Parts Inspection Equipment

Abstract

To support the effective threshold adjustment work in a component inspection device.SOLUTION: A component inspection device comprises: storage means which stores at least a non-defective product image; generation means which generates a defective product image by using a machine learning model; setting means which allows a user to set a parameter for a component inspection; and output means which outputs an inspection result by performing an inspection on the non-defective product image stored in the storage means and the defective product image generated by the generation means by using the parameter.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to a component inspection device. [Background technology]

[0002] Component inspection devices are used to inspect the quality of components soldered to printed circuit boards. In such component inspection devices, it is common to adjust (tune) the threshold value of the inspection logic in order to reduce false positives (overlooking or missing) (Patent Document 1, etc.).

[0003] The tuning process involves registering images of good and defective products that actually occur in the inspection process as model images, virtually applying the inspection logic to the model images (called model testing), and manually adjusting the threshold so that good and defective images can be detected correctly. Defective product images are essential in the tuning process, and the inspection logic is matured by repeatedly finely adjusting the threshold while accumulating defective product images, which requires a lot of time and skilled techniques.

[0004] However, with recent technological innovation, the performance of surface mount manufacturing equipment has improved dramatically, and the defect rate has been reduced to the level of a few ppm. Therefore, it is not easy to collect images of actual defective products. In particular, when setting up a new manufacturing line, there are almost no images of defective products, making it difficult to adjust the threshold appropriately and increasing the risk of false detection. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2018-77147 A Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to provide a technique for assisting in effectively performing threshold adjustment work in a component inspection device. [Means for solving the problem]

[0007] The present invention enables effective threshold adjustment for component inspection by generating defective product images using a machine learning model and performing a teaching process using the generated defective product images. More specifically, a component inspection device according to one aspect of the present invention includes a storage means for storing at least good product images, a generation means for generating defective product images using a machine learning model, a setting means for allowing a user to set parameters for component inspection, and an output means for inspecting the good product images stored in the storage means and the defective product images generated by the generation means using the parameters and outputting the inspection results.

[0008] In this way, by generating defective product images using machine learning technology, it is possible to increase the number of defective product images available for the teaching process, and it becomes possible to appropriately adjust the threshold value.

[0009] In the present invention, the generating means may generate a defective product image by applying a machine learning model to at least any of the good product images stored in the storage means. For the machine learning model, for example, a generative model based on a generative adversarial network (GAN) can be used.

[0010] In this way, by generating a defective product image based on a good product image, it is possible to generate a defective product image similar to the good product image. In other words, it is possible to generate an image having a defect similar to a defect that may actually occur, so that the teaching process can be performed appropriately.

[0011] In the present invention, the system may further include a designation means for allowing a user to designate a type of defect in a defective image generated by the generation means, and the generation means may generate a defective image having a type of defect designated by the designation means. The designation means may display a graphical user interface (GUI) such as a designation screen for allowing a user to designate a plurality of feature quantities representing defects, and allow the user to designate the type of defect to be generated using the GUI. The designation screen may be configured to allow a user to designate an area in a feature space consisting of a plurality of feature quantities, and further, an image representing a defect corresponding to the feature space may be displayed in at least one of the areas in the feature space. Alternatively, instead of designating an area in the feature quantity space, the system may be configured to allow a user to designate values ​​of a plurality of feature quantities individually.

[0012] According to this configuration, it is possible to generate defective product images having defects that the user considers necessary for the teaching process, thereby enabling threshold adjustment that enables more appropriate inspection.

[0013] In the present invention, it is preferable to further include a suggestion means for suggesting a type of defective product image to be generated. For example, when the number of registered defective product images of a specific type is small, the suggestion means proposes to generate a defective product image of the type. As a specific method, the suggestion means can generate a defective product image for each of a plurality of types of defects by the generation means, perform a clustering process to classify the generated defective product images into a plurality of clusters, determine which cluster each of the defective product images stored in the storage means belongs to, and propose a defect corresponding to a cluster having a predetermined number or less of existing defective product images as a type of defective product image to be generated. In addition, when generating a defective product image corresponding to a certain cluster, it is preferable to add random noise to generate an image having unspecified defects in the cluster.

[0014] According to this configuration, it is possible to prompt the user to generate the missing defective product images, thereby enabling more appropriate threshold adjustment.

[0015] In the present invention, the output means may output, as inspection results, an image of the inspection object, information indicating whether the image of the inspection object was generated by the generation means, and information indicating whether the image of the inspection object is determined to be a good product or a defective product using the parameters.

[0016] Displaying such inspection results allows the user to make appropriate threshold adjustments and generate missing defect images.

[0017] Another aspect of the present invention is a method for assisting in parameter setting for part inspection in a part inspection device, the method including a generation step of generating defective images from stored good product images using a machine learning model, a setting step of accepting parameter settings for part inspection from a user, and an output step of inspecting the stored good product images and the generated defective product images using the parameters, and outputting the inspection results.

[0018] The present invention can be understood as a method including at least a part of the above-mentioned processing, or as a program for implementing the above-mentioned method, or a recording medium on which the program is non-temporarily recorded. Note that the above-mentioned means and processing can be combined with each other as much as possible to constitute the present invention. Effect of the Invention

[0019] According to the present invention, it is possible to assist in effectively performing the threshold adjustment work in a component inspection device. [Brief description of the drawings]

[0020] [Figure 1] FIG. 1 is a diagram showing functional blocks of a component inspection device according to a first embodiment; [Diagram 2] A diagram showing the hardware configuration of a component inspection device. [Diagram 3] A diagram showing the overall flow of the teaching process (threshold adjustment process) performed by the component inspection device. [Figure 4]A diagram showing the flow of the defective product image generation process. [Diagram 5] Example of a screen for outputting model test results and adjusting thresholds used in the teaching process [Figure 6] An example of a designation screen for designating a defective product image to be generated in the defective product image generation process. [Figure 7] An example of a designation screen for designating a defective product image to be generated in the defective product image generation process. [Figure 8] An example of a designation screen for designating a defective product image to be generated in the defective product image generation process. [Figure 9] FIG. 13A is a diagram showing functional blocks of a component inspection device according to a second embodiment, and FIG. 13B is an example of a designation screen for designating a defective product image to be generated. [Figure 10] (A) and (B) are diagrams for explaining a process for proposing defects to be generated in the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] <Application Examples> With reference to FIG. 1, one application example of the present invention will be described. FIG. 1 shows a part inspection device 1 to which the present invention is applied, and is characterized by a support technology for appropriately performing parameter adjustment (tuning work) for part inspection. The storage unit 40 stores image data of good products and image data of defective products that are generated on an actual manufacturing line and are captured by the imaging unit 10. With the recent progress in manufacturing technology, the defect occurrence rate has decreased, and it is not easy to collect necessary defective product images. Therefore, in the present invention, the defective product image generating unit 20 generates a defective product image based on the good product image data 41 using machine learning technology. The user is allowed to specify via the specifying unit 21 what type of defect should be included in the image to be generated. In this way, it is possible to generate a defective product image that is similar to an actual good product image and has a defect of the type specified by the user.

[0022] The user can perform appropriate tuning by adjusting the threshold value using the teaching unit 30 using the defective product images generated in this manner, and actually captured images of good and defective products.

[0023] First Embodiment A component inspection apparatus 1 according to an embodiment of the present invention will be described in detail with reference to Figs. 1 to 5. Fig. 1 is a diagram showing functional blocks of the component inspection apparatus 1. Fig. 2 is a diagram showing a hardware configuration of the component inspection apparatus 1. Fig. 3 is a diagram showing an overall flow of a teaching process (threshold adjustment process) performed by the component inspection apparatus 1. Fig. 4 is a diagram showing a flow of a process for generating a defective product image. Figs. 5 to 8 are examples of display screens used in the teaching process, Fig. 5 is a screen mainly for model test results and threshold adjustment, and Figs. 6 to 8 are designation screens for designating a defective product image to be generated in the defective product image generation process.

[0024] Although the component inspection device 1 also has a function of actually carrying out component inspection based on the teaching results, the following mainly describes the functions of the component inspection device 1 related to the teaching process.

[0025] (composition) As shown in FIG. 1, the part inspection device 1 includes an imaging unit 10, a defective product image generating unit 20, a teaching unit 30, and a storage unit 40.

[0026] The imaging unit 10 captures images of components (e.g., printed circuit boards with components mounted thereon) on a manufacturing line. An image selected by a user from among the captured images is labeled as either a good-quality image or a defective-quality image, and then stored in the storage unit 40 as a model image.

[0027] The defective image generating unit 20 is a functional unit that generates defective images using machine learning technology. The defective image generating unit 20 has a designation unit 21 for receiving designation from a user of the type of defect of the defective image to be generated, and a generation unit 22 for generating the designated defective image.

[0028] The designation unit 21 displays a screen (GUI) on which the user can designate the type of defect to be generated, and receives input from the user.

[0029] The generating unit 22 applies a machine learning model to the good product image stored in the storage unit 40 to generate a defective product image having a defect of the type specified by the specifying unit 21. The machine learning model used by the generating unit 22 is, for example, a generative model based on a generative adversarial network (GAN). In the GAN, a generator that generates an image similar to a correct image and a classifier that distinguishes between a correct image and a generated image are trained. The generator is trained to generate an image closer to the correct image, and the classifier is trained to more accurately distinguish between a correct image and a generated image. The generating unit 22 uses the generator to generate and output a defective product image to which a specified type of defect feature quantity is transferred to an actual good product image. In this way, a defective product image having a specified type of defect for a part similar to the good product image, that is, a defective product image that is likely to occur in reality, can be generated.

[0030] The teaching unit 30 is a functional unit used by the user to set parameters (thresholds) for component inspection. The teaching unit 30 has an inspection unit 31, a result output unit 32, and a threshold setting unit 33. The inspection unit 31 performs an inspection (test) on a model image stored in the storage unit 40 using currently set logic and thresholds. Hereinafter, a test on a model image is also referred to as a model test. The result output unit 32 displays the inspection result by the inspection unit 31 to the user. Although details will be described later, the inspection result includes an image of the component to be inspected, information indicating whether the image of the component to be inspected was generated by the defective product image generating unit 20, and information indicating whether the component is determined to be a good product or a defective product using the current thresholds.

[0031] The storage unit 40 stores good product image data 41, defective product image data 42, and an inspection threshold value 43. The good product image data 41 is image data actually photographed by the imaging unit 10. The defective product image data 42 includes image data actually photographed by the imaging unit 10 and image data generated by the defective product image generating unit 20, and is stored in association with information indicating whether the image was actually photographed or generated. The user specifies whether the image photographed by the imaging unit 10 is to be registered as a model image of a good product image or a defective product image. The inspection threshold value 43 is a threshold value used in the part inspection logic, and can be changed by the user via the teaching unit 30.

[0032] Fig. 2 is a diagram showing a hardware configuration of the part inspection device 1. As shown in Fig. 2, the part inspection device 1 has the same configuration as a general computer (information processing device), and includes a CPU (Central Processing Unit) 51, a ROM (Read Only Memory) 52, a RAM (Random Access Memory) 53, a storage The system includes a storage 54, a keyboard 55, a mouse 56, a monitor 57, and a communication interface 58. Each component is connected to each other via a bus 59 so as to be able to communicate with each other.

[0033] In this embodiment, the ROM 52 or the storage 54 stores a learning program for executing the learning process of the learning model. The CPU 51 is a central processing unit, and executes various programs and controls each configuration. That is, the CPU 51 reads a program from the ROM 52 or the storage 54, and executes the program using the RAM 53 as a working area. The CPU 51 controls each of the above configurations and performs various arithmetic processing according to the program recorded in the ROM 52 or the storage 54. The ROM 52 stores various programs and various data. The RAM 53 temporarily stores programs or data as a working area. The storage 54 is composed of a HDD (Hard Disk Drive), an SSD (Solid State Drive), or a flash memory, and stores various programs including an operating system and various data. The keyboard 55 and the mouse 56 are examples of input devices, and are used to perform various inputs. The monitor 57 is, for example, a liquid crystal display, and displays a user interface. The monitor 57 may adopt a touch panel system and function as an input unit. The communication interface 58 is an interface for communicating with other devices, and uses standards such as Ethernet (registered trademark), FDDI, or Wi-Fi (registered trademark).

[0034] (process) Fig. 3 is a flowchart showing the overall flow of the teaching process performed by the part inspection device 1. Fig. 4 is a flowchart showing the flow of the defective product image generation process. Fig. 5 is an example of a screen used in the teaching process.

[0035] First, the screen of FIG. 5 will be described. The screen 100 includes an enlarged image display area 110, a result output area 120, a part number designation area 130, a logic / threshold setting area 140, a test execution button 150, and a defective product image generation button 160. The part number designation area 130 is a GUI for designating a part (part number) for which a threshold is to be set, and the part is designated by the user checking a check box. The logic / threshold setting area 140 is a GUI for the user to select an inspection item and input a threshold value to be used in the inspection item. The result output area 120 displays the image used in the test, the inspection result with the current settings, and information indicating whether the image is an actual good product image, an actual defective product image, or a generated defective product image. The enlarged image display area 110 displays an enlarged image selected from the multiple images to be inspected, and also displays height profiles in the X-axis and Y-axis directions at the cursor position. By pressing the test execution button 150, inspection is performed in the part number designation area 130 and the logic / threshold setting area 140. By pressing the defective product image generation button 160, a defective product image can be generated.

[0036] Next, the teaching process will be described with reference to Fig. 3. When the teaching process starts, the teaching unit 30 displays a screen 100. Note that the processing order in the flowchart shown in Fig. 3 is merely an example, and it is not necessary to perform the processes according to this order, and the processing order may be changed as appropriate.

[0037] In step S1, teaching unit 30 accepts the designation of a part number to be a target of a model test via part number designation area 130. In the example of FIG.

[0038] In step S2, the teaching unit 30 accepts the designation of an inspection item for the model test via the logic / threshold setting area 140. In the example of FIG. 5, electrode wetness at the end center is selected as the inspection item. When the inspection item is selected, The current threshold values ​​for are displayed in the logic threshold settings area 140 and the values ​​are editable.

[0039] Once the part number and inspection logic are specified, the teaching unit 30 may display a model image of the specified part number and whether each model image is pass or fail for the specified inspection item.

[0040] In step S3, if the user determines that the number of defective product images is insufficient for performing the model test, the user presses the defective product image generation button 160 to generate a defective product image and register it as a model image.

[0041] The defective product image generating process in step S3 will be described in detail with reference to Fig. 4. This process is executed by the defective product image generating unit 20.

[0042] In step S31, the designation unit 21 displays a screen for designating the type of defect to be generated. Figs. 6 to 8 are examples of the screen for designating the type of defect to be generated. Here, three types of defects will be described: too little or too much solder, floating electrodes, and non-wetting of electrodes. These defects are typical examples of defects for which it is difficult to set an appropriate threshold value in teaching, so it is preferable to generate defective product images for such defects, but defective product images may also be generated for other defects.

[0043] Figures 6 and 7 show screens that allow the user to specify the type of defect to be generated by specifying an area in the feature space consisting of three feature quantities that indicate defects. In these examples, the user can specify the type of defect to be generated from an area defined by dividing the amount of solder into three levels, the degree of electrode lift into two levels, and the wetness / non-wetness of the electrode into two levels. Note that the amount of solder and the degree of electrode lift can be divided into more or fewer levels. The wetness / non-wetness of the electrode is basically binary information, but it can be divided into three or more levels. Note that some combinations of features are not possible, so they cannot be selected on the screen.

[0044] In the example of FIG. 6, the electrodes are divided into wet and non-wet, and the types of defects are displayed as areas in a space with the horizontal axis representing the amount of solder and the vertical axis representing the degree of floating of the electrodes. Here, a total of 10 areas, ID11 to ID15 and ID21 to ID25, can be selected. Also, for each area, an image showing an example of a defect corresponding to this feature space is displayed. This allows the user to easily understand what type of defect will be generated. The user selects what type of defect to generate on this specification screen. Here, a frame 601 indicates the defect selected by the user.

[0045] The example in Fig. 7 is a screen for specifying the type of defect by specifying an area in the feature amount space, similar to Fig. 6, but it is a screen for the user to specify an area in a three-dimensional space with the three axes of solder amount, degree of floating, and wetness / non-wetness. Unlike Fig. 6, Fig. 7 does not display images showing examples of defects corresponding to each area, but it may be so arranged that they are displayed.

[0046] Fig. 8 is a diagram showing another example of the specification screen. In this specification screen, the user can directly input values ​​for the amount of solder, the degree of floating of the electrode, and the wetness / non-wetness of the electrode individually. In the example of Fig. 8, a GUI is used in which the amount of solder and the degree of floating are input using scroll bars, but other GUIs such as spin up / down control or inputting numerical values ​​into text boxes may also be used. Also, since the wetness / non-wetness of the electrode is treated as binary information, it is convenient to specify it using check boxes.

[0047] In step S32, the designation unit 21 accepts the type of defect to be generated from the user. Specifically, the designation unit 21 acquires the type of defect that was input when the NG generation button 603 in Figs. 6 to 8 was pressed as the type of defect to be generated.

[0048] In step S33, generation unit 22 acquires an image of the target part number from storage unit 40. The target part number is the part number specified in step S1. Generation unit 22 acquires one or more pieces of data for the target part number from non-defective product image data 41 stored in storage unit 40. If there are many corresponding images in storage unit 40, they may be selected randomly or according to a predetermined rule.

[0049] In step S34, the generating unit 22 generates a defective product image from the image acquired in step S33. As described above, the generating unit 22 has a generator of a generative adversarial network (GAN), and can generate a defective product image that actually occurs on-site by inputting the acquired good product image to the generator and adding the feature amount of the defect specified by the user as noise. The noise to be added is determined, for example, randomly from within the specified feature amount space. Therefore, even if the specified type of defect is the same, fluctuations occur in the defective product image that is actually generated, and it is possible to generate defective product images that have the same type of defect but different details.

[0050] In step S35, the generation unit 22 presents the generated defective product image to the user. Specifically, in the screens of Fig. 6 to Fig. 8, the defective product image is displayed in the three-dimensional viewer units 602a and 602b. In the three-dimensional viewer units 602a and 602b, the user can arbitrarily specify the display position, magnification rate, and viewpoint direction of the generated image to check the generated defects.

[0051] In step S36, it is determined whether the user registers the generated defective product image as a model image. The user checks the defective product image using the three-dimensional viewer units 602a and 602b, and if the user determines that the defective product image is appropriate for use in teaching, he or she presses the "Add to Library" button 604. If the button 604 is pressed, the process proceeds to step S37.

[0052] In step S37, the generated defective product image is stored as a model image in the storage unit 40. At this time, the defective product image is stored in the storage unit 40 so that it can be identified as a defective product image generated by the defective product image generating unit 20.

[0053] In step S38, the user determines whether or not to generate the next defective product image. If the user wishes to continue generating the next defective product image, the process returns to step S32, and if the user wishes to end the generation, the process ends.

[0054] In this way, in the defective product image generation process in step S3, a machine learning-based generator A pseudo defective product image having the defect specified by the user is generated and registered as a model image. Note that the generated image does not necessarily have to be a defective product image, and a non-defective product image can be generated by a similar process, and a non-defective product image having desired characteristics can be generated and registered as a model image.

[0055] Returning to Fig. 3, the explanation will be continued. In step S4, the user presses test execution button 150 to execute an inspection of the model image using the current inspection content and inspection threshold. Specifically, inspection unit 31 executes an inspection of the inspection logic specified in step S2 for each of the model images stored in storage unit 40.

[0056] In step S5, the result output unit 32 outputs the inspection result of step S4 to the result output area 120 of the screen 100. The result output area 120 displays the model image on which the model test was performed. The image, the result of the model test, and information about the model image are displayed. The information about the model image includes information indicating whether the model image is a good product image or a defective product image, and if the model image is a defective product image, information indicating whether the model image is an actually photographed defective product image or a defective product image generated by the defective product image generating unit 20.

[0057] 5, five model images 121a to 121e are displayed in the result output area 120, and their respective inspection results 122a to 122e and information 123a to 123e relating to the model images are displayed. Here, the model images 121a and 121c are actual images of good products, the model images 121b and 121d are actual images of defective products, and the model image 121e is an image of a defective product generated by the defective product image generating unit 20. It can also be seen that while the model images 121a and 121b were appropriately judged as good or defective, the model image 121c was erroneously detected as a good product (overlooking) and the model image 121d was erroneously detected as a defective product (overlooking).

[0058] In step S6, the user judges whether the inspection is being performed correctly, i.e., whether the threshold needs to be adjusted. If the threshold needs to be adjusted, the user changes the threshold value in the logic / threshold setting area 140. The threshold setting unit 33 stores the threshold changed by the user in the storage unit 40. After that, the process returns to step S4 and executes the model test again.

[0059] In the flowchart of FIG. 3, when threshold adjustment is necessary, only the threshold value is changed, but a defective product image generating process in step S3 may be executed to add a defective product image.

[0060] The user repeats the above process to adjust the thresholds so that the part inspection can be performed correctly. The above process can be performed for various part numbers and inspection logics.

[0061] According to this embodiment, pseudo defective product images can be generated using a machine learning model, and can be used in threshold adjustment processing (teaching work). It is difficult to collect defective product images, especially when a new production line is installed, but automatic generation makes it possible to make up for the lack of good product images and adjust the threshold appropriately. In addition, images with specific defects may be required to check whether the threshold adjustment is performed appropriately, and such defective product images can be specified and generated by the user, making it easy to adjust the threshold appropriately from this perspective as well.

[0062] In addition, since the generation of defective product images is performed using a generative model of a generative adversarial network, the generated defective product images are close to actual images of good products, making it possible to generate defective product images that resemble defects that occur in the actual field.

[0063] <Second embodiment> In the first embodiment, the user must specify what kind of defective product image should be generated. The component inspection device according to the present embodiment further has a function of suggesting what kind of defective product image should be generated.

[0064] Fig. 9(A) is a diagram showing functional blocks of a part inspection device 2 according to this embodiment. Compared to the first embodiment (Fig. 1), this embodiment is different in that the defective product image generating unit 20 further includes a proposing unit 23, but is otherwise similar. Therefore, the following description will mainly focus on the proposing unit 23.

[0065] FIG. 9B shows an example of a designation screen on which a user designates the type of defective product image to be generated. In this embodiment, the designation screen includes a button for executing a function of proposing the type of defect to be generated. 10A. When the button 901 is pressed, the proposing unit 23 executes a generated image proposing process shown in FIG.

[0066] In step S101, the proposing unit 23 generates a defective product image for each of the plurality of defect IDs using the generating unit 22. Here, it is assumed that a plurality of defective product images, for example, about 10 defective product images, are generated for each defect.

[0067] In step S102, the proposing unit 23 performs principal component analysis or the like on the generated defective product image to extract features and map them in a feature space. In step S103, the proposing unit 23 performs a clustering process on the defective product image in the feature space to classify the defective product image into a plurality of clusters. The clustering method is not particularly limited, and for example, k-means or the like may be used.

[0068] In step S104, the suggestion unit 23 extracts features from the model image (library image) stored in the storage unit 40, maps them on the feature space in the same manner as in step S102, and determines which cluster the model image belongs to. In step S105, the suggestion unit 23 identifies a cluster in which no library image exists within the region. In step S106, the suggestion unit 23 displays an alert on the GUI for a defect type corresponding to a cluster in which no model image exists within the region, and suggests to the user that a defective product image of this defect type be generated. Note that instead of proposing a defect type corresponding to a cluster in which no model image exists at all, the suggestion unit 23 may suggest a defect type in which the number of model images existing is equal to or less than a predetermined number.

[0069] Fig. 10(B) is a diagram for explaining the above process. The areas in Fig. 10(B) show clusters (areas for each defect) obtained by mapping the defective product images generated in step S101 onto a two-dimensional space and performing clustering processing. Ten clusters (areas) are obtained corresponding to the ten types of defects. The points in Fig. 10(B) are the points where the model images are mapped onto the two-dimensional space. From this result, it can be seen that there is no model image included in the cluster with ID12.

[0070] Therefore, the suggestion unit 23 displays an alert 902 for ID 12 on the defect type designation screen (FIG. 9(B)) to indicate that the defect of ID 12 is missing and therefore an image of this defect should be generated. Note that, when multiple types of defects are missing, an alert may be displayed for each defect.

[0071] According to this embodiment, it is proposed to automatically determine which types of defective product images are missing and generate them, thereby reducing the user's workload and allowing even inexperienced users to appropriately adjust the threshold.

[0072] <Other> The above-described embodiment merely describes an exemplary configuration of the present invention. The present invention is not limited to the above-described specific embodiment, and various modifications are possible within the scope of the technical concept thereof.

[0073] For example, in the above description, a defective product image is generated, but a non-defective product image may be generated in the same manner. Although a large number of non-defective product images are available, there are cases where a non-defective product image required for threshold adjustment cannot be obtained. Therefore, if a non-defective product image having desired characteristics is generated by a method similar to that described above, the threshold adjustment can be performed more appropriately.

[0074] In the above example, the object of inspection is a printed circuit board with mounted components, but what kind of components are being inspected? In addition, the amount of solder, floating of electrodes, and non-wetting of electrodes have been given as examples of types of defects, but other types of defects may also be inspected.

[0075] In the above example, the defective product images are generated by a machine learning model based on a generative adversarial network (GAN), but the defective product images may be generated by using any machine learning model other than GAN, particularly a deep generative model. Examples of the deep generative model include a generative adversarial network (GAN), a variational autoencoder (VAE), and a flow-based generative model.

[0076] <Additional Notes> 1. A storage means (40) for storing at least a non-defective product image (41); A generating means (22) for generating a defective product image using a machine learning model; A setting means (21, 140) for allowing a user to set parameters for part inspection; an output means (32, 120) for performing an inspection on the non-defective product images stored in the storage means and the defective product images generated by the generation means using the parameters, and outputting the inspection results; A part inspection device (1).

[0077] 2. A method for assisting in parameter setting for part inspection in a part inspection device, comprising: A generation step (S3) of generating defective product images using a machine learning model for the stored non-defective product images; a setting step (S7) of accepting settings of parameters for part inspection from a user; an output step (S5) of inspecting the stored non-defective product images and the generated defective product images using the parameters (S4) and outputting the inspection results; and, methods of assistance. [Explanation of symbols]

[0078] 1,2: Parts inspection equipment 10: Imaging unit 20: Defective product image generation section 21: Specification section 22: Generation section 23: Proposal section 30: Teaching section 31: Inspection section 32: Result output section 33: Threshold setting section 40: Storage unit 41: Image data of non-defective products 42: Image data of defective products 43: Inspection threshold

Claims

1. A storage means for storing actually photographed images of good products and actually photographed images of defective products; A generating means for generating a defective product image using a machine learning model; A setting means for allowing a user to set parameters for part inspection; an output means for performing an inspection on the non-defective product images and the defective product images stored in the storage means and the defective product images generated by the generation means, using the parameters, and outputting the inspection results as support information for adjusting the parameters; Equipped with The output means outputs, as the inspection results, an image of the inspection object, information indicating whether the defective image of the inspection object is generated by the generation means or actually photographed, and information indicating whether the inspection object is judged to be a good product or a defective product using the parameters.

2. The generating means applies a machine learning model to at least any of the non-defective product images stored in the storage means to generate a defective product image.

2. The component inspection apparatus according to claim 1.

3. The machine learning model is a deep generative model.

3. The component inspection device according to claim 1 or 2.

4. A designation means for allowing a user to designate a type of defect in the defective product image generated by the generating means is further provided, The generating means generates a defective product image having a defect of the type designated by the designating means. The component inspection device according to any one of claims 1 to 3.

5. the designation means displays a designation screen on which a user can designate a plurality of feature quantities representing defects; 5. The component inspection apparatus according to claim 4.

6. the designation screen is configured to allow a user to designate an area in a feature space consisting of a plurality of feature quantities; 6. The component inspection apparatus according to claim 5.

7. An image representing a defect corresponding to the feature space is displayed in at least one of the regions in the feature space.

7. The component inspection apparatus according to claim 6.

8. the designation screen is configured to allow a user to designate values ​​of a plurality of feature quantities individually; 6. The component inspection apparatus according to claim 5.

9. A suggestion means for suggesting a type of defective product image to be generated is further provided. The component inspection device according to any one of claims 1 to 8.

10. The suggestion means includes: generating a defective product image for each of a plurality of types of defects by the generating means; A clustering process is performed to classify the generated defective product images into multiple clusters. determining to which cluster each of the defective product images stored in the storage means belongs; A defect corresponding to a cluster having a predetermined number or less of existing defective product images is proposed as a type of defective product image to be generated.

10. The component inspection apparatus according to claim 9.

11. A storage means for storing at least a non-defective product image; A designation means for allowing a user to designate a type of defect in a defective product image to be generated; A generating means for generating a defective product image having a defect type specified by the specifying means by using a machine learning model; A setting means for allowing a user to set parameters for part inspection; an output means for performing an inspection on the non-defective product images stored in the storage means and the defective product images generated by the generation means using the parameters, and outputting the inspection results as support information for adjusting the parameters; A suggestion means for suggesting a type of defective product image to be generated; Equipped with The suggestion means includes: A clustering process is performed to classify defective product images into a plurality of clusters for each of a plurality of types of defects that can be designated by the designation means; determining to which cluster each of the defective product images stored in the storage means belongs; A type of defect corresponding to a cluster having a predetermined number or less of existing defective product images is proposed as a type of defective product image to be generated. Parts inspection equipment.

12. 1. A method for supporting parameter setting for part inspection in a part inspection device, comprising: A generation step of generating a defective product image from the stored non-defective product image by using a machine learning model; a setting step of accepting settings of parameters for part inspection from a user; an output step of inspecting the stored actually photographed images of non-defective products, the actually photographed images of defective products, and the generated images of defective products using the parameters, and outputting the inspection results as support information for adjusting the parameters; Including, In the output step, an image of the inspection target, information indicating whether the defective product image of the inspection target is generated in the generation step or actually photographed, and information indicating whether the product is determined to be a non-defective product or a defective product using the parameters are output as the inspection result. How to help.

13. A method for assisting parameter setting for part inspection in a part inspection device, comprising: A step of receiving, from a user, a designation of a type of defect of a defective product image to be generated; A generating step of generating a defective product image having a specified defect by using a machine learning model on the stored non-defective product image; a setting step of accepting settings of parameters for part inspection from a user; an output step of inspecting the stored non-defective product images and the generated defective product images using the parameters, and outputting the inspection results as support information for adjusting the parameters; A suggestion step of proposing a type of defective image to be generated; Including, In the proposing step, A clustering process is performed to classify the generated defective product images into a plurality of clusters for each of a plurality of specifiable types of defects; determining to which cluster each of the stored defective product images belongs; A type of defect corresponding to a cluster having a predetermined number or less of existing defective product images is proposed as a type of defective product image to be generated. How to help.

14. A computer program product for causing a computer to carry out the steps of the method according to claim 12 or 13.

Images