10 (c) is from Figure 10 It was obtained by extracting only the "G" channel from the image (a). Figure 10 (d) is from Figure 10 It was obtained by extracting only the "B" channel from the image (a). Figure 10 (e) is from Figure 10 The image of (b) and Figure 10 The difference is obtained from the image of (c).

[0071] exist Figure 10 In image (b), rust 213 is not visible, making it difficult to distinguish between region 211 and rust 213. Figure 10 (c) and Figure 10 In image (d), the difference between the pixel values ​​of dirt 212 and rust 213 is small, making it difficult to distinguish between dirt 212 and rust 213. On the other hand, in Figure 10In image (e), dirt 212 has been removed. The difference between the pixel values ​​of region 211 and the pixel values ​​of rust 213 is also large enough to be distinguishable. Therefore, compared with... Figure 10 (a)~ Figure 10 Compared to image (d), rust 213 can be detected more easily regardless of the adhesion of dirt 212. That is, by using the image obtained from the difference between the image of the extracted "R" channel and the image of the extracted "G" channel for inspection, the detection of defective products becomes easier.

[0072] Figure 11 (a) and Figure 11 (b) represents images obtained by photographing a portion of article 220a and a portion of article 220b, respectively. Figure 11 (a) and Figure 11 In (b), the RGB image is converted to grayscale. Items 220a and 220b belong to the same category. Item 220a is a qualified product. Item 220b is a defective product. Figure 11 As shown in (b), article 220b includes defect 221.

[0073] Figure 11 (c)~ Figure 11 (f) represents the processed image after conditions have been applied. Specifically, Figure 11 (c) is from Figure 11 The image of (a) and Figure 11 The difference between the images of (b) is obtained. Figure 11 (d) and Figure 11 (e) are respectively for Figure 11 (a) and Figure 11 The image in (b) was obtained by applying a scatter filter (filter size 5×5). Figure 11 (f) is based on Figure 11 The image of (d) and Figure 11 It is obtained by the difference of the image of (e).

[0074] like Figure 11 As shown in (c), the difference in pixel values ​​for defect 221 is small between the acceptable and unacceptable product images. The smaller the difference between the acceptable and unacceptable product images, the more difficult it is to detect defect 221. Figure 11 (c) and Figure 11 As can be seen from the comparison of (f), for the pixel value of defect 221, the difference between the qualified and unqualified product images after applying the dispersion filter is larger than the difference between the original qualified and unqualified product images. Therefore, the detection of defect 221 becomes easier.

[0075] The advantages of the implementation method are explained.

[0076] As mentioned above, processing the images of captured items can improve the accuracy of determining whether they are acceptable or not. However, the appropriate processing method for improving accuracy varies depending on the type of item. If the user has to search for the appropriate processing method for each type of item, it places a heavy burden on the user. There is also a possibility that the user may not find the appropriate processing method. Furthermore, even if the user determines a processing method to be appropriate, there are situations where it may not actually improve the accuracy of the determination.

[0077] To address this technical problem, the determination device 10 of the implementation method is as follows: Figure 6 As shown, the processing group of steps S12 to S15 is repeated. Through the repetition of this processing group, multiple conditions are automatically evaluated. Based on the evaluation of each condition, the judgment device 10 determines the conditions to be used in the inspection. Thus, conditions suitable for improved judgment accuracy can be automatically identified.

[0078] Furthermore, in the evaluation of each condition, one or more judgment results are selected from the judgment results of whether multiple qualified product images are qualified or not, and the judgment results of whether multiple unqualified product images are qualified or not. As a result, the judgment device 10 can select the conditions more suitable for inspection.

[0079] To further improve the accuracy of the inspection, it is preferable to set a suitable threshold. The judgment device 10 can also be configured to... Figure 6 In step S13 of the flowchart shown, multiple threshold adjustment processes are performed.

[0080] Figure 12 This is a flowchart illustrating the method for adjusting the threshold of the determination device in the implementation method.

[0081] When the determination device 10 acquires multiple processed images in step S12, it sets initial values ​​for multiple thresholds (step S13a). The determination device 10 determines whether the items captured in at least a portion of the qualified product processed images are qualified or not (step S13b, first sub-processing).

[0082] For ease of explanation, the images of the captured items that are judged as either qualified or unqualified will be referred to as "qualified product processing images" or "unqualified product processing images," respectively. The images of the captured items that are judged as either qualified or unqualified will also be referred to as "unqualified product processing images." The judging device 10 determines whether there is a qualified product processing image that is judged as unqualified during the judging process (step S13c).

[0083] When there is a qualified product processed image determined to be unqualified, the determination apparatus 10 relaxes at least a part of the threshold values (step S13d, a second sub-process). Through the relaxation of the threshold values, it becomes more difficult to determine a qualified product processed image as unqualified. Specifically, in the relaxation of the threshold values, one or two selected from increasing the upper limit and decreasing the lower limit is performed. After at least a part of the threshold values are relaxed, the determination apparatus 10 executes step S13b again.

[0084] Steps S13b to S13d are repeated until all qualified product processed images are determined to be qualified. If there is no processed image determined to be unqualified in step S13c, the determination apparatus 10 ends the process. The adjustment of a plurality of threshold values is completed. Then, using the adjusted plurality of threshold values, execute Figure 6 step S14 of the flowchart shown in.

[0085] Initial values of the threshold values are set as strict values such that an item captured in at least one qualified product processed image is determined to be unqualified. Therefore, in the first step S13b, at least one qualified product processed image is determined to be unqualified. A plurality of threshold values may also be set such that items captured in all qualified product processed images are determined to be unqualified.

[0086] In image-based inspection, more importance is attached to not misdetermining qualified products as unqualified than to not misdetermining unqualified products as qualified. For example, when an inspection mainly aimed at screening out obvious defects is performed, it is more important to ensure smooth production by not misdetermining qualified products as unqualified than to prevent unqualified products from flowing to subsequent processes. According to the above threshold adjustment method, the threshold values are adjusted until each qualified product processed image is determined to be qualified. According to the embodiment, the plurality of threshold values are set in such a manner that all qualified product processed images are determined to be qualified. Thereby, in actual inspection, the possibility of misdetermining qualified products as unqualified can be reduced.

[0087] Preferably, after the adjustment process, the condition is evaluated by using the qualification determination result for the unqualified product processed image obtained with the adjusted threshold values. If the threshold values are excessively relaxed, an unqualified product processed image may be misdetermined as qualified. By evaluating the condition using the qualification determination result for the unqualified product processed image, it can be confirmed whether the threshold values are excessively relaxed. For example, as a result of excessive relaxation of the threshold values, a condition with decreased evaluation is not adopted.

[0088] According to the above method, more suitable conditions for inspection can be found, including the adjustment of threshold values.

[0089] Hereinafter, the processing of the determination apparatus 10 in the adjustment process will be specifically described.

[0090] The initial value for the threshold can be a pre-set fixed value or set based on the average value and the dispersion. The average value is calculated by averaging the pixel values ​​at the same coordinate in multiple qualified product processing images. As the dispersion, the standard deviation, variance, or mean deviation of the pixel values ​​at the same coordinate in multiple qualified product processing images can be used. To strictly set the initial value, the dispersion is set to be small. For example, using the average value μ and the standard deviation σ, (μ-σ) and (μ+σ) are set as the threshold. By using the average value and the dispersion to set the initial value, the number of repetitions of steps S13b to S13d required for subsequent threshold adjustment can be reduced compared to the case where the initial value is fixed.

[0091] In setting the initial values, multiple defective product processing images can also be used. After setting multiple thresholds as initial values, the determination device 10 determines whether the items captured in the multiple defective product processing images are qualified or not. If all defective product processing images are determined to be unqualified, the determination device 10 allows each of the set initial values. If any defective product processing image is determined to be qualified, the determination device 10 narrows all thresholds by a predetermined amount. That is, the determination device 10 reduces the upper limit of the pixel value by a predetermined amount, or increases the lower limit of the pixel value by a predetermined amount. The determination device 10 may also narrow the threshold only for pixel units that are not determined to be abnormal.

[0092] In step S13b, the determination device 10 can also exclude qualified product processing images that have already been determined as qualified from the objects of determination in previous repeated processing. Even qualified product processing images that have been determined as qualified once will be determined as qualified in subsequent determination processes. By excluding qualified product processing images that have been determined as qualified from the objects, the amount of computation required for repeating steps S13b to S13d can be reduced.

[0093] Figure 13 This is a flowchart illustrating another adjustment method of the determination device for the implementation method.

[0094] In step S13b of the first iteration, the judging device 10 can judge whether the images of all the prepared qualified products are qualified or not, or it can judge whether the images of only a portion of the qualified products are qualified or not.

[0095] For example, such as Figure 13As shown, after step S13a, the judging device 10 selects a predetermined number of qualified product processing images from a plurality of qualified product processing images, either non-randomly or randomly (step S13e). The judging device 10 executes step S13b on the selected qualified product processing images. If there are qualified product processing images that are judged as unqualified in step S13b, the judging device 10 executes step S13d. If all the selected qualified product processing images are judged as qualified in step S13b, the judging device 10 determines whether all the prepared qualified product processing images have been selected (step S13f). If there are qualified product processing images that have not yet been selected, the judging device 10 executes step S13e again. At this time, the judging device 10 excludes qualified product processing images that have been judged as good from the selected objects.

[0096] In step S13b, when only a portion of the qualified product processing images are judged as qualified, the computational load required for threshold adjustment can be reduced. On the other hand, if step S13d is performed based only on the judgment results of a portion of the qualified product processing images, it is possible to relax some thresholds beyond what is necessary.

[0097] As an example, in the initial step S13e, a first qualified product processing image with relatively few abnormal pixel units is selected. In step S13b, the first qualified product processing image is determined to be unqualified. Based on the first qualified product processing image, step S13d is executed, and the minority threshold is relaxed. In the subsequent step S13e, a second qualified product processing image with relatively many abnormal pixel units is selected. In step S13b, the second qualified product processing image is determined to be unqualified. Based on the second qualified product processing image, step S13d is executed, and the majority threshold is relaxed. In this example, after the minority threshold is relaxed, the majority threshold is further relaxed.

[0098] In contrast, if all qualified product processing images are selected at once, and the threshold of the majority is relaxed based on the second qualified product processing image, it is possible that the threshold of the minority may not be relaxed. Therefore, in order to suppress excessive relaxation of the threshold and improve the determination accuracy, it is preferable to determine whether all qualified product processing images are qualified or not in step S13b once, and adjust the threshold based on the determination result.

[0099] For example, when relaxing the threshold, the determination device 10 extracts all abnormal pixel units from the qualified product processing images that were determined to be unqualified in step S13b, and relaxes the threshold for these abnormal pixel units. Specifically, when the number of qualified product processing images determined to be unqualified is two or more, the determination device 10 obtains the sum set of abnormal pixel units in these qualified product processing images. The determination device 10 relaxes the threshold for the abnormal pixel units contained in the sum set. According to this method, the threshold can be relaxed effectively, and the amount of computation required for threshold adjustment can be reduced.

[0100] When relaxing the threshold, the determination device 10 can also select a portion of the qualified product processing images from those determined to be unqualified in step S13b, based on abnormal regions. The determination device 10 relaxes the threshold for abnormal pixel units in the selected qualified product processing images. For example, the determination device 10 selects the qualified product processing image with the largest abnormal region area. The determination device 10 relaxes the threshold for abnormal pixel units in the selected qualified product processing images. According to this method, it is possible to prevent the threshold from being relaxed beyond what is necessary, thereby improving the accuracy of the qualification / disqualification determination.

[0101] The user can also pre-limit the pixel units for which the threshold can be relaxed. In step S13d, the determination device 10 relaxes the threshold for unrestricted pixel units. Relaxation restrictions can also be set for at least a portion of the thresholds. For example, an upper limit or a lower limit can be set for the threshold. In step S13d, no thresholds that have already reached their limits are relaxed. By setting restrictions on pixel units or thresholds, it is possible to prevent the determination of pixel units where strict checks for abnormalities are desired from becoming too slow.

[0102] For example, the size of the relaxed threshold is preset by the user. When relaxing the threshold, the determination device 10 adds the user's preset value to the upper limit of the pixel value and subtracts the preset value from the lower limit of the pixel value. As a result, the range between the lower and upper limits is expanded, and the threshold is relaxed.

[0103] The size of the relaxed threshold can also be determined based on one or two of the pixel values ​​of the corresponding pixel units and the area of ​​the set of pixel units for which the threshold is relaxed. For example, the greater the difference between the pixel value and the threshold, the greater the threshold is relaxed. The larger the area of ​​the set of pixel units for which the threshold is relaxed, the greater the threshold is relaxed. This reduces the time required to adjust multiple thresholds.

[0104] Figure 14 This is a flowchart illustrating another adjustment method of the determination device for the implementation method.

[0105] The determination device 10 can also use the defective product processing images to determine whether multiple thresholds are appropriate. The determination device 10 executes steps S13b to S13d. After step S13d, the determination device 10 uses multiple thresholds to determine whether the items captured in each defective product processing image are qualified or not (step S13g, third sub-processing). The determination device 10 calculates the ratio of the number of defective product processing images determined to be qualified to the total number of determined defective product processing images. The determination device 10 determines whether this ratio is less than a predetermined ratio (step S13h). The predetermined ratio can also be set to zero. That is, even if one defective product processing image is determined to be good, the ratio is greater than or equal to the predetermined ratio.

[0106] The acceptance or rejection of items captured in the defective product processing image is determined in the same way as that of items captured in the acceptable product processing image. Specifically, the determination device 10 compares multiple pixel values ​​in the defective product processing image with multiple thresholds, extracting abnormal pixel units that deviate from the thresholds. The determination device 10 infers abnormal regions based on the extracted abnormal pixel units. Based on the characteristics of the abnormal regions, the determination device 10 determines whether the items captured in the defective product processing image are acceptable or not.

[0107] If an item captured in a specified proportion of multiple defective product processing images is determined to be acceptable, the determination device 10 stops the repetitive processing. The determination device 10 may also notify the user of the suspension of repetitive processing (step S13i). For example, the determination device 10 may notify the user by displaying data to the display device 12 or sending data to a specific terminal device.

[0108] The threshold for classifying non-conforming products as acceptable in processing images is excessively relaxed, indicating that at least a portion of the thresholds have been overly broadened. When these thresholds are applied to actual inspections, it is possible to misclassify non-conforming products as acceptable. By halting repetitive processing, useless calculations can be avoided. For example, when repetitive processing is halted, in... Figure 6 In the flowchart shown, steps S14 and S15 regarding this condition are omitted. This condition is not selected in step S17.

[0109] Furthermore, notifying the user through the judgment device 10 can draw the user's attention. An example of an overly relaxed threshold is the presence of defective product images mixed in with acceptable product images prepared by the user. Notification allows the user to confirm whether the threshold adjustment method is inadequate.

[0110] exist Figure 14In the flowchart shown, step S13g can also be executed before step S13d. For example, step S13g can also be executed simultaneously with step S13b. When steps S13g and S13b are executed simultaneously, step S13h can also be executed after steps S13b and S13g and before step S13c.

[0111] Figure 15 It is a schematic diagram representing the hardware structure.

[0112] The determining device 10 includes, for example, a Figure 15 The hardware configuration is shown. Figure 15 The processing device 90 shown includes a CPU 91, a ROM 92, a RAM 93, a storage device 94, an input interface 95, an output interface 96, and a communication interface 97.

[0113] ROM 92 stores programs used to control the operation of the computer. ROM 92 contains the programs required for the computer to perform the aforementioned processes. RAM 93 functions as a storage area for the expanded programs stored in ROM 92.

[0114] CPU 91 includes processing circuitry. CPU 91 uses RAM 93 as its working memory and executes programs stored in at least one of ROM 92 or storage device 94. During program execution, CPU 91 controls various structures and performs various processes via system bus 98.

[0115] Storage device 94 stores the data required for executing the program and the data obtained by executing the program.

[0116] The input interface (I / F) 95 connects the processing device 90 to the input device 95a. The input I / F 95 is, for example, a serial bus interface such as USB. The CPU 91 can read various data from the input device 95a via the input I / F 95.

[0117] Output interface (I / F) 96 connects the processing device 90 to the display device 96a. Output I / F 96 may be a video output interface such as Digital Visual Interface (DVI) or High-Definition Multimedia Interface (HDMI). The CPU 91 can send data to the display device 96a via output I / F 96, causing the display device 96a to display images.

[0118] The communication interface (I / F) 97 connects the server 97a external to the processing device 90 to the processing device 90. The communication I / F 97 is, for example, a network card such as a LAN card. The CPU 91 can read various data from the server 97a via the communication I / F 97. The camera 99 takes pictures of objects and saves the images to the server 97a.

[0119] Storage device 94 includes one or more selected from Hard Disk Drive (HDD) and Solid State Drive (SSD). Input device 95a includes one or more selected from mouse, keyboard, microphone (voice input), and touchpad. Display device 96a includes one or more selected from monitor and projector. A device that combines the functions of both input device 95a and display device 96a can also be used, similar to a touchpad.

[0120] Storage device 94 can also function as storage device 20. Input device 95a can also function as input device 11. Display device 96a can also function as display device 12. Camera 99 can also function as imaging device 30.

[0121] The processing of the various types of data described above can also be recorded as programs that can be executed by a computer on a disk (floppy disk and hard disk, etc.), optical disk (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, etc.), semiconductor memory, or other non-transitory computer-readable storage medium.

[0122] For example, information recorded on a recording medium can be read by a computer (or embedded system). The recording format (storage format) on the recording medium is arbitrary. For instance, a computer reads a program from the recording medium and, based on that program, causes the CPU to execute the instructions described in the program. In a computer, program retrieval (or reading) can also be performed via a network.

[0123] The implementation method includes the following features.

[0124] (Feature 1)

[0125] A determining device that uses captured images of an item to determine whether the item is qualified or not.

[0126] The determining device repeatedly processes the group of conditions and, based on the evaluation of each of the multiple conditions, determines one of the conditions.

[0127] The processing group includes:

[0128] The first process involves selecting any one of the conditions from condition data that includes multiple conditions, wherein the condition specifies the channel to be used and the preprocessing.

[0129] The second process involves applying the selected conditions to multiple qualified product images and multiple unqualified product images respectively to obtain multiple processed images;

[0130] The third process involves comparing the multiple pixel units and multiple thresholds contained in each of the multiple processed images to determine whether the items captured in each of the multiple processed images are qualified or not; and

[0131] The fourth process involves using one or more judgment results selected from the judgment results of whether the plurality of qualified product images are qualified or not, and the judgment results of whether the plurality of unqualified product images are qualified or not, to determine the evaluation for the selected condition.

[0132] (Feature 2)

[0133] According to the determination device of feature 1, in the fourth process, the evaluation for the selected condition is determined by using two or more of the following: the number of items determined to be qualified for the plurality of qualified product images, the number of items determined to be unqualified for the plurality of qualified product images, the number of items determined to be qualified for the plurality of unqualified product images, and the number of items determined to be unqualified for the plurality of unqualified product images.

[0134] (Feature 3)

[0135] According to the determination device described in feature 1 or 2, an adjustment process for adjusting the plurality of thresholds is also performed between the second process and the third process.

[0136] (Feature 4)

[0137] The determination device according to feature 3.

[0138] The plurality of processed images includes a plurality of qualified product processed images obtained by applying the selected conditions to the plurality of qualified product images.

[0139] In the adjustment process, a first sub-process and a second sub-process are performed, and the second sub-process is repeated until the item captured in each of the plurality of qualified product processing images in the first sub-process is determined to be qualified, thereby adjusting the plurality of thresholds.

[0140] In the first sub-processing, the qualification or non-qualification of items captured in at least a portion of the qualified product processing images is determined based on the plurality of thresholds.

[0141] In the second sub-processing, if the item captured in one or more of the plurality of qualified product processing images is determined to be unqualified, at least a portion of the plurality of thresholds is relaxed.

[0142] In the third process, the plurality of pixel units contained in each of the plurality of processed images are compared with the plurality of adjusted thresholds respectively.

[0143] (Feature 5)

[0144] According to the determination device of feature 4, in the adjustment process, the initial value of the plurality of thresholds is set such that the item captured in at least one of the plurality of qualified product processing images is determined to be unqualified.

[0145] (Feature 6)

[0146] According to the determination device described in feature 5, in the adjustment process, for each pixel unit, the average value and dispersion of pixel values ​​among the plurality of qualified product images are calculated, and the initial value is set based on the average value and the dispersion.

[0147] (Feature 7)

[0148] The determining device according to any one of features 4 to 6

[0149] In the first sub-processing, for each of the plurality of processed images, one or more abnormal pixel units that deviate from the threshold are extracted from the plurality of pixel units.

[0150] In the second sub-process, at least a portion of one or more of the thresholds for the one or more abnormal pixel units is relaxed.

[0151] (Feature 8)

[0152] The determining device according to any one of features 4 to 7

[0153] In the first sub-processing, for each of the at least a portion of the qualified product processing images among the plurality of qualified product processing images...

[0154] Extract abnormal pixel units that have deviated from the threshold from the plurality of pixel units;

[0155] The abnormal region is inferred based on the abnormal pixel units;

[0156] Based on the characteristics of the abnormal area, the quality of the item is determined.

[0157] (Feature 9)

[0158] The determining device according to feature 8.

[0159] In the second sub-process,

[0160] Select the qualified product processing image from the plurality of qualified product processing images with the largest area of ​​the abnormal region;

[0161] The threshold for the abnormal pixel unit in one of the multiple qualified product processing images is relaxed.

[0162] (Feature 10)

[0163] According to the determination device of feature 8 or 9, the feature of the abnormal region is selected from at least one of the area of ​​the abnormal region, the shape of the abnormal region, and the distribution of the pixel values ​​in the abnormal region.

[0164] (Feature 11)

[0165] According to any one of features 4 to 10, in the repetition of the first sub-process and the second sub-process, the first sub-process is not executed when the qualified product processing image of the article that has been determined to be qualified is captured.

[0166] (Feature 12)

[0167] The determining device according to any one of features 4 to 11,

[0168] The plurality of processed images comprises a plurality of defective product processed images obtained by applying the selected conditions to the plurality of defective product images.

[0169] In the adjustment process, a third sub-process is also performed, in which the passability or failability of the items captured in the multiple defective product processing images is determined based on the multiple thresholds.

[0170] Repeat the first sub-process, the second sub-process, and the third sub-process.

[0171] If, in the third sub-process, the item is determined to be qualified if it is captured at a specified ratio or higher in the multiple defective item processing images, the repeated processing is stopped.

[0172] (Feature 13)

[0173] According to the determination device of feature 12, the initial values ​​of the plurality of thresholds are set such that each of the items captured in the plurality of defective product processing images during the first processing is determined to be defective.

[0174] (Feature 14)

[0175] The determination device according to any one of features 4 to 13 further performs an inspection process in which the qualification of an item captured in an image is determined using one of the determined plurality of conditions and the adjusted plurality of thresholds.

[0176] (Feature 15)

[0177] A determination system having:

[0178] The determining device according to any one of features 1 to 14; and

[0179] Camera device.

[0180] Based on the embodiments described above, a determination apparatus and a determination system are provided that can automatically determine the conditions for image processing. Furthermore, the same effect can be obtained by having a computer execute the determination method described above.

[0181] The above examples illustrate several embodiments of the present invention, but these embodiments are merely illustrative and not intended to limit the scope of the invention. These new embodiments can be implemented in various other ways, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included within the scope or spirit of the invention, and are included within the scope of the invention as described in the claims and its equivalents. Furthermore, the foregoing embodiments can be combined with each other for implementation.

[0182] [Explanation of reference numerals in the attached figures]

[0183] 1: Judgment system; 10: Judgment device; 11: Input device; 12: Display device; 20: Storage device; 30: Camera device; 40: Conveying device; 90: Processing device; 91: CPU; 92: ROM; 93: RAM; 94: Storage device; 95: Input interface; 95a: Input device; 96: Output interface; 96a: Display device; 97: Communication interface; 97a: Server; 98: System bus; 99: Camera; 100 :Conditional Data, 120:Channel, 130:Preprocessing, 150:Window, 160:Channel Candidate, 161:Checkbox, 162:Channel Candidate, 163:Input Field, 170:Preprocessing Candidate, 172-174:Input Field, 200a, 200b:Item, 201-203:Area, 210:Item, 211:Area, 213:Rust, 220a, 220b:Item, 221:Defect, A, A1, A2:Item.

Claims

1. A determining device that uses an image of an article captured by photography to determine whether the article is qualified or not. The determining device repeatedly processes the group of conditions and, based on the evaluation of each of the multiple conditions, determines one of the conditions. The processing group includes: The first process involves selecting any one of the conditions from condition data that includes multiple conditions, wherein the condition specifies the channel to be used and the preprocessing. The second process involves applying the selected conditions to multiple qualified product images and multiple unqualified product images respectively to obtain multiple processed images; The third process involves comparing the results of multiple pixel units and multiple thresholds contained in each of the multiple processed images to determine whether the items captured in each of the multiple processed images are qualified or not. as well as The fourth process involves using one or more judgment results selected from the pass / fail determination results for the plurality of qualified product images and the pass / fail determination results for the plurality of unqualified product images to determine the evaluation for the selected condition. Between the second process and the third process, an adjustment process is also performed to adjust the plurality of thresholds. The plurality of processed images includes a plurality of qualified product processed images obtained by applying the selected conditions to the plurality of qualified product images. In the adjustment process, a first sub-process and a second sub-process are performed, and the second sub-process is repeated until the item captured in each of the plurality of qualified product processing images in the first sub-process is determined to be qualified, thereby adjusting the plurality of thresholds. In the first sub-processing, the qualification or non-qualification of items captured in at least a portion of the qualified product processing images is determined based on the plurality of thresholds. In the second sub-processing, if the item captured in one or more of the plurality of qualified product processing images is determined to be unqualified, at least a portion of the plurality of thresholds is relaxed. In the third process, the plurality of pixel units contained in each of the plurality of processed images are compared with the plurality of adjusted thresholds respectively.

2. The determining device according to claim 1, wherein, In the fourth process, the evaluation for the selected condition is determined by using two or more of the following: the number of items judged as qualified based on the plurality of qualified product images, the number of items judged as unqualified based on the plurality of qualified product images, the number of items judged as qualified based on the plurality of unqualified product images, and the number of items judged as unqualified based on the plurality of unqualified product images.

3. The determining device according to claim 1, wherein, In the adjustment process, the initial values ​​of the plurality of thresholds are set such that the item captured in at least one of the plurality of qualified product processing images is determined to be unqualified.

4. The determining device according to claim 3, wherein, In the adjustment process, for each pixel unit, the average value and dispersion of pixel values ​​among the plurality of qualified product images are calculated, and the initial value is set based on the average value and the dispersion.

5. The determining device according to claim 1, wherein, In the first sub-processing, for each of the at least a portion of the qualified product processing images of the plurality of qualified product processing images, one or more abnormal pixel units that deviate from the threshold are extracted from the plurality of pixel units. In the second sub-process, at least a portion of one or more of the thresholds for the one or more abnormal pixel units is relaxed.

6. The determining device according to claim 1, wherein, In the first sub-processing, for each of the at least a portion of the qualified product processing images of the plurality of qualified product processing images, Extract the abnormal pixel units that deviate from the threshold from the plurality of pixel units. Abnormal regions are inferred based on the aforementioned abnormal pixel units. Based on the characteristics of the abnormal area, the quality of the item is determined.

7. The determining device as described in claim 6, In the second sub-process, Select the qualified product processing image from the plurality of qualified product processing images that has the largest area of ​​the abnormal region. The threshold for the abnormal pixel unit in one of the multiple qualified product processing images is relaxed.

8. The determining device according to claim 6, wherein, The feature of the abnormal region is selected from at least one of the area of ​​the abnormal region, the shape of the abnormal region, and the distribution of pixel values ​​in the abnormal region.

9. The determining device according to claim 1, wherein, In the repetition of the first sub-process and the second sub-process, the first sub-process is not executed for the qualified product processing image of the item that has been captured and has been determined to be qualified.

10. The determining device according to claim 1, wherein, The plurality of processed images comprises a plurality of defective product processed images obtained by applying the selected conditions to the plurality of defective product images. In the adjustment process, a third sub-process is also performed, in which the passability or failability of the items captured in the multiple defective product processing images is determined based on the multiple thresholds. Repeat the first sub-process, the second sub-process, and the third sub-process. If, in the third sub-process, the item is determined to be qualified if it is captured at a specified ratio or higher in the multiple defective item processing images, the repeated processing is stopped.

11. The determining device according to claim 10, wherein, The initial values ​​of the plurality of thresholds are set such that each of the items captured in the plurality of defective product processing images during the first processing is determined to be defective.

12. The determining device according to claim 1, wherein, The system also performs an inspection process in which one of the determined conditions and the adjusted thresholds are used to determine whether the item captured in the image is qualified or not.

13. A determination system, comprising: The determining device according to claim 1 or 2; and Camera device.

14. A method for determining whether an item is qualified or not, using an image of the item captured by a camera. The determination method causes the computer to repeatedly process the group of conditions and, based on the evaluation of each of the multiple conditions, determine one of the conditions. The processing group includes: The first process involves selecting any one of the conditions from condition data that includes multiple conditions, wherein the condition specifies the channel to be used and the preprocessing. The second process involves applying the selected conditions to multiple qualified product images and multiple unqualified product images respectively to obtain multiple processed images; The third process involves comparing the results of multiple pixel units and multiple thresholds contained in each of the multiple processed images to determine whether the items captured in each of the multiple processed images are qualified or not. as well as The fourth process involves using one or more judgment results selected from the pass / fail determination results for the plurality of qualified product images and the pass / fail determination results for the plurality of unqualified product images to determine the evaluation for the selected condition. Between the second process and the third process, an adjustment process is also performed to adjust the plurality of thresholds. The plurality of processed images includes a plurality of qualified product processed images obtained by applying the selected conditions to the plurality of qualified product images. In the adjustment process, a first sub-process and a second sub-process are performed, and the second sub-process is repeated until the item captured in each of the plurality of qualified product processing images in the first sub-process is determined to be qualified, thereby adjusting the plurality of thresholds. In the first sub-processing, the qualification or non-qualification of items captured in at least a portion of the qualified product processing images is determined based on the plurality of thresholds. In the second sub-processing, if the item captured in one or more of the plurality of qualified product processing images is determined to be unqualified, at least a portion of the plurality of thresholds is relaxed. In the third process, the plurality of pixel units contained in each of the plurality of processed images are compared with the plurality of adjusted thresholds respectively.

15. A storage medium storing a program that enables a computer to determine the quality of an article using an image of the article being captured. The program causes the computer to repeatedly process the group of conditions and, based on the evaluation of each of the multiple conditions, determine one of the conditions. The processing group includes: The first process involves selecting any one of the conditions from condition data that includes multiple conditions, wherein the condition specifies the channel to be used and the preprocessing. The second process involves applying the selected conditions to multiple qualified product images and multiple unqualified product images respectively to obtain multiple processed images; The third process involves comparing the results of multiple pixel units and multiple thresholds contained in each of the multiple processed images to determine whether the items captured in each of the multiple processed images are qualified or not. as well as The fourth process involves using one or more judgment results selected from the pass / fail determination results for the plurality of qualified product images and the pass / fail determination results for the plurality of unqualified product images to determine the evaluation for the selected condition. Between the second process and the third process, an adjustment process is also performed to adjust the plurality of thresholds. The plurality of processed images includes a plurality of qualified product processed images obtained by applying the selected conditions to the plurality of qualified product images. In the adjustment process, a first sub-process and a second sub-process are performed, and the second sub-process is repeated until the item captured in each of the plurality of qualified product processing images in the first sub-process is determined to be qualified, thereby adjusting the plurality of thresholds. In the first sub-processing, the qualification or non-qualification of items captured in at least a portion of the qualified product processing images is determined based on the plurality of thresholds. In the second sub-processing, if the item captured in one or more of the plurality of qualified product processing images is determined to be unqualified, at least a portion of the plurality of thresholds is relaxed. In the third process, the plurality of pixel units contained in each of the plurality of processed images are compared with the plurality of adjusted thresholds respectively.

Images