Image processing apparatus and image processing method

Abstract

The learning efficiency or the processing speed of an image processing device is improved. An image processing device (100) is provided with an acquisition unit (110), a division unit (120), a determination unit (140), and an output unit (150). The acquisition unit (110) acquires an original image in which an object to be inspected is reflected. A dividing unit (120) generates a plurality of divided images by dividing the original image on the basis of the features of the object to be inspected reflected in the original image. The determination unit (140) determines, for each of the plurality of divided images, whether or not the qualification of the inspection object can be determined using the rule information, and determines, for a first divided image among the plurality of divided images, whether or not the qualification of the inspection object can be determined using the rule information. And determining whether the inspection object is qualified or not by using the rule information. The output unit (150) outputs, from among the plurality of segmented images, a second segmented image in which the qualification of the inspection object cannot be determined using the rule information, to a learning module (160) which is capable of machine learning or which is obtained as a learning result.

Description

Technical Field

[0001] The present disclosure relates to an image processing device and an image processing method. Background Art

[0002] Conventionally, there is a device that uses an image of an inspection object such as an electronic component to learn or evaluate the inspection object (for example, refer to Patent Document 1).

[0003] Patent document 1 discloses an object evaluation system including a plurality of image acquisition devices, an image output device, and a learning module. The plurality of image acquisition devices have cameras. The plurality of image acquisition devices acquire a first image obtained by photographing an object from a first direction by the camera, and a second image obtained by photographing the object from a second direction. The image output device acquires the first image and the second image. In addition, the image output device inputs the first image and the second image into a learning module capable of machine learning or a learning module obtained as a learning result. More specifically, the first image is input from a first part of an input unit provided in the learning module, and the second image is input from a second part of the input unit.

[0004] Prior Art Literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2018-132962 Summary of the invention

[0007] When a learning module capable of machine learning is made to perform machine learning, for example, it is desired to improve learning efficiency so that machine learning can be performed with a smaller amount of information. In addition, in the learning module obtained as a learning result, that is, the learning module that has performed machine learning, it is desired to improve the processing speed.

[0008] The present disclosure provides an image processing device and an image processing method capable of improving learning efficiency or processing speed.

[0009] An image processing device according to one embodiment of the present disclosure comprises: an acquisition unit, a segmentation unit, a determination unit, and an output unit. The acquisition unit acquires an original image showing an inspection object. The segmentation unit segments the original image based on the features of the inspection object shown in the original image, thereby generating a plurality of segmented images. The determination unit determines, for each of the plurality of segmented images, whether rule information can be used to determine whether the inspection object is qualified or not, and determines whether the inspection object is qualified or not using the rule information for the first segmented image among the plurality of segmented images for which the rule information can be used to determine whether the inspection object is qualified or not using the rule information. The output unit outputs the second segmented image among the plurality of segmented images for which the rule information cannot be used to determine whether the inspection object is qualified or not to a learning module capable of machine learning or obtained as a learning result.

[0010] In addition, in an image processing method involved in one aspect of the present disclosure, an original image reflecting an inspection object is obtained. And, based on the features of the inspection object reflected in the original image, the original image is segmented to generate a plurality of segmented images. And, for each of the plurality of segmented images, it is determined whether the rule information can be used to determine whether the inspection object is qualified or not, and for the first segmented image among the plurality of segmented images in which the rule information can be used to determine whether the inspection object is qualified or not, the rule information is used to determine whether the inspection object is qualified or not. Furthermore, the second segmented image among the plurality of segmented images in which the rule information cannot be used to determine whether the inspection object is qualified or not is output to a learning module that can perform machine learning or is obtained as a learning result.

[0011] According to the present disclosure, it is possible to provide an image processing device and an image processing method that can improve learning efficiency or increase processing speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a block diagram showing an overview of a system including an image processing device according to an embodiment.

[0013] Figure 2 This is a block diagram showing the structure of an image processing device according to an embodiment.

[0014] Figure 3 The diagram is for explaining a processing procedure for determining whether an inspection object is passable or not, which is executed by the image processing device according to the embodiment.

[0015] Figure 4 It is a diagram showing a display image according to the embodiment.

[0016] Figure 5 This is a flowchart showing the processing procedure of the image processing device according to the embodiment. DETAILED DESCRIPTION

[0017] Hereinafter, the embodiments of the present disclosure will be described with reference to the accompanying drawings. In addition, the embodiments described below represent a specific example of the present disclosure. Therefore, the numerical values, shapes, materials, structural elements, configuration positions of structural elements, and connection methods shown in the following embodiments are examples and are not intended to limit the present disclosure. Therefore, among the structural elements in the following embodiments, the structural elements that are not recorded in the independent claims are described as arbitrary structural elements.

[0018] In addition, each figure is a schematic diagram and does not necessarily represent the exact diagram. In addition, in each figure, the same reference numerals are given to the substantially same structures as those in other figures, and repeated descriptions are omitted or simplified.

[0019] (Implementation Method)

[0020] [structure]

[0021] First, the configuration of the image processing device 100 according to the embodiment will be described.

[0022] Figure 1 1 is a block diagram showing an overview of a system including the image processing apparatus 100 according to the embodiment.

[0023] The image processing device 100 is a device that determines whether the inspection object 400 is acceptable based on an image (original image) generated by an imaging device 200 such as a camera capturing an inspection object (workpiece) 400. The image processing device 100 performs a predetermined image processing on the original image obtained by capturing the inspection object 400, and performs image analysis on the image on which the predetermined image processing is performed, thereby determining whether the inspection object 400 reflected in the image is acceptable.

[0024] The so-called pass / fail determination of the inspection object 400 refers to, for example, the determination of whether the inspection object 400 has scratches, notches, cracks, etc., or the determination of whether appropriate text is printed at an appropriate position, etc. The image processing device 100 uses rule information that specifies a given threshold value, etc. to determine whether the inspection object reflected in the image is pass / fail.

[0025] The rule information is information used to determine whether the inspection object 400 is acceptable. The rule information includes, for example, information indicating a range of a brightness threshold value, information indicating a range of a color threshold value, information indicating printed text, and information indicating an appearance. The image processing device 100 uses the rule information to determine whether the inspection object 400 is acceptable. Figure 3 , Figure 4 The pass / fail determination of the inspection object 400 shown in the segmented image 310 described later is performed.

[0026] In addition, for example, the image processing device 100 causes the display device 210 to display an image in which the inspection object 400 cannot be judged as pass or fail using the rule information, that is, an image in which the pass or fail judgment cannot be performed. For example, the inspector confirms the image and inputs the pass or fail judgment result of the inspection object 400 shown in the image using the input device 220. The image processing device 100 obtains the judgment result and performs machine learning, thereby being able to judge the pass or fail of the inspection object 400 shown in the image using AI (Artificial Intelligence) or the like.

[0027] The image processing device 100 is, for example, a computer such as a personal computer or a tablet terminal. Specifically, for example, the image processing device 100 is implemented by a communication interface for communicating with the camera 200, the display device 210, and the input device 220, a non-volatile memory storing a program, a volatile memory as a temporary storage area for executing the program, an input / output port for transmitting and receiving signals, a processor for executing the program, etc. The communication interface can be implemented by a connector connected with a communication line to enable wired communication, or can be implemented by an antenna and a wireless communication circuit to enable wireless communication.

[0028] Figure 2 It is a block diagram showing the structure of the image processing device 100 according to the embodiment. Figure 3 The diagram is for explaining a processing procedure for determining whether the inspection object 400 is acceptable or not, which is executed by the image processing device 100 according to the embodiment.

[0029] like Figure 2 As shown, the image processing device 100 includes an acquisition unit 110 , a segmentation unit 120 , a generation unit 130 , a determination unit 140 , an output unit 150 , a learning module 160 , a reception unit 170 , an association unit 180 , and a storage unit 190 .

[0030] The acquisition unit 110 is to acquire Figure 3 The processing unit of the original image 300 of the inspection object 400 is shown. Figure 3 As shown in (a) of FIG. 8 , the acquisition unit 110 acquires one or more original images 300 from the imaging device 200 , for example.

[0031] The inspection target 400 is, for example, an electronic component such as an IC (Integrated Circuit).

[0032] In addition, the inspection target 400 may be any target object such as a substrate instead of an electronic component.

[0033] Furthermore, for example, when the acquisition unit 110 acquires a plurality of original images, the inspection object shown in each original image may be the same inspection object or may be the same type of inspection object but different inspection objects.

[0034] Alternatively, the acquisition unit 110 may acquire the original image 300 from a server device or the like via a communication interface provided in the image processing device 100 .

[0035] The segmentation unit 120 is a processing unit that generates a plurality of segmented images 310 by segmenting the original image 300 based on the features of the inspection target object 400 reflected in the original image 300 .

[0036] Furthermore, when the acquisition unit 110 acquires a plurality of original images, the division unit 120 may divide each original image into the same number of divided images or into different numbers of divided images.

[0037] Furthermore, the characteristics of the inspection object 400 may be set arbitrarily.

[0038] The segmentation unit 120 divides the inspection object 400 into Figure 3 The original image 300 showing the inspection object 400 shown in (a) is segmented. Figure 3 As shown in (b) of FIG. 1 , the segmentation unit 120 segments the original image 300 into: (i) a portion of the inspection object 400 where characters such as “ABCDE” are printed (or engraved); Figure 3 (ii) an end portion of the inspection object 400 including a boundary portion between the inspection object 400 and the background in the original image 300 ( Figure 3 and (iii) a surface portion of the inspection object 400 excluding the second region and the third region ( Figure 3 In this example, the segmentation unit 120 segments the first region of the original image 300 into 12 images, the second region into 1 image, and the third region into 4 images.

[0039] For example, in the inspection of the surface portion (for example, the inspection of a portion of the inspection object 400 that is flat and has a relatively flat surface brightness value), the presence or absence of defects such as scratches on the surface is inspected.

[0040] Furthermore, for example, in the inspection of the end (for example, the inspection of the boundary between the inspection object and the background), the end of the inspection object 400 is inspected for defects such as chipping and deformation.

[0041] In addition, for example, in the inspection of printing (for example, the inspection of a characteristic portion of the shape of characters), the characters are inspected for defects such as skewness, white spots, and positional deviation.

[0042] In this way, the content of the inspection may be different depending on the position of the inspection object 400, that is, the feature of the inspection object 400 (more specifically, the feature of each position of the inspection object 400). For example, the segmentation unit 120 segments the original image 300 according to the content of each inspection, that is, each feature of the inspection object 400.

[0043] In addition, the number of segmented images 310 generated by the segmentation unit 120 is not particularly limited. Figure 3 In the example shown, the dividing unit 120 divides the original image 300 into three regions, namely, a first region, a second region, and a third region. However, the original image 300 may be divided into two regions or may be divided into four or more regions.

[0044] In addition, Figure 3 In the example shown, the features of the inspection target 400 are the printed portion, the end portion, and the surface portion as described above (i) to (iii), but the features are not limited to these examples and can be arbitrarily set.

[0045] In addition, the shape (more specifically, the outer shape) of the segmented image 310 generated by the segmentation unit 120 can be arbitrary. In this embodiment, the segmentation unit 120 divides the original image into a plurality of segmented images 310, each of which is rectangular. For example, the segmentation unit 120 can also divide the original image 300 into segmented images 310 in an L shape.

[0046] In addition, the features of the inspection object 400 may be obtained from the inspector via the input device 220, for example, by obtaining information representing the features. In addition, for example, the segmentation unit 120 may also use a recognition model (feature recognition model) that has been machine-learned for recognizing the features of the inspection object 400 to determine the features in the original image 300. The feature recognition model is, for example, an inference model that has been machine-learned using the original image and training data (so-called annotation information) representing the features of the inspection object contained in the original image as learning data. The feature recognition model is, for example, a machine learning model that uses a neural network such as deep learning (Deep Learning) (for example, a convolutional neural network (CNN)), but may also be other machine learning models.

[0047] The feature recognition model is stored in advance in the storage unit 190 , for example.

[0048] Figure 4 3 is a diagram showing a display image 330 according to the embodiment. The generator 130 generates the inspection image 320 (see Figure 4 ), the inspection image 320 includes one or more segmented images 310 selected based on the characteristics of the inspection object 400 from among the plurality of segmented images 310 generated by the segmentation unit 120. Specifically, the generation unit 130 classifies the plurality of segmented images 310 based on the characteristics of the inspection object 400, and generates a plurality of inspection images 320 by combining one or more segmented images 310 classified into the same group from among the plurality of segmented images 310. In the present embodiment, the one or more segmented images 310 are two or more segmented images 310. That is, the generation unit 130 generates the inspection image 320 including the two or more segmented images 310. For example, the generation unit 130 performs pre-processing such as reduction and emphasis on the segmented images 310 generated by the segmentation unit 120 for maintaining the detection rate, and then directly arranges the segmented images 310 without gaps according to each segmented image 310 corresponding to the characteristics of the inspection object 400, thereby generating the inspection image 320.

[0049] For example, the generator 130 classifies the plurality of segmented images 310 generated by the segmentation unit 120 into a plurality of groups grouped for each feature of the inspection target 400. The generator 130 further generates an inspection image 320 in which the segmented images 310 are grouped for each of the classified groups.

[0050] Each group is a group of segmented images 310 for each feature of the inspection target 400 such as the printed portion, the end portion, and the surface portion as described above (i) to (iii), for example.

[0051] Furthermore, the generating unit 130 may perform image processing such as changing the size, brightness, or color of the divided image 310 .

[0052] For example, the generation unit 130 may also perform an enhancement on the defects included in the inspection object 400 on the segmented image 310 (for example, Figure 3 Such image processing includes, for example, image processing to increase the brightness difference, processing to increase the contrast, or processing to increase the size of a portion estimated to be a defect.

[0053] For example, the generator 130 generates the inspection image 320 on which a given process is performed based on the brightness of one or more segmented images 310. For example, the generator 130 generates the inspection image 320 in which the average value of the brightness of one or more segmented images 310 is calculated as the given process, and the brightness of one or more segmented images 310 is corrected and arranged based on the calculated average value. In addition, for example, as the given process, when two or more segmented images 310 are arranged, the generator 130 corrects the brightness gradient of at least a part of the adjacent segmented images 310 among the two or more segmented images 310. In addition, the part for which the brightness gradient is corrected may be one or both of the adjacent segmented images. In addition, for example, as the given process, the generator 130 determines the arrangement of the two or more segmented images 310 so that the brightness difference between the adjacent segmented images 310 among the two or more segmented images 310 is minimized, and generates the inspection image 320 in which the two or more segmented images 310 are arranged according to the determined arrangement.

[0054] In addition, for example, when two or more segmented images 310 are arranged, the generation unit 130 generates the inspection image 320 by performing a blurring process such as alpha blending between adjacent segmented images 310 among the two or more segmented images 310. In addition, the so-called adjacent segmented images 310 refer to, for example, a portion where images in the adjacent segmented images 310 are connected to each other in at least one of the adjacent segmented images 310. The so-called adjacent segmented images 310 may also refer to, for example, an image that is arranged between the adjacent segmented images 310 and has been blurred.

[0055] In addition, for example, the generation unit 130 may also perform image processing to change the size of the segmented image (specifically, reduce it). For example, the generation unit 130 may perform processing to reduce one or more segmented images 310 and improve the brightness contrast of the one or more segmented images 310 before generating the inspection image 320. In addition, the generation unit 130 may perform processing to improve the brightness contrast after reducing the segmented image 310, or may perform processing to improve the brightness contrast before reducing the segmented image 310.

[0056] In addition, the same image processing may be performed on each divided image 310, or different image processing may be performed on each divided image 310. In addition, different image processing may be performed on each group.

[0057] For example, after performing a predetermined image process on the divided image 310 , the generating unit 130 determines the arrangement of the divided images 310 , and generates the inspection image 320 arranged in accordance with the determined arrangement.

[0058] Furthermore, the generating unit 130 may generate the inspection image based on a plurality of original images, or may generate the inspection image based on a single original image.

[0059] Alternatively, the generator 130 may determine the arrangement of the plurality of divided images 310 regardless of the positional relationship of the original image 300 , and generate the inspection image 320 in which the plurality of divided images 310 are arranged in the determined arrangement.

[0060] In addition, for example, Figure 3 As shown in (b) of FIG. 3 , in a portion that is not included in any of the first region, the second region, or the third region in the original image 300 , a segmented image 310 may be generated as the fourth region, or may be discarded.

[0061] The arrangement of the segmented images 310 in the inspection image 320 may be determined arbitrarily. For example, the generator 130 determines the arrangement of one or more segmented images 310 based on the characteristics of the inspection object 400, and generates one or more inspection images in which the one or more segmented images 310 are arranged according to the determined arrangement.

[0062] The number of segmented images 310 included in the inspection image 320 may be one or more. For example, the generation unit 130 generates the inspection image 320 for each feature of the inspection object 400. For example, the generation unit 130 generates the inspection image 320 including the segmented image 310 of the printing portion, the inspection image 320 including the segmented image 310 of the end portion, and the inspection image 320 including the segmented image 310 of the surface portion. The generation unit 130 may generate the inspection image 320 in which the segmented image 310 of the printing portion, the segmented image 310 of the end portion, and the segmented image 310 of the surface portion are respectively collected at different positions (for example, the upper part, the middle part, and the lower part, or the left part, the middle part, and the right part, etc. in the inspection image 320).

[0063] The determination unit 140 is a processing unit that determines whether the inspection object 400 can be determined to be acceptable or not using rule information for each of the plurality of segmented images 310. Specifically, the determination unit 140 determines whether the inspection object 400 shown in the segmented image 310 can be determined to be acceptable or not using rule information, more specifically, without using machine learning. For example, for the first segmented image among the plurality of segmented images 310 in which the inspection object 400 can be determined to be acceptable or not using rule information, the determination unit 140 determines the acceptable or not of the inspection object 400 (more specifically, the acceptable or not of the portion of the inspection object 400 shown in the segmented image 310) using the rule information.

[0064] For example, the determination unit 140 performs image measurement such as brightness measurement on the segmented image 310, and if the value of the measured brightness or the like is a value greater than a given first threshold value or less than a second threshold value that is smaller than the first threshold value, the inspection object 400 in the segmented image 310 is determined to be unqualified. In addition, for example, if the value of the measured brightness or the like is a value less than a given first threshold value and greater than a second threshold value, the determination unit 140 determines that the inspection object 400 in the segmented image 310 is qualified.

[0065] In addition, the content of the image measurement may be the brightness gradient, brightness variance, brightness deviation, or the size of a portion including a given number or more of pixels in a given brightness range of the segmented image 310. In addition, for example, the determination unit 140 may determine that the inspection object 400 in the segmented image 310 including pixels of a specific brightness is unqualified, or may determine that the inspection object 400 in the segmented image 310 including a given number or more of pixels in a given brightness range is unqualified. In addition, for example, the determination unit 140 may determine that the inspection object 400 in the segmented image 310 cannot be determined to be qualified or not, for example, if the brightness measurement or the like cannot be performed on the segmented image 310.

[0066] These measurement contents, threshold values, and determination methods may be stored in the storage unit 190 as rule information, for example.

[0067] Furthermore, for example, when the generator 130 generates a plurality of inspection images 320 each including a plurality of segmented images 310 , the determiner 140 determines whether the inspection object 400 can be determined to be pass / fail using the rule information for each of the plurality of inspection images 320 .

[0068] The output unit 150 is a processing unit that outputs the segmented image 310 and the result of the pass / fail determination of the inspection object 400. For example, the output unit 150 outputs the second segmented image, for which the rule information cannot be used to determine the pass / fail of the inspection object 400 among the plurality of segmented images 310, to a learning module 160 that can perform machine learning or is obtained as a learning result. In addition, the output unit 150 causes the display device 210 to display the second segmented image, for example, by outputting the second segmented image to the display device 210. For example, the output unit 150 causes the display device 210 to display the inspection image 320 generated by the generation unit 130 including the second segmented image. In the present embodiment, the output unit 150 outputs a display image 330 (see FIG. 2 ) including the inspection image 320 to the display device 210. Figure 4), so that the display device 210 displays the display image 330. The output unit 150 may also output the inspection image 320 to the storage unit 190 and cause the storage unit 190 to store the inspection image 320.

[0069] Furthermore, the output unit 150 may cause the display device 210 to display the first divided image by outputting the determination result of the determination unit 140 on the inspection target 400 in the first divided image to the display device 210 .

[0070] For example, Figure 3 As shown in (c) of FIG. 1 , the determination unit 140 determines whether or not the inspection object 400 shown in the divided image 310 can be determined to be acceptable for each of the 12 divided images 310 . Figure 3 As shown in (d), the output unit 150 causes the display device 210 to display a segmented image 311 (for example, an inspection image 320 or a display image 330 including the segmented image 311) among the 12 segmented images 310 in which the determination unit 140 cannot determine whether the inspection object 400 reflected in the segmented image 310 is pass or fail, or outputs the segmented image 311 to the learning module 160.

[0071] The learning module 160 is a learning module that can perform machine learning or is obtained as a result of learning. Specifically, the learning module 160 is a processing unit that uses a recognition model (defect recognition model) that takes the segmented image 310 as an input and outputs a result of determining whether the inspection object 400 reflected in the input segmented image 310 is qualified or not.

[0072] The defect recognition model is, for example, an inference model that performs machine learning using an image and training data representing the pass / fail determination result of the inspection object 400 shown in the image as learning data. The defect recognition model is, for example, a machine learning model using a neural network such as deep learning, but may also be other machine learning models.

[0073] For example, the learning module 160 is a learning module obtained as a learning result, that is, a learning module that has been machine-learned. In this case, for example, the output unit 150 outputs the result of the learning module 160 on the pass / fail determination of the inspection object 400 in the second segmented image to the display device 210. Thus, for example, the result of the pass / fail determination of the inspection object 400 is displayed on the display device 210.

[0074] In addition, the output unit 150 may output both the qualified determination result and the unqualified determination result to the display device 210 as the qualified or unqualified determination result of the inspection object 400, or may output only the qualified determination result or only the unqualified determination result. In addition, the output unit 150 may output the original image 300, the segmented image 310, or the inspection image 320 that has been determined together with the determination result. In addition, the output unit 150 may output an identifier such as a manufacturing number that uniquely indicates the inspection object 400 together with the determination result. The identifier may be the file name of the file of the original image 300, or may be text printed in the printing unit, etc.

[0075] The receiving unit 170 is a processing unit that receives the operation of the inspector. The receiving unit 170 receives the operation of the inspector, for example, via the input device 220. The receiving unit 170 receives, for example, the result of the determination of whether the second segmented image is acceptable or not. The inspector observes the inspection image 320 displayed on the display device 210, for example, and uses the input device 220 to input information indicating that a segmented image 310 including the defect 500 is included in a plurality of segmented images 310 (for example, the second segmented image) included in the inspection image 320. The receiving unit 170 receives, for example, the input as a determination result. In this case, for example, the output unit 150 outputs the second segmented image and the result of the determination of whether the inspection object 400 in the second segmented image received by the receiving unit 170 to the learning module 160, so that the learning module 160 performs machine learning.

[0076] Figure 4 It is a diagram showing a display image 330 according to the embodiment.

[0077] The display image 330 is an image displayed on the display device 210. The display image 330 includes an inspection image area and a UI (User Interface) area.

[0078] The inspection image region is located in the center of the display image 330 and is a region where the inspection image 320 is arranged.

[0079] The UI area is located at the periphery of the display image 330 and is an area where the operation image is configured. For example, the central field of view is an elliptical area 340. Therefore, the UI area is configured at the periphery of the display image 330. Figure 4 In the case of the horizontally long display image 330 shown, the center field of view may be arranged at both left and right ends of the display image 330. The center field of view refers to a region near the center of the field of view of a person (here, the operator).

[0080] For example, the inspector operates the input device 220 to move, for the multiple segmented images 310 included in the inspection image 320, a segmented image 310 that shows a qualified product (for example, a segmented image 310 that does not include the defect 500) to an area marked with "OK", and moves a segmented image 310 that shows an unqualified product (for example, a segmented image 310 that includes the defect 500) to an area marked with "NG".

[0081] The input device 220 outputs the result of the movement of the segmented image 310 as a result of determining whether the inspection object 400 shown in the segmented image 310 is good or bad to the image processing device 100. Alternatively, the input device 220 may output to the image processing device 100 a result of determining that the inspection object 400 in all the segmented images 310 that have not been moved to "NG" is good or bad.

[0082] For example, the output unit 150 outputs the second divided image and the pass / fail determination result of the inspection object 400 in the second divided image received by the receiving unit 170 to the learning module 160 , thereby causing the learning module 160 to perform machine learning.

[0083] The defect recognition model is stored in advance in the storage unit 190 , for example.

[0084] In addition, the learning module 160 may be a learning module capable of machine learning, that is, a learning module that starts machine learning from now on, or a learning module obtained as a learning result, that is, a learning module that has completed machine learning. In other words, the image processing device 100 may be a device that uses the second segmented image to make the learning module 160 perform machine learning, or a device that makes the learning module 160 determine whether the inspection object 400 reflected in the second segmented image is qualified or not. The image processing device 100 may include the learning module 160, or may not include the learning module 160 but be connected to a device that includes the learning module 160 so as to be able to communicate.

[0085] In addition, Figure 4 The following structure is shown in which the result of the pass or fail judgment of the inspection object 400 in the split image 310 (projected in the split image 310) is input by moving the split image 310 to either the UI area of ​​"OK" or the UI area of ​​"NG", but the structure may also be a structure in which the result of the pass or fail judgment of the inspection object 400 in the split image 310 is input by clicking buttons such as "OK" and "NG".

[0086] Furthermore, the output unit 150 may cause the display device 210 to display information indicating the position of the divided image 310 in the original image 300 on which the divided image 310 is based.

[0087] The associating unit 180 associates the plurality of segmented images 310 with the pass / fail determination results of the inspection object 400 in the plurality of segmented images 310 and stores the associations in the storage unit 190. The pass / fail determination results of the inspection object 400 may be the results based on the determination unit 140, the results based on the learning module 160, or the results received by the receiving unit 170.

[0088] In addition, for example, the associating unit 180 associates the unqualified image determined by the determination unit 140 or the learning module 160 as the inspection object 400 unqualified among the plurality of segmented images 310 with the determination result indicating the meaning of unqualified, and stores it in the storage unit 190. Thereafter, for example, the output unit 150 outputs the unqualified image to the display device 210, so that the display device 210 displays the unqualified image. In addition, for example, the receiving unit 170 receives the determination result of whether the inspection object 400 in the unqualified image is qualified. In these cases, for example, when the determination result of whether the inspection object 400 in the unqualified image received by the receiving unit 170 is qualified, the associating unit 180 changes the determination result associated with the unqualified image and stored in the storage unit 190 to the determination result indicating the meaning of qualified. That is, the image processing device 100 receives information from the inspector whether the determination result of the inspection object 400 by the determination unit 140 or the learning module 160 is correct, and if it is wrong, corrects the determination result and stores it in the storage unit 190.

[0089] In addition, for example, in the case where the learning module 160 makes an erroneous judgment on whether the inspection object 400 is qualified or not, the learning module 160 may also use the unqualified image and the judgment result received by the receiving unit 170 (specifically, the correct judgment result received from the inspector) as learning data for machine learning.

[0090] Furthermore, the learning module 160 may perform machine learning using the images and determination results stored in the storage unit 190 as learning data.

[0091] In addition, when the determination result of the determination unit 140 on the pass / fail of the inspection object 400 is erroneous, the rule information (for example, the threshold of brightness etc. included in the rule information) is considered to be inappropriate. The receiving unit 170 may also receive new rule information, for example, update the rule information stored in the storage unit 190 to the new rule information.

[0092] Furthermore, the associating unit 180 may cause the storage unit 190 to store all of the plurality of segmented images 310 generated by the segmentation unit 120, or may cause the storage unit 190 to store only a part of them. For example, the associating unit 180 may cause the storage unit 190 to store the same number of segmented images 310 for which the result of the determination of the pass or fail of the inspection object 400 is pass and the segmented images 310 for which the determination result is fail in the plurality of segmented images 310. For example, the associating unit 180 associates all of the segmented images 310 for which the result of the determination of the pass or fail of the inspection object 400 is fail and the same number of segmented images 310 for which the result of the determination of the pass or fail of the inspection object 400 is pass with the determination result, and stores them in the storage unit 190.

[0093] The segmented images 310 stored in the storage unit 190 by the associating unit 180 and the segmented images 310 in which the result of the pass / fail determination of the inspection object 400 is passed can be selected arbitrarily. For example, the associating unit 180 may randomly select from a plurality of segmented images 310 including the inspection object 400 determined to be passed and store them in the storage unit 190. In addition, for example, the associating unit 180 may determine the segmented images 310 stored in the storage unit 190 including the inspection object 400 determined to be passed based on the segmented images 310 in which the result of the pass / fail determination of the inspection object 400 is failed in the original image 300. For example, the associating unit 180 may cause the storage unit 190 to store the segmented images 310 in which the result of the pass / fail determination of the inspection object 400 is passed and located near the segmented images 310 in which the result of the pass / fail determination of the inspection object 400 is failed in the original image 300. In addition, for example, the associating unit 180 may store a segmented image 310 in an original image 300 different from the segmented image 310 (that is, a different inspection object 400 from the inspection object 400 determined to be unqualified) at the same position as the segmented image 310 in which the inspection object 400 is determined to be unqualified, and a segmented image 310 reflecting the inspection object 400 determined to be qualified. In this way, the associating unit 180 may cause the storage unit 190 to store a segmented image 310 having features close to the segmented image 310 in which the inspection object 400 is determined to be unqualified, and a segmented image 310 in which the inspection object 400 is determined to be qualified. In addition, for example, the associating unit 180 may perform brightness measurement of the segmented image 310, and cause the storage unit 190 to store a segmented image 310 having a brightness value of a given value.

[0094] In addition, the associating unit 180 may associate the plurality of segmented images 310 with information indicating the features of the inspection object 400 and store them in the storage unit 190. That is, the associating unit 180 may cause the storage unit 190 to store the plurality of segmented images 310 for each feature such as the surface portion, the end portion, or the printed portion.

[0095] The acquisition unit 110 , the division unit 120 , the generation unit 130 , the determination unit 140 , the output unit 150 , the learning module 160 , the reception unit 170 , and the association unit 180 are implemented by, for example, one or more processors.

[0096] The storage unit 190 is a storage device that stores programs executed by the processing units such as the acquisition unit 110, the segmentation unit 120, the generation unit 130, the determination unit 140, the output unit 150, the learning module 160, the reception unit 170, and the association unit 180 for performing each process, information required for the process, and a plurality of segmented images 310 or inspection images 320. The storage unit 190 is implemented, for example, by a HDD (Hard Disk Drive) or a semiconductor memory.

[0097] The imaging device 200 is a camera that generates an original image by imaging an inspection object, and is implemented by, for example, a CMOS (Complementary Metal Oxide Semiconductor) image sensor or the like.

[0098] The display device 210 is a display that displays an image based on the control of the image processing device 100 (more specifically, the output unit 150). The display device 210 displays, for example, an inspection image, a display image, or a determination result of the determination unit 140 or the learning module 160. The display device 210 is implemented by, for example, a display device such as a liquid crystal panel or an organic EL (Electro Luminescence) panel.

[0099] The input device 220 is a user interface for receiving an inspector's operation and is implemented by a mouse, a keyboard, a touch panel, or a hardware button.

[0100] [Processing order]

[0101] Next, a processing procedure of the image processing device 100 according to the embodiment will be described.

[0102] Figure 5 This is a flowchart showing the processing procedure of the image processing device 100 according to the embodiment.

[0103] First, the acquisition unit 110 acquires the original image 300 showing the inspection object 400 ( S110 ).

[0104] Next, the segmentation unit 120 segments the original image 300 acquired by the acquisition unit 110 based on the features of the inspection target 400 reflected in the original image 300 , thereby generating a plurality of segmented images 310 ( S120 ).

[0105] For example, the segmentation unit 120 performs position alignment so that the inspection object 400 included in the original image 300 is located at a given position. The method of position alignment may be a general method such as edge extraction or pattern consistency, and is not particularly limited. The segmentation unit 120 may also perform position alignment of the inspection object 400 based on the brightness of each position (for example, each pixel) of the original image 300. Next, the segmentation unit 120 specifies the area of ​​each feature of the inspection object 400, such as the surface portion, end portion, and print portion described above. The designation of the area may be specified by the inspector, or may be automatically specified by the segmentation unit 120 using the above-mentioned machine learning, etc. The segmentation unit 120 segments the original image 300 according to each specified area, thereby generating a plurality of segmented images 310.

[0106] Furthermore, the segmentation unit 120 may segment the designated area so that there is no gap, or may segment the original image 300 so that the segmented images 310 partially overlap, that is, the segmented images 310 partially contain the same image portion.

[0107] The size of the divided image 310 can be arbitrarily determined and is not particularly limited. The size of the divided image 310 can be predetermined with a specific numerical value or can be determined according to the size of the inspection object 400, such as 1 / 10 of the inspection object 400.

[0108] In addition, the characteristics of the inspection object 400 can be determined by the combination of colors. In addition, the classification of characteristics does not need to cover all patterns included in the original image 300. For example, even if the inspection object 400 is composed of 5 colors, 3 colors are set as the characteristics of the inspection object 400. That is, the characteristics of the inspection object 400 do not need to be set in a manner that includes all the characteristics of the inspection object 400.

[0109] Furthermore, the segmentation unit 120 may store the original image 300 serving as a basis for the segmented image 310 and information indicating the position of the segmented image 310 in the original image 300 in the storage unit 190 .

[0110] In addition, the segmentation unit 120 may not perform position alignment of the inspection object 400 in the original image 300. For example, the original image 300 may be segmented so that the segmented image 310 has a given size. The given size may be determined arbitrarily and is not particularly limited. Information indicating the given size may also be stored in the storage unit 190, for example.

[0111] In addition, the segmentation unit 120 may determine how to classify the segmented images 310 for each feature based on the color when the feature of the inspection object 400 is color, for example, when the original image 300 is segmented so that the segmented images 310 have a given size without aligning the position of the inspection object 400 in the original image 300. For example, when the surface of the inspection object 400 is black, the text portion is white, and the background is gray, the segmentation unit 120 may determine the segmented image 310 containing only black as the surface portion, the segmented image 310 containing both black and white as the end portion, the segmented image 310 containing only white as the printed portion, and the segmented image 310 containing only gray as the other portion.

[0112] Next, the determination unit 140 determines whether or not the inspection object 400 can be determined to be acceptable using the rule information for each of the plurality of segmented images 310 ( S130 ).

[0113] Next, the determination unit 140 determines the pass / fail status of the inspection object 400 using the rule information for the first divided image among the plurality of divided images 310 for which the pass / fail status of the inspection object 400 can be determined using the rule information ( S140 ).

[0114] Next, the output unit 150 outputs the second segmented image, for which the inspection object 400 cannot be judged as pass / fail using the rule information, among the plurality of segmented images 310 to the learning module 160 capable of machine learning or obtained as a learning result ( S150 ).

[0115] Thus, the output unit 150 causes the learning module 160 to perform machine learning using the second divided image, or causes the learning module 160 to determine whether the inspection target 400 shown in the second divided image is acceptable.

[0116] [Summarize]

[0117] As described above, the image processing device 100 involved in the embodiment comprises: an acquisition unit 110, which acquires an original image 300 reflecting an inspection object 400; a segmentation unit 120, which segments the original image 300 based on the features of the inspection object 400 reflected in the original image 300, thereby generating a plurality of segmented images 310; a determination unit 140, which determines, for each of the plurality of segmented images 310, whether rule information can be used to determine whether the inspection object 400 is qualified or not, and for the first segmented image among the plurality of segmented images 310 for which rule information can be used to determine whether the inspection object 400 is qualified or not, uses the rule information to determine whether the inspection object 400 is qualified or not; and an output unit 150, which outputs the second segmented image among the plurality of segmented images 310 for which rule information cannot be used to determine whether the inspection object 400 is qualified or not, to a learning module 160 that can perform machine learning or is obtained as a learning result.

[0118] Thus, the image processing device 100 uses the rule information to make a determination of whether the inspection object is qualified or not, for images that can be easily determined using the rule information, and allows the learning module 160 to learn for images that cannot be determined using the rule information, or allows the learning module 160 to make the determination if the learning module 160 has completed learning. Therefore, the learning module 160 can learn only for images that cannot be determined using the rule information, or can make the determination only for images that cannot be determined using the rule information. Therefore, according to the image processing device 100, since the learning module 160 is not allowed to process images that can be determined using the rule information, the learning efficiency of the learning module 160 can be improved, or the processing speed of the learning module 160 can be improved (that is, the speed for obtaining the result of the determination can be improved).

[0119] Furthermore, for example, the output unit 150 outputs the second divided image to the display device 210 to cause the display device 210 to display the second divided image.

[0120] This makes it possible to notify the inspector of an image for which the pass / fail determination of the inspection object cannot be performed using the rule information.

[0121] In addition, for example, the image processing device 100 also includes: a receiving unit 170, which receives the pass / fail judgment result of the second segmented image, and an output unit 150 outputs the second segmented image and the pass / fail judgment result of the inspection object 400 in the second segmented image received by the receiving unit 170 to the learning module 160, so that the learning module 160 performs machine learning.

[0122] This allows the learning module 160 to learn images for which the pass / fail determination of the inspection object cannot be performed using the rule information.

[0123] In addition, for example, the image processing apparatus 100 further includes a correlating unit 180 that correlates the plurality of segmented images 310 and the pass / fail determination results of the inspection object 400 in the plurality of segmented images 310 , and stores the associations in the storage unit 190 .

[0124] Thus, for example, by an inspector checking the segmented image and the determination result, the inspector can check whether the determination using the rule information and the determination by the learning module 160 are appropriately performed.

[0125] For example, the associating unit 180 stores the same number of divided images 310 for which the inspection object 400 is judged as acceptable and the same number of divided images 310 for which the inspection object 400 is judged as unacceptable in the storage unit 190 .

[0126] Therefore, in machine learning, it is preferable that the number of images determined to be qualified and the number of images determined to be unqualified are equal. Thus, for example, by making the learning module 160 learn these images and store them, the learning module 160 can learn effectively, and the learned images can be confirmed by inspectors, etc.

[0127] For example, the image processing device 100 includes the learning module 160 obtained as a learning result. In this case, for example, the output unit 150 outputs the result of the learning module 160 determining whether the inspection object 400 in the second divided image is pass or fail to the display device 210 .

[0128] Thereby, for example, an inspector can confirm the result of the pass / fail determination of the inspection object and discard the inspection object determined to be unqualified.

[0129] In addition, for example, the image processing device 100 further includes: an associating unit 180 that associates an unqualified image determined by the determination unit 140 or the learning module 160 as the inspection object 400 as unqualified among the plurality of segmented images 310 with a determination result indicating that the inspection object 400 is unqualified, and stores the result in the storage unit 190; and a receiving unit 170 that receives the determination result of whether the inspection object 400 in the unqualified image is qualified or not. In addition, for example, the output unit 150 outputs the unqualified image to the display device 210. In these cases, for example, when the determination result of whether the inspection object 400 in the unqualified image received by the receiving unit 170 is qualified or not, the associating unit 180 changes the determination result associated with the unqualified image and stored in the storage unit 190 to a determination result indicating that the inspection object 400 is qualified.

[0130] Therefore, for example, in a case where the image processing device 100 makes an incorrect judgment on whether the inspection object is qualified or not, the accuracy of the judgment on whether the inspection object is qualified or not by the image processing device 100 is improved by causing the learning module 160 to perform machine learning or change the rule information based on the judgment result received by the receiving unit 170.

[0131] In addition, for example, the image processing device 100 further includes: a generating unit 130 that classifies the plurality of segmented images 310 based on the features of the inspection object 400, and generates a plurality of inspection images 320 formed by combining one or more segmented images 310 classified into the same group among the plurality of segmented images 310. In this case, for example, the determining unit 140 determines whether the inspection object 400 can be determined to be acceptable using the rule information for each of the plurality of inspection images 320.

[0132] Thus, for example, the inspection image in which the segmented images are aggregated for each feature is output to the learning module 160, thereby simplifying the neural network in the learning module 160. For example, compared with the case where the learning module 160 is made to learn using an image that is not aggregated for each feature, the learning module 160 can be made to learn so that the determination result can be derived efficiently (for example, the processing speed becomes faster).

[0133] Furthermore, for example, the output unit 150 outputs the determination result of the inspection target 400 in the first divided image by the determination unit 140 to the display device 210 .

[0134] This makes it possible to notify the inspector of an image that enables pass / fail determination of the inspection object using the rule information.

[0135] In addition, in the image processing method involved in the embodiment, an original image 300 reflecting the inspection object 400 is obtained (S110), the original image 300 is segmented based on the characteristics of the inspection object 400 reflected in the original image 300, thereby generating a plurality of segmented images 310 (S120), and for each of the plurality of segmented images 310, it is determined whether rule information can be used to determine whether the inspection object 400 is qualified (S130), and for the first segmented image among the plurality of segmented images 310 in which rule information can be used to determine whether the inspection object 400 is qualified, the rule information is used to determine whether the inspection object 400 is qualified (S140), and the second segmented image among the plurality of segmented images 310 in which rule information cannot be used to determine whether the inspection object 400 is qualified is output to a learning module 160 that can perform machine learning or is obtained as a learning result (S150).

[0136] This achieves the same effects as those of the image processing device 100 .

[0137] In addition, these general or specific methods can be implemented by systems, methods, integrated circuits, computer programs, or non-temporary recording media such as computer-readable CD-ROMs, or by any combination of systems, methods, integrated circuits, computer programs, and recording media.

[0138] (Other implementations, etc.)

[0139] As mentioned above, although embodiment was described, this disclosure is not limited to the said embodiment.

[0140] Therefore, the structural elements recorded in the drawings and detailed descriptions include not only the structural elements necessary for solving the problem, but also structural elements that are not necessary for solving the problem in order to illustrate the above-mentioned technology. Therefore, it should not be directly determined that these non-essential structural elements are essential just because they are recorded in the drawings and detailed descriptions.

[0141] For example, the image processing device 100 may not include the generating unit 130 and the associating unit 180 .

[0142] In addition, for example, the image processing device 100 may further include a learning unit (processing unit) for learning a recognition model. In this case, the above-mentioned training data may also be stored in the storage unit 190 .

[0143] In addition, for example, the image processing device 100 may perform image processing such as the brightness correction described above before dividing the original image 300 .

[0144] In addition, for example, the image processing device 100 may also have a function of determining whether the inspector can properly extract the defect. For example, the image processing device 100 may also determine whether the position of the defect indicated by the position information received by the receiving unit 170 is correct. Specifically, the image processing device 100 may also determine whether the position of the defect indicated by the position information received by the receiving unit 170 is consistent with the position of the defect in the inspection image 320 displayed on the display device 210. For example, the image processing device 100 may also display the determination result on the display device 210 by outputting the correctness determination result to the output unit 150.

[0145] Furthermore, the information indicating the position of the defect in the inspection image displayed on the display device 210 may be stored in advance in the storage unit 190 , or may be generated by the image processing device 100 performing image processing on the inspection image 320 .

[0146] In addition, for example, the image processing device 100 may output a segmented image 310, among the plurality of segmented images 310, for which the determination unit 140 and the learning module 160 cannot determine the pass / fail status of the inspection object 400 in the segmented image 310, to the display device 210 for display. For example, the learning module 160 may be configured to output information indicating that the pass / fail determination cannot be made when the accuracy of the pass / fail determination result is lower than a given value.

[0147] In addition, in the above-mentioned embodiment, the image processing device 100 is implemented as a single device, but it can also be implemented by multiple devices. In the case where the image processing device is implemented by multiple devices, the structural elements of the image processing device described in the above-mentioned embodiment can also be distributed to multiple devices. For example, the image processing device can also be implemented as a client server system. In this case, the client device is a portable terminal that performs image acquisition, user operation reception, image display, etc., and the server device is an information terminal that performs information processing such as generating a segmented image based on the original image.

[0148] In addition, for example, the image processing device 100, the display device 210, and the input device 220 may be arranged in the same place or in different places. For example, the imaging device 200 and the image processing device 100 may be arranged in a factory, and the display device 210 and the input device 220 may be arranged in an office outside the factory or in the home of an inspector. Alternatively, for example, the imaging device 200 may be arranged in a factory, the image processing device 100 may be arranged in an office outside the factory as a server, and the display device 210 and the input device 220 may be arranged in the home of an inspector. In addition, there may be multiple display devices 210, and the first segmented image and the second segmented image may be displayed on independent display devices.

[0149] In the above-mentioned embodiment, the process executed by a specific processing unit may be executed by another processing unit. In addition, the order of a plurality of processes may be changed, or a plurality of processes may be executed in parallel.

[0150] In addition, in the above-mentioned embodiments, each structural element (each processing unit) can also be implemented by executing a software program suitable for each structural element. Each structural element can also be implemented by a program execution unit such as a CPU (Central Processing Unit) or a processor reading and executing a software program recorded in a recording medium such as a hard disk or a semiconductor memory.

[0151] In addition, each structural element may be realized by hardware. Each structural element may also be a circuit (or integrated circuit). Multiple circuits may constitute one circuit as a whole, or they may be independent circuits. In addition, each of these circuits may be a general circuit or a dedicated circuit.

[0152] In addition, the overall or specific aspects of the present disclosure may also be implemented by a system, device, method, integrated circuit, computer program, or non-temporary recording medium such as a computer-readable CD-ROM. In addition, it may also be implemented by any combination of a system, device, method, integrated circuit, computer program, and recording medium.

[0153] For example, the present disclosure may be implemented as an image recognition method executed by a computer such as an image processing device. In addition, the present disclosure may be implemented as a program for causing a computer to execute the image recognition method, or as a computer-readable non-transitory recording medium recording such a program.

[0154] In addition, the present disclosure also includes various modifications that can be conceived by those skilled in the art to the embodiments, or embodiments achieved by arbitrarily combining the structural elements and functions in the embodiments within the scope not departing from the gist of the present disclosure.

[0155] Industrial Applicability

[0156] The present disclosure is useful as an image processing device capable of generating an inspection image for visual inspection by an inspector based on an image generated by a camera or the like.

[0157] -Explanation of symbols-

[0158] 100 Image processing device

[0159] 110 Acquisition Department

[0160] 120 Division

[0161] 130 Generation Department

[0162] 140 Judgment Department

[0163] 150 Output

[0164] 160 learning modules

[0165] 170 Receiving Department

[0166] 180 Establishment of the Related Department

[0167] 190 Storage

[0168] 200 Filming Device

[0169] 210 Display device

[0170] 220 Input device

[0171] 300 Original image

[0172] 310, 311 Segmented Image

[0173] 320 Check Image

[0174] 330 Display image

[0175] 340 Area

[0176] 400 Inspection object

[0177] 500 defects.

Claims

1. An image processing device comprising: an acquisition unit that acquires an original image showing the inspection object; a segmentation unit that segments the original image based on features of the inspection object reflected in the original image, thereby generating a plurality of segmented images; a determination unit that determines, for each of the plurality of segmented images, whether the rule information can be used to determine whether the inspection object is qualified, and determines whether the inspection object is qualified using the rule information for a first segmented image among the plurality of segmented images for which the rule information can be used to determine whether the inspection object is qualified; and The output unit outputs a second segmented image, for which the rule information cannot be used to determine whether the inspection object is acceptable, among the plurality of segmented images, to a learning module capable of machine learning or obtained as a learning result.

2. The image processing device according to claim 1, wherein: The output unit outputs the second divided image to a display device, thereby causing the display device to display the second divided image.

3. The image processing device according to claim 2, wherein: The image processing device further comprises: a receiving unit for receiving a result of determining whether the second divided image is acceptable or not; The output unit outputs the second divided image and the pass / fail determination result of the inspection object in the second divided image received by the receiving unit to the learning module, so that the learning module performs machine learning.

4. The image processing device according to claim 1, wherein: The image processing device further includes a correlating unit that correlates the plurality of segmented images with pass / fail determination results of the inspection object in the plurality of segmented images, and stores the correlation results in a storage unit.

5. The image processing device according to claim 4, wherein: The associating unit stores the same number of divided images of which the result of determination of the pass or fail of the inspection object is passed and the divided images of which the result of determination is failed in the storage unit.

6. The image processing device according to claim 1, wherein: The image processing device includes the learning module obtained as a learning result, The output unit outputs a result of the learning module determining whether the inspection object in the second divided image is pass / fail to a display device.

7. The image processing device according to claim 6, wherein: The image processing device further comprises: an associating unit that associates the unqualified image determined by the determination unit or the learning module as the inspection object being unqualified among the plurality of segmented images with the determination result indicating unqualified, and stores the associated image in the storage unit; and a receiving unit, receiving a result of determining whether the inspection object in the unqualified image is qualified or not, The output unit outputs the unqualified image to the display device, When the pass / fail judgment result of the inspection object in the unqualified image received by the receiving unit is passed, the associating unit changes the judgment result associated with the unqualified image and stored in the storage unit to a judgment result indicating pass.

8. The image processing device according to claim 1, wherein: The image processing device further comprises: a generating unit that classifies the plurality of segmented images based on the features of the inspection object and generates a plurality of inspection images by combining one or more segmented images classified into the same group among the plurality of segmented images; The determination unit determines, for each of the plurality of inspection images, whether or not the inspection object can be determined to be pass / fail using the rule information.

9. The image processing device according to any one of claims 1 to 5 or 8, wherein: The output unit outputs a determination result of the inspection object in the first divided image by the determination unit to a display device.

10. The image processing device according to claim 6 or 7, wherein: The output unit outputs a determination result of the inspection object in the first divided image by the determination unit to the display device.

11. An image processing method, Acquire the original image of the inspection object, The original image is segmented based on the features of the inspection object reflected in the original image, thereby generating a plurality of segmented images, For each of the plurality of segmented images, determining whether the inspection object can be determined to be qualified or not by using the rule information, and determining whether the inspection object is qualified or not by using the rule information for a first segmented image among the plurality of segmented images for which the rule information can be used to determine whether the inspection object is qualified or not, A second segmented image, for which the rule information cannot be used to determine whether the inspection object is acceptable, among the plurality of segmented images is output to a learning module capable of machine learning or obtained as a learning result.

Images