Inspection support system, inspection support method, and program

Abstract

The purpose of the present disclosure is to inhibit the need for a complex design. An inspection assistance system (1) comprises an image acquisition unit (11) and an image generation unit (12). The image acquisition unit (11) acquires a standard image (A1) relating to a target (T1). The standard image (A1) is an image in which a condition parameter (P1) of a process condition relating to the state of a surface of the target (T1) is set to a standard value. The image generation unit (12) uses the standard value as a reference to generate a plurality of evaluation images (B1) which relate to the target (T1) and in which the condition parameter (P1) has been changed on the basis of the standard image (A1) and a prescribed image generation model.

Description

[Technical field]

[0001] The present disclosure generally relates to an inspection support system, an inspection support method, and a program, and more particularly, to an inspection support system, an inspection support method, and a program for an inspection of a surface condition of an object. [Background technology]

[0002] Patent Document 1 discloses an inspection standard determination device. This inspection standard determination device determines inspection standards relating to appearance features of a defect candidate area of ​​a sample, such as the size of a scratch or crack, color difference, etc., based on a psychometric curve, in order to determine whether the area is defective or not based on the feature values. In the inspection standard determination device, an image presenting means presents a standard sample and one of the target samples to an inspector, and the inspector is asked to compare the appearance features of the defect candidate areas of the two samples with each other and to answer which is larger. The inspector's answer is acquired by an input means. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2007-333709 A Summary of the Invention

[0004] The inspection standard determination device of Patent Document 1 requires design of appearance features in advance. However, in the inspection of the surface condition of an object, the appearance features can be enormous. For example, in the case of a surface coating, the features can be multidimensional if they include color (brightness, saturation, hue), gradation, graininess, glitteriness, glossiness, matteness, and the like. Therefore, it is difficult to express the texture felt by humans with a simple feature, and even if it could be expressed with a feature, the number of trials of subjective evaluation would be enormous.

[0005] The present disclosure has been made in consideration of the above-mentioned reasons, and aims to provide an inspection support system, an inspection support method, and a program that do not require complex design.

[0006] Books An inspection support system according to another aspect of the disclosure includes an image acquisition unit and an image generation unit. The image acquisition unit acquires a standard image of an object, the standard image having a condition parameter in a process condition related to a surface state of the object set to a standard value. The image generation unit generates a plurality of evaluation images of the object by varying the condition parameter based on a predetermined image generation model and the standard image with the standard value as a reference. The condition parameter is set so that the plurality of evaluation images become evaluation targets for a subjective evaluation of appearance features of the surface of the object.

[0007] Books An inspection support method according to another aspect of the disclosure includes an image acquisition process and an image generation process. In the image acquisition process, a standard image of an object is acquired, in which a condition parameter in a process condition related to a surface state of the object is set to a standard value. In the image generation process, a plurality of evaluation images of the object are generated by changing the condition parameter based on a predetermined image generation model and the standard image, with the standard value as a reference. The condition parameter is set so that the plurality of evaluation images become evaluation targets for a subjective evaluation of the appearance features of the surface of the object.

[0008] A program according to one embodiment of the present disclosure includes the steps of: Inspection It is a program for executing the inspection support method. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 is a schematic block diagram of an examination support system according to an embodiment. [Diagram 2] FIG. 2 is a conceptual diagram of the entire system including the above-mentioned inspection support system. [Diagram 3]Fig. 3A is a graph for explaining a standard image and an evaluation image in the above-mentioned examination support system, and Fig. 3B is a conceptual diagram relating to display and evaluation of the above-mentioned standard image and evaluation image. [Figure 4] FIG. 4 is a graph for explaining the determination of the inspection standard in the inspection support system. [Diagram 5] FIG. 5 is a flowchart for explaining an operation example 1 in the above-mentioned examination support system. [Figure 6] FIG. 6 is a flowchart for explaining an operation example 2 in the above-mentioned examination support system. [Figure 7] FIG. 7 is a graph for explaining a first modified example of the above test support system. [Figure 8] FIG. 8 is a conceptual diagram for explaining a second modified example of the above test support system. [Figure 9] FIG. 9 is a graph for explaining a third modified example of the above test support system. [Figure 10] Fig. 10A is a graph for explaining a standard image and an evaluation image in another modified example 1 of the above examination support system. Fig. 10B is a conceptual diagram regarding display and evaluation of the standard image and the evaluation image in another modified example 1 of the above examination support system. [Figure 11] FIG. 11 is a conceptual diagram for explaining another modified example 2 (variation of display mode) of the above-mentioned examination support system. [Figure 12] FIG. 12 is a conceptual diagram for explaining another modified example 3 of the above inspection support system (screen display of the inspection standard process). DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] (1) Overview Each drawing described in the following embodiments is a schematic drawing, and the size and thickness ratios of each component in each drawing do not necessarily reflect the actual dimensional ratios.

[0011] The examination support system 1 according to this embodiment includes an image acquisition unit 11 and an image generation unit 12, as shown in FIG.

[0012] The image acquisition unit 11 acquires a standard image A1 (see FIGS. 3A and 3B) of the target T1 (see FIG. 2), in which a condition parameter P1 (see FIG. 3A) in the process conditions related to the surface condition of the target T1 is set to a standard value. In the present embodiment, as an example, the target T1 is assumed to be an automobile part. However, the target T1 is not limited to an automobile part, and may be any object having a surface. In addition, the "surface condition" of the target T1 here is assumed to be a condition related to the painting of the surface, as an example. Therefore, the process conditions are painting conditions. The condition parameter P1 relates to at least one of the amount of paint discharged, the pressure at which the paint is atomized, the painting distance to the surface of the target T1, the number of coats, and the drying speed of the paint. However, the "surface condition" may be, other than painting, for example, a condition related to plating or a condition related to decorative molding.

[0013] In Fig. 2, the painted state of a part of the outer surface of the automobile part (target T1) is merely conceptually represented by a circular picture with dot hatching, and the circular picture is not intended to represent that the target T1 is spherical. Similarly, in Fig. 3A and Fig. 3B, etc., an image of the painted state of the target T1 is conceptually represented by a circular picture with dot hatching.

[0014] The standard image A1 is, for example, a captured image of the target T1 captured by the imaging unit 2 (see FIG. 2). In other words, the standard image A1 is a captured image of the actual target T1 (sample product) that has been actually painted by the painting system 300 (see FIG. 2) with the condition parameter P1 set to a standard value. However, the standard image A1 is not limited to a captured image of the actual product, and may be a pseudo image (CG image) with the condition parameter P1 set to a standard value. The standard image A1 is, for example, a single still image, but is not particularly limited thereto and may be a moving image.

[0015] The image generating unit 12 generates a plurality of evaluation images B1 (see FIG. 3) of the target T1 by varying the condition parameter P1 based on a predetermined image generation model M1 and standard image A1 with a standard value as a reference. Here, the image generation model M1 is, for example, a function model with the condition parameter P1 as a variable. The evaluation image B1 is an image generated by adjusting the color density (pixel value) of three colors in the color space RGB based on the standard image A1. However, the color space is not limited to RGB, and may be, for example, XYZ or Lab (Lab color space).

[0016] According to this configuration, a plurality of evaluation images B1 are generated using the condition parameters P1 in the process conditions. Therefore, by using the evaluation images B1, the inspection standard can be determined more easily than when the inspection standard is determined by designing complex appearance features, for example. As a result, the inspection support system 1 has the advantage that complex design is unlikely to be required.

[0017] The inspection support method according to the present embodiment includes an image acquisition process (image acquisition step) and an image generation process (image generation step). In the image acquisition process (image acquisition step), a standard image A1 of the object T1 is acquired, in which a condition parameter P1 in a process condition related to the surface state of the object T1 is set to a standard value. In the image generation process (image generation step), a plurality of evaluation images B1 of the object T1 are generated by changing the condition parameter P1 based on a predetermined image generation model M1 and the standard image A1 with the standard value as a reference. According to this configuration, it is possible to provide an inspection support method that does not require a complex design. This inspection support method is used on a computer system (inspection support system 1). In other words, this inspection support method can also be embodied as a program. The program according to the present embodiment is a program for causing one or more processors to execute the inspection support method according to the present embodiment.

[0018] (2) Details Hereinafter, an entire system (painting control system 100) including the inspection support system 1 and its peripheral configuration according to this embodiment will be described in detail with reference to Figures 1 and 2. Note that at least a part of the peripheral configuration may be included in the configuration of the inspection support system 1.

[0019] (2.1) Overall Structure As shown in FIG. 2, the paint management system 100 includes an inspection support system 1, a paint system 300, and an imaging unit 2 (imaging system).

[0020] The inspection support system 1 has a function of supporting the creation of so-called "limit samples" that indicate the limit of whether a painted product is "good (OK)" or "bad (NG)" in terms of quality as an inspection standard. Specifically, for example, an automobile parts manufacturer may create a limit sample (sample product) for a painted product of a certain part, and share the information with a customer such as an automobile manufacturer to obtain approval for the manufacture of the painted product. Furthermore, an inspector may check the limit sample when inspecting a painted product that has completed the painting process during operation of an actual production line, thereby enabling the inspection work to be performed stably and accurately. However, the creation of limit samples may require a huge amount of cost and time. The inspection support system 1 is configured to support the creation of limit samples.

[0021] For ease of explanation, below, the work involved in creating limit samples is referred to as "sample creation work," and the person who performs this is referred to as "creator H1" (see Figures 2 and 3B). Additionally, the work of inspecting painted products on an actual production line is referred to as "inspection work," and the person who performs this is referred to as "inspector." For explanation purposes, when there is no need to distinguish between creator H1 and inspector, they may simply be referred to as "user." Note that creator H1 and inspector may be the same person.

[0022] In this embodiment, the test support system 1 includes a processing unit 10, an operation unit 3, a display unit 4, a first storage unit 5, a second storage unit 6, a learning unit 7, and a pass / fail judgment unit 8 (inference unit) as shown in FIG. 1. The main functions of the test support system 1 (functions of the processing unit 10, the first storage unit 5, the second storage unit 6, the learning unit 7, and the pass / fail judgment unit 8, etc.) are assumed to be provided in a server 200 (see FIG. 2) as an example. The "server" here is assumed to be composed of one server device. In other words, it is assumed that the main functions of the test support system 1 are provided in one server device. However, the "server" may be composed of multiple server devices. Specifically, the functions of the processing unit 10, the first storage unit 5, the second storage unit 6, the learning unit 7, and the pass / fail judgment unit 8 may be provided in individual server devices, or two or more of these functions may be provided in one server device. In addition, such a server device may build, for example, a cloud (cloud computing). Some functions of the inspection support system 1 may be provided in a distributed manner in a personal computer, a notebook computer, a tablet terminal, or the like, other than the server. The server device may be installed in a factory where at least one of the painting process and the painting inspection is performed, or may be installed outside the factory (for example, at the business headquarters). When multiple functions of the inspection support system 1 are provided in separate server devices, it is desirable that each server device is connected to the other server devices so as to be able to communicate with each other.

[0023] The imaging unit 2 (imaging system) is a system for generating an image (digital image) of the surface of the target T1. In this embodiment, the imaging unit 2 captures the surface of the target T1 illuminated by, for example, a lighting device to generate an image of the surface of the target T1. The imaging unit 2 includes, for example, one or more RGB cameras. The camera includes one or more image sensors. The camera may include one or more line sensors. As shown in FIG. 2, the imaging unit 2 is connected to a network NT1 and can communicate with the server 200 via the network NT1. The network NT1 is not particularly limited and may be constructed by wired communication via a communication line or wireless communication. The wired communication is, for example, wired communication via a twisted pair cable, a dedicated communication line, or a LAN (Local Area Network) cable. The wireless communication is, for example, wireless communication conforming to standards such as Wi-Fi (registered trademark), Bluetooth (registered trademark), ZigBee (registered trademark), or low-power wireless (specified low-power wireless) that does not require a license, or wireless communication such as infrared communication.

[0024] In this embodiment, as an example, the same imaging unit 2 is used for both the "sample making work" and the "inspection work", but a different imaging unit may be used for each work.

[0025] The painting system 300 is a system for painting the surface of the target T1. That is, the painting system 300 executes a painting process for the target T1. The painting system 300 includes one or more painting units (painting robots). The painting robots may be of a conventionally known configuration, and therefore detailed description thereof will be omitted. As shown in FIG. 2, the painting system 300 is connected to a network NT1 and is capable of communicating with the server 200 via the network NT1. The communication between the painting system 300 and the server 200 is not particularly limited, and may be performed by wired communication via a communication line, or may be performed by wireless communication, similar to the communication between the imaging unit 2 and the server 200.

[0026] In this embodiment, as an example, the same painting system 300 is used for both the "sample creation work" and the "inspection work", but different painting systems may be used for each work.

[0027] (2.2) Configuration of the Inspection Support System The configuration of the test support system 1 (the processing unit 10, the operation unit 3, the display unit 4, the first storage unit 5, the second storage unit 6, the learning unit 7, and the pass / fail determination unit 8) will be described in more detail below.

[0028] The processing unit 10 can be realized by a computer system including one or more processors (microprocessors) and one or more memories. In other words, the one or more processors execute one or more programs (applications) stored in one or more memories to function as the processing unit 10. Here, the programs are pre-recorded in the memory of the processing unit 10, but they may be provided via a telecommunications line such as the Internet or recorded on a non-transitory recording medium such as a memory card.

[0029] The processing unit 10 executes processing related to the imaging unit 2 and the painting system 300. It is assumed that the functions of the processing unit 10 are in the server 200. The processing unit 10 also has an image acquisition unit 11, an image generation unit 12, an evaluation acquisition unit 13, a standard determination unit 14, an output unit 15, and a condition determination unit 16, as shown in Fig. 1. That is, the processing unit 10 has a function as the image acquisition unit 11, a function as the image generation unit 12, a function as the evaluation acquisition unit 13, a function as the standard determination unit 14, a function as the output unit 15, and a function as the condition determination unit 16.

[0030] The image acquisition unit 11 is configured to acquire a standard image A1 of the target T1. The standard image A1 is an image of the target T1 captured by the imaging unit 2. In other words, the standard image A1 is an image of the actual target T1 that has been actually painted by the painting system 300 as a preliminary preparation for a sample creation work, with the condition parameter P1 set to a standard value. In the sample creation work, the inspection support system 1 receives information on the standard image A1 from the imaging unit 2.

[0031] Here, the condition parameter P1 relates to the coating conditions of the target T1, including at least one of the paint discharge amount, the pressure at which the paint is atomized (atomization pressure), the coating distance to the surface of the target T1, the number of coats, and the drying speed of the paint.

[0032] For example, in a painting process, the painting system 300 applies multiple layers of paint to an object T1 (e.g., the surface of a vehicle body) while changing at least one of the color and type of paint and adjusting the discharge amount, atomization pressure, painting distance, number of coats, etc. The discharge amount is, for example, the amount (L / min) discharged from the spray gun at the tip of the painting robot. The atomization pressure is the pressure of the paint atomized (atomized) by air pressurized by an air compressor and sent to the spray gun. The painting distance is, for example, the distance between the spray gun and the object T1.

[0033] The above-mentioned layers may include, for example, an anti-rust electrocoating (first layer), a first base coat (second layer), a second base coat (third layer), and a clear coat (fourth layer) in order from the surface side of the target T1. The thickness of the coating is controlled by adjusting the discharge amount, coating distance, and number of coatings. If the second or third layer is thin, the base is easily visible, and if the fourth layer is thin, the gloss increases. By providing the second to fourth layers with controlled coating thickness in this way, the color development, aesthetics, gloss, etc. are improved. The atomization pressure affects the granulation (graininess) of the paint. The granulation (graininess) of the paint affects unique surface finishes such as roughness. The drying speed affects the uniformity of the orientation of aluminum flakes, for example, to better express a metallic feel.

[0034] In the following, as an example, we focus on the paint discharge amount (L / min) and assume that the condition parameter P1 (painting parameter) is a parameter related to the discharge amount. If there are 10 setting levels (levels "1" to "10") of the degree of discharge amount that can be set for a certain layer (e.g., the third layer) among the above-mentioned multiple layers, the standard value is the value of the condition parameter P1 that corresponds to the discharge amount of the middle standard level "5". Hereinafter, the value of the condition parameter P1 that corresponds to the discharge amount of the standard level is called the first standard value P11 (see FIG. 3A). Note that FIG. 3A is a graph with the horizontal axis representing the condition parameter P1 (painting parameter p) and the vertical axis representing image data I (e.g., color density (pixel value) related to RGB).

[0035] The image generating unit 12 is configured to generate a plurality of (five in FIG. 3A) evaluation images B1. The five evaluation images B1 are images obtained by varying the condition parameter P1 at regular intervals based on the image generation model M1 and the standard image A1 with a standard value as a reference. Here, the image acquiring unit 11 further acquires the standard image A1 in which the condition parameter P1 is set to a second standard value P12 different from the first standard value P11 as a standard value. The image generating unit 12 generates a plurality of evaluation images B1 in which the condition parameter P1 is varied between the first standard value P11 and the second standard value P12. As an example in FIG. 3A, the second standard value P12 is a value of the condition parameter P1 corresponding to the discharge amount of the highest level "10" among, for example, 10 levels of the degree of the discharge amount that can be set. The degree of the second standard value P12 is not particularly limited as long as it is different from the first standard value P11, but it is preferable that the second standard value P12 is a value extremely different from the first standard value P11. Particularly in this embodiment, the purpose is to determine the limit of "OK" or "NG" as an inspection standard. Therefore, it is preferable that the first standard value P11 is a value that can be judged sufficiently as "OK" even by visual inspection, and the second standard value P12 is a value that can be judged sufficiently as "NG" even by visual inspection.

[0036] In Fig. 3A, image data I1 located at the origin is image data (hereinafter also referred to as first standard image A11) of the actual object (sample product) of the target T1 actually painted by the painting system 300 using the first standard value P11 (discharge amount) of the first painting condition. Also in Fig. 3A, image data I2 is image data (hereinafter also referred to as second standard image A12) of the actual object (sample product) of the target T1 actually painted by the painting system 300 using the second standard value P12 (discharge amount) of the second painting condition.

[0037] The image generation model M1 is a function model with a condition parameter P1 as a variable. Here, if the coating parameter, which is the condition parameter P1, is p (variable), the image data I (color density) of the evaluation image B1 is determined by a function f(p) (image generation model M1). In other words, the function f(p) (approximately) defines the change characteristics of the color density for RGB with respect to the discharge amount (condition parameter P1) for a certain layer (e.g., the third layer), and is obtained by measurement verification, simulation results, etc. Information related to the image generation model M1 is stored (stored) in advance in the first storage unit 5.

[0038] For ease of understanding, it is assumed that the coating conditions focus on the discharge amount (condition parameter P1) for the third layer and only change it, while the condition parameters P1 for the other layers, such as the discharge amount, number of coats, and atomization pressure, are fixed to standard values ​​to generate the evaluation image B1. However, the evaluation image B1 may be generated by simultaneously changing two or more condition parameters P1. For example, when changing the discharge amount and the number of coats simultaneously, the image data I (color density) of the evaluation image B1 can be prepared by preparing a function f(p) that defines the change characteristics of the color density with respect to the discharge amount and number of coats.

[0039] The first storage unit 5 and the second storage unit 6 each include a rewritable non-volatile memory such as an Electrically Erasable Programmable Read-Only Memory (EEPROM).

[0040] The first storage unit 5 stores information on various coating conditions. The first storage unit 5 also stores a plurality of types of image generation models M1. That is, the first storage unit 5 stores not only the function f(p) of the discharge amount for the third layer, but also functions f(p) of the discharge amount for the first layer, second layer, and fourth layer, and further stores a large number of functions f(p) for the atomization pressure, coating distance, number of coats, and drying speed in addition to the discharge amount. The second storage unit 6 stores (stores) a trained model M2, which will be described later. In this embodiment, the first storage unit 5 and the second storage unit 6 are separate storage units, but may be one common storage unit. At least one of the first storage unit 5 and the second storage unit 6 may be a memory of the processing unit 10.

[0041] Furthermore, the image data I (color density) of the evaluation image B1 may be simply calculated, for example, by the following formula (1) (image generation model M1).

[0042] I = ΔI×α+I1 Equation (1) In formula (1), ΔI is the difference obtained by subtracting the image data I1 (color density) of the first standard image A11 from the image data I2 (color density) of the second standard image A12. α is a normalized value of the coating parameter p, which can take a value between 0 and 1. In other words, the coating parameter p's discharge amount, number of coats, atomization pressure, etc. have different unit amounts, so α is a scaling of the values ​​that can be taken by the first standard image A11 to the second standard image A12 for the discharge amount, number of coats, atomization pressure, etc., from 0 to 1.

[0043] When formula (1) is used, the image generating unit 12 identifies the color densities (pixel values) of RGB for each of the first standard image A11 and the second standard image A12, and calculates the difference ΔI. The image generating unit 12 generates multiple evaluation images B1 for the target T1 by changing α, multiplying α by the difference ΔI each time, and adding the result to the image data I1 (color density) of the first standard image A11 that serves as the base. FIG. 3A shows five evaluation images B1 simply generated by the above formula (1). Therefore, the color densities of the five evaluation images B1 gradually increase linearly (proportionally) in response to an increase in the painting parameter p.

[0044] The evaluation acquisition unit 13 is configured to acquire evaluation information regarding two or more stages of evaluation results for a plurality (here, five) of evaluation images B1. Here, as an example, it is assumed that the evaluation results are in two stages, "good (OK)" and "poor (NG)". The evaluation results are subjective evaluations by the creator H1. Specifically, as shown in FIG. 3B, the creator H1 visually evaluates each evaluation image B1, and uses the operation unit 3 to input the evaluation results (OK or NG) for each evaluation image B1.

[0045] The display unit 4 is a liquid crystal display or an organic EL (Electro-Luminescence) display. The display unit 4 may be a touch panel display. The display unit 4 may be attached to an information terminal 9 (see FIG. 3B) such as a personal computer used by a user. The information terminal 9 may be a notebook computer or a tablet terminal. The display unit 4 displays a standard image A1 and a plurality of evaluation images B1.

[0046] The server 200 and the information terminal 9 can communicate with each other via the network NT1. During sample creation work, the information terminal 9 receives information on the standard image A1 and the evaluation image B1 from the server 200 and displays them on the screen of the display unit 4. The creator H1 can visually confirm the standard image A1 and the evaluation image B1 through the display unit 4. In particular, in this embodiment, as shown in FIG. 3B, the display unit 4 simultaneously displays the standard image A1 and each evaluation image B1 on one screen so that they can be easily compared visually. The display unit 4 displays various information other than the standard image A1 and the evaluation image B1.

[0047] The operation unit 3 includes a mouse, a keyboard, a pointing device, and the like. The operation unit 3 is provided, for example, in an information terminal 9 used by a user. If the display unit 4 is a touch panel type display, it may also function as the operation unit 3. The creator H1 visually compares the standard image A1 and each evaluation image B1 displayed on the display unit 4, evaluates each evaluation image B1 as "good (OK)" or "no good (NG)," and inputs the evaluation results to the inspection support system 1 via the operation unit 3. The processing unit 10 associates the input evaluation results with the evaluation images B1, and stores them in a memory unit (for example, the first memory unit 5).

[0048] The standard determination unit 14 is configured to determine an inspection standard for the surface condition of the target T1 based on the evaluation information. 3A and 4 show the evaluation result in which creator H1 has evaluated evaluation images B11 to B13 of five evaluation images B1 (B11 to B15) as "OK" and evaluated evaluation images B14 and B15 as "NG". For convenience, in the drawings, a circle mark is marked near evaluation images B11 to B13 corresponding to "OK" and a cross mark is marked near evaluation images B14 and B15 corresponding to "NG".

[0049] The standard determination unit 14 identifies the boundary where the evaluation result changes from "OK" to "NG" among the multiple evaluation images B1 arranged in a straight line. The standard determination unit 14 sets the evaluation image B1 (evaluation image B13 in FIG. 4) corresponding to the evaluation result "OK" that is closest to "NG" as the inspection standard. The processing unit 10 stores information on the image data I3 (color density) and the inspection standard value P13 (third painting condition) of the evaluation image B13 that is the inspection standard in a storage unit (e.g., the first storage unit 5).

[0050] The output unit 15 is configured to output information about a condition parameter P1 corresponding to the inspection standard (in FIG. 4, an inspection standard value P13). That is, the output unit 15 outputs information about the third painting condition to the outside (for example, the information terminal 9). In other words, the server 200 transmits information about the third painting condition to the information terminal 9. The information terminal 9 presents information about the third painting condition from the display unit 4. The information about the third painting condition includes not only the inspection standard value P13 related to the discharge amount of a certain layer that was subject to change, but also condition parameters P1 such as the discharge amount of other layers that were fixed to standard values, the number of coats, and the atomization pressure.

[0051] The creator H1 creates the target T1, which is a limit sample, based on the presented third painting conditions. That is, the creator H1 inputs information about the presented third painting conditions to the painting system 300 via a user interface. As a result, the painting system 300 paints according to the third painting conditions, and the target T1 (limit sample) is created. The limit sample created in this way has a painted state that is very close to the evaluation image B13 generated by the image generation unit 12.

[0052] The destination of the information on the third painting condition by the output unit 15 may be directly to the painting system 300, not to the information terminal 9. The painting system 300 may paint according to the third painting condition received directly from the server 200, and create a target T1 (limit sample).

[0053] By feeding back the information on the third painting condition thus output to the painting process, the creator H1 can create and check an actual product (limit sample) that meets the inspection standard.

[0054] In this embodiment, the first storage unit 5 stores a plurality of candidate models N1 (see FIG. 1) corresponding to a plurality of standard values ​​in the condition parameter P1, respectively. The plurality of candidate models N1 here may include, for example, a plurality of image generation models M1 that have been applied to determining the inspection standard in the past.

[0055] The condition determination unit 16 determines the similarity between the standard value and each of the multiple standard values, and if there is a value that is highly similar to the standard value among the multiple standard values, selects the candidate model N1 corresponding to that value from among the multiple candidate models N1 as a predetermined image generation model M1. Specifically, the condition determination unit 16 compares the absolute value (|pp'|) of the difference between the first standard value P11 (painting parameter p) for the discharge amount of interest and each of the multiple first standard values ​​P11 (painting parameter p') corresponding to the multiple candidate models N1 with a threshold value. If there is a candidate model N1 in the first storage unit 5 whose absolute value (|pp'|) is smaller than the threshold value, the condition determination unit 16 selects that candidate model N1 as the image generation model M1.

[0056] It is preferable that the condition determination process by the condition determination unit 16 is executed as a preparation for the sample creation work. If the first storage unit 5 does not contain a candidate model N1 whose absolute value (|pp'|) is smaller than the threshold, the server 200 notifies the information terminal 9 of that effect. The creator H1 prepares a new image generation model M1 and registers the information about it in the inspection support system 1 via the operation unit 3.

[0057] In this way, the condition determination unit 16 determines the similarity and selects the image generation model M1, which saves the creator H1 the trouble of creating and selecting a new image generation model M1. As a result, the inspection criteria can be determined more efficiently.

[0058] The learning unit 7 generates a trained model M2 (see FIG. 1) by using image data to which a pass / fail label is added regarding the surface condition (here, the coating condition) as training data. This "pass / fail label" is a label based on the inspection standard determined by the standard determination unit 14.

[0059] "Learning data" is learning data used for machine learning of a model. This "model" is a program that, when input data on an object to be identified (surface state of object T1) is input, estimates the state of the object to be identified and outputs an estimation result (identification result). "Learning model" refers to a model for which machine learning using learning data has been completed. "Learning data (set)" is a data set that combines input data (image data) input to the model and labels assigned to the input data, and is so-called teacher data. That is, in this embodiment, the learning model M2 is a model for which machine learning using supervised learning has been completed.

[0060] The learning unit 7 has a function of generating a trained model M2 for the target T1. The learning unit 7 generates the trained model M2 based on a plurality of labeled training data (image data). It is assumed that the trained model M2 includes, for example, a model using a neural network, or a model generated by deep learning using a multilayer neural network. The neural network may include, for example, a convolutional neural network (CNN) or a Bayesian neural network (BNN). The trained model M2 is realized by implementing a trained neural network in an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). The trained model M2 is not limited to a model generated by deep learning. The trained model M2 may be a model generated by a support vector machine, a decision tree, or the like.

[0061] The multiple learning data are generated by labeling the multiple evaluation images B1 generated under various painting conditions with the evaluation results "OK" or "NG" based on the inspection standard determined by the standard determination unit 14. That is, in the example of FIG. 4, learning data is generated in which the image data of the evaluation images B11 to B13 are labeled with "OK". Also, learning data is generated in which the image data of the evaluation images B14 and B15 are labeled with "NG". It is also possible to generate learning data in which the image data of the first standard image A11 is labeled with "OK". Also, learning data is generated in which the image data of the second standard image A12 is labeled with "NG".

[0062] In other words, when multiple evaluation images B1 are used as learning data, the work of labeling (labeling) is automatically completed at the time when the inspection standard is determined. This reduces the time and effort required for the user to prepare new learning data and label the data for the inspection support system 1 via a user interface such as the operation unit 3. The learning unit 7 uses multiple labeled learning data to perform machine learning to determine whether the coating condition of the target T1 is good or bad, thereby generating a trained model M2. The trained model M2 generated by the learning unit 7 is stored (memorized) in the second storage unit 6.

[0063] The learning unit 7 can improve the performance of the trained model M2 by re-learning using the newly acquired labeled learning data (evaluation image B1). For example, if an evaluation image B1 is generated under new painting conditions, it is possible to cause the learning unit 7 to re-learn about the new evaluation image B1.

[0064] The pass / fail judgment unit 8 judges pass / fail of the inspection image C1 of the object T1 using the trained model M2. That is, the inspection support system 1 has a function of automatically discriminating pass / fail of the paint state of the object T1 that has undergone a painting process in an actual production line, for example. The imaging unit 2 sequentially captures images of the object T1 that has undergone a painting process in the inspection process, and transmits the captured images (inspection images C1) to the server 200. The processing unit 10 transmits the discrimination result of the object T1 to a terminal (for example, an information terminal 9) used by an inspector. Here, if the discrimination result is "NG (fail)", the server 200 notifies the information terminal 9 of a warning message. Furthermore, the server 200 transmits a signal to a management facility that manages the production line to discard the object T1 whose discrimination result is "NG (fail)" (or to stop the operation of a conveying device such as a conveyor so that an inspector can visually confirm it).

[0065] By performing machine learning for pass / fail judgment using learning data labeled based on the inspection standards determined in this manner by the standard determination unit 14, the accuracy of pass / fail judgment regarding the surface condition is improved.

[0066] Incidentally, when a manufacturer of the object T1 manufactures the object T1 at multiple bases (factories), there is a possibility that the inspection standards for the surface condition inspection at each base may vary. Therefore, for example, the server 200 at a certain base may transmit information on the inspection standards determined by that server 200 to the servers 200 at other bases via a wide area network such as the Internet, thereby sharing the information. In this case, a unified inspection standard can be established at multiple bases.

[0067] (2.3) Operation The following describes operations (operation examples 1 and 2) of the paint management system 100 equipped with the inspection support system 1. Note that the order of processing in each of the following operation examples is merely an example and is not particularly limited. In addition, some of the processing in each operation example may be omitted, or other processing may be added.

[0068] <Example 1: Creating a sample> Hereinafter, a first operational example regarding sample creation will be described with reference to FIG. 5 (flowchart).

[0069] First, the creator H1 prepares a standard image A1. That is, the painting system 300 performs painting based on the first painting conditions (first standard value P11) on the object T1 to create an actual object (sample product) (step S1). The imaging unit 2 captures an image of the actual object based on the first painting conditions (step S2: generation of the first standard image A11). The imaging unit 2 transmits the first standard image A11 to the server 200 of the inspection support system 1. As a result, the image acquisition unit 11 of the processing unit 10 acquires the first standard image A11 of the object T1 set to the first standard value P11 (image acquisition process).

[0070] The painting system 300 also performs painting based on the second painting conditions (second standard value P12) on the target T1 (the target T1 in step S1 is a separately prepared object) to create an actual object (step S3). The imaging unit 2 captures an image of the actual object based on the second painting conditions (step S4: generation of a second standard image A12). The imaging unit 2 transmits the second standard image A12 to the server 200 of the inspection support system 1. As a result, the image acquisition unit 11 of the processing unit 10 acquires the second standard image A12 of the target T1 set to the second standard value P12 (image acquisition process).

[0071] The inspection support system 1 compares the absolute value (|pp'|) of the difference between the first standard value P11 (painting parameter p) of interest and each of the multiple first standard values ​​P11 (painting parameter p') corresponding to the multiple candidate models N1 with a threshold value. If there is a candidate model N1 in the first storage unit 5 whose absolute value (|pp'|) is smaller than the threshold value (S5: Yes), the inspection support system 1 selects the candidate model N1 as the image generation model M1 (step S6).

[0072] If there is no candidate model N1 whose absolute value (|pp'|) is smaller than the threshold (S5: No), the inspection support system 1 notifies the creator H1 of the result. The creator H1 prepares a new image generation model M1 and inputs the information about it to the inspection support system 1. That is, the inspection support system 1 obtains a new image generation model M1 (step S7).

[0073] The examination support system 1 generates a plurality of evaluation images B1 of the target T1 by varying the condition parameter P1 based on the image generation model M1 with the first and second standard values ​​P11, P12 as references (step S8: image generation process).

[0074] The inspection support system 1 displays the standard image A1 (e.g., the first standard image A11) and a plurality of evaluation images B1 on the display unit 4 (step S9). The creator compares the displayed standard image A1 with each of the plurality of evaluation images B1, evaluates each evaluation image B1 as "OK" or "NG," and inputs the evaluation result. That is, the inspection support system 1 obtains the evaluation result for each evaluation image B1 (step S10).

[0075] The inspection support system 1 determines an inspection standard for the surface state of the object T1 based on the evaluation result (step S11). The inspection support system 1 outputs information on the third coating condition (inspection standard value P13) to the information terminal 9 (step S12).

[0076] The creator H1 prepares a limit sample based on the information of the third painting condition. That is, the painting system 300 performs painting based on the third painting condition on the object T1 and creates a limit sample (step S13).

[0077] For example, an automobile parts manufacturer can easily obtain approval for the manufacture of a painted product (target T1) by sharing the limit sample created in this way with a customer such as an automobile manufacturer. Also, when inspecting a painted product that has completed the painting process during operation of an actual production line, an inspector can check the limit sample, so that the inspection work can be performed stably and accurately. There is a concern that creating limit samples can be extremely costly and time-consuming, but the inspection support system 1 allows limit samples to be created efficiently.

[0078] <Example 2: Inspection> Hereinafter, a second operational example regarding the inspection (inspection process) will be described with reference to FIG. 6 (flowchart).

[0079] During the operation of the production line, the painting system 300 sequentially performs painting on the object T1 based on a predetermined painting condition (e.g., a first painting condition) in a painting process to produce a painted product (a manufactured product or a semi-finished product). The imaging unit 2 sequentially captures images of the painted product that has undergone the painting process (generation of inspection images C1). The imaging unit 2 sequentially transmits the captured inspection images C1 to the server 200 of the inspection support system 1.

[0080] The inspection support system 1 sequentially acquires the inspection images C1 (step S21). Then, the inspection support system 1 (good / bad judgment unit 8) uses the trained model M2 to identify the good / bad paint condition of the object T1 that has undergone the painting process based on the sequentially acquired inspection images C1 (step S22). If the identification result is "OK" (S23: Yes), the inspection support system 1 does not notify a warning message or the like. Then, if the inspection process has not ended (step S24: No), the inspection support system 1 acquires the next inspection image C1 and identifies the good / bad paint condition (return to S21).

[0081] On the other hand, if the identification result is "NG" (S23: No), the inspection support system 1 notifies the information terminal 9 of a warning message (step S25). Furthermore, the inspection support system 1 transmits a stop signal to the management equipment to temporarily stop the operation of the transport device that transports the object T1 (step S26). The inspector goes to the site of the inspection process and visually checks the actual object, and after completing the check, performs an operation to restore the operation of the transport device (step S27). As a result, the inspection process is resumed. Alternatively, if the identification result is "NG", a mechanism for removing the object T1 may be provided to remove it without temporarily stopping the equipment such as the transport device. When the inspection process is completed (step S24: Yes), the process ends.

[0082] <Advantages> When inspecting the surface condition using appearance features, the amount of data for the appearance features can be enormous. For example, in the case of surface coating, the features can be multidimensional if they include color (brightness, saturation, hue), gradation, graininess, glitteriness, glossiness, matteness, etc. Therefore, it is difficult to express the texture felt by humans with a simple feature, and even if it could be expressed with a feature, the number of trials for subjective evaluation would be enormous.

[0083] In contrast, in the inspection support system 1 according to the present embodiment, a plurality of evaluation images B1 are generated using a condition parameter P1 in the process conditions. Therefore, when the evaluation image B1 is used, the inspection standard can be determined more easily than when, for example, a design related to complex appearance features is performed to determine the inspection standard, since low-dimensional parameters (discharge amount, number of coats, etc.) are handled. As a result, the inspection support system 1 has the advantage that a complex design is unlikely to be required.

[0084] Furthermore, in this embodiment, a display unit 4 is provided that displays the standard image A1 and the evaluation image B1, allowing the user to visually check the standard image A1 and the evaluation image B1, making it easier to determine the inspection criteria.

[0085] In the present embodiment, the standard image A1 is an image of the target T1 captured by the imaging unit 2, so that the standard image A1 can be prepared more easily than when the standard image A1 is a CG image, for example. Also, a more accurate inspection standard can be determined.

[0086] Particularly in this embodiment, the image generating unit 12 generates an evaluation image B1 in which the condition parameter P1 is changed between the first standard value P11 and the second standard value P12. This makes it possible to more clearly define the directionality of the change in the paint condition of the target T1. In other words, it becomes easier to quantify the direction in which the color density of the paint changes, and the change characteristics of the surface condition (painted condition) relative to the condition parameter P1 can be linearly approximated.

[0087] (3) Modifications The above embodiment is merely one of various embodiments of the present disclosure. The above embodiment can be modified in various ways depending on the design, etc., as long as the object of the present disclosure can be achieved. In addition, the same function as the examination support system 1 according to the above embodiment may be embodied in an examination support method, a computer program, or a non-transitory recording medium on which a computer program is recorded, etc.

[0088] Modifications of the above embodiment are listed below. The modifications described below can be combined as appropriate and applied. Hereinafter, the above embodiment may be referred to as a "basic example."

[0089] The inspection support system 1 in the present disclosure includes a computer system. The computer system is mainly composed of a processor and a memory as hardware. The processor executes a program recorded in the memory of the computer system, thereby realizing the function of the inspection support system 1 in the present disclosure. The program may be pre-recorded in the memory of the computer system, may be provided through an electric communication line, or may be provided by being recorded in a non-transitory recording medium such as a memory card, an optical disk, or a hard disk drive that can be read by the computer system. The processor of the computer system is composed of one or more electronic circuits including a semiconductor integrated circuit (IC) or a large-scale integrated circuit (LSI). The integrated circuits such as IC or LSI referred to here are called by different names depending on the degree of integration, and include integrated circuits called system LSI, VLSI (Very Large Scale Integration), or ULSI (Ultra Large Scale Integration). Furthermore, a field-programmable gate array (FPGA) that is programmed after the manufacture of the LSI, or a logic device that can reconfigure the connection relationship inside the LSI or the circuit partition inside the LSI, can also be adopted as a processor. The electronic circuits may be integrated in one chip or distributed among multiple chips. The chips may be integrated in one device or distributed among multiple devices. The computer system referred to here includes a microcontroller having one or more processors and one or more memories. Thus, the microcontroller is also composed of one or more electronic circuits including a semiconductor integrated circuit or a large-scale integrated circuit.

[0090] Furthermore, it is not essential that the multiple functions of the test support system 1 are concentrated in one housing. For example, the components of the test support system 1 may be distributed across multiple housings.

[0091] Conversely, multiple functions in the inspection support system 1 may be integrated in one housing. Furthermore, at least some of the functions of the inspection support system 1, for example, some of the functions of the inspection support system 1 may be realized by cloud (cloud computing) or the like.

[0092] (3.1) First Modification The test support system 1 of the first modified example will be described below with reference to Fig. 7. In the following description, components that are substantially the same as those in the basic example will be appropriately denoted by the same reference symbols, and descriptions thereof may be omitted as appropriate.

[0093] In the basic example, the image generating unit 12 generates a plurality of evaluation images B1 by varying the condition parameter P1 on the premise that the change characteristic of the color density with respect to RGB with respect to the condition parameter P1 is linear. However, in reality, the change characteristic of the color density may not increase linearly with respect to the increase in the condition parameter P1, and may increase gradually, for example, in a curved manner.

[0094] Here, the broken line L1 shown in Fig. 7 is the same as the change characteristic in Fig. 3 and Fig. 4 in the basic example. As in the basic example, the inspection support system 1 determines the image data I3 (color density) and the inspection standard value P13 (third coating condition) in the evaluation image B13 (see Fig. 3 and Fig. 4) as the inspection standard. Then, when the coating system 300 paints according to the third coating condition and creates the target T1 (limit sample), the captured image X1 (see Fig. 7) of the limit sample captured by the imaging unit 2 may have a difference in color density compared to the evaluation image B13 generated by the image generation unit 12. In the example of Fig. 7, the color density of the captured image X1 is lower than that of the evaluation image B13.

[0095] In short, the "true change characteristics" in response to changes in coating conditions may have an "error (deviation)" from the ideal straight line (dashed line L1), as shown by the solid curve Y1 in Figure 7. However, the "true change characteristics" of the solid curve Y1 is actually unknown, and it may not be easy to obtain it by calculation or the like.

[0096] Therefore, this modified example differs from the basic example in that a series of steps, including setting a standard value, generating and displaying an evaluation image B1, obtaining an evaluation result by creator H1, and determining an inspection standard based on the evaluation result, are repeated until the "error" is eliminated (for example, until it becomes equal to or less than a predetermined value). Hereinafter, this series of steps will be referred to as the "standard determination process."

[0097] First, the processing unit 10 compares the difference in color density between the evaluation image B13 and its (provisional) limit sample captured image X1 in the first reference determination process (the difference between image data I3 and image data I3') with a predetermined value.

[0098] In the example of FIG. 7, since the difference is equal to or greater than a predetermined value, the "error" is not eliminated in the first reference determination process, and therefore the inspection support system 1 of this modified example executes the second reference determination process. Here, in the second reference determination process, the processing unit 10 sets the inspection reference value P13 of the third painting condition as a new standard value. That is, the image generation unit 12 sets the captured image X1 to the origin, i.e., the first standard image A11. The condition determination unit 16 determines the similarity regarding the new standard value (inspection reference value P13). If there is a candidate model N1 corresponding to a standard value that is highly similar to the new standard value among the multiple candidate models N1, the condition determination unit 16 selects it as the image generation model M1 to be applied in the second reference determination process. If there is no candidate model N1 that is highly similar, the creator H1 prepares a new image generation model M1 and registers it in the inspection support system 1.

[0099] The image generating unit 12 generates a plurality of evaluation images B1 again using the image generation model M1 to be applied in the second reference determination process. The dashed line L2 (straight line) in FIG. 7 indicates the change characteristic of the color density with respect to the condition parameter P1 by the newly applied image generation model M1. The inspection support system 1 displays the generated plurality of evaluation images B1 again on the display unit 4, and determines the inspection standard based on the evaluation result of the creator H1. It is assumed that the inspection support system 1 determines the inspection standard value P14 (fourth painting condition) of the evaluation image B16 (see FIG. 7) as the inspection standard. Then, the painting system 300 creates an object T1 (limit sample) by painting according to the fourth painting condition, and generates a captured image X2 (see FIG. 7) by capturing the limit sample. The processing unit 10 compares the difference in color density between the evaluation image B16 and the captured image X2 of the (provisional) limit sample in the second reference determination process (the difference between the image data I4 and the image data I4') with a predetermined value.

[0100] In the example of FIG. 7, since the difference is equal to or greater than the predetermined value, the "error" is still not eliminated even in the second reference determination process, so the inspection support system 1 executes the third reference determination process. Here, in the third reference determination process, the processing unit 10 sets the inspection reference value P14 of the fourth painting condition as a new standard value. That is, the image generation unit 12 sets the captured image X2 as the origin, i.e., the first standard image A11. The condition determination unit 16 determines the similarity regarding the new standard value (inspection reference value P14). If there is a candidate model N1 corresponding to a standard value that is highly similar to the new standard value among the multiple candidate models N1, the condition determination unit 16 selects it as the image generation model M1 to be applied in the third reference determination process. If there is no candidate model N1 that is highly similar, the creator H1 prepares a new image generation model M1 and registers it in the inspection support system 1.

[0101] The image generating unit 12 generates a plurality of evaluation images B1 again using the image generation model M1 to be applied in the third reference determination process. The dashed line L3 (straight line) in FIG. 7 indicates the change characteristic of the color density with respect to the condition parameter P1 by the newly applied image generation model M1. The inspection support system 1 displays the generated plurality of evaluation images B1 again on the display unit 4, and determines the inspection standard based on the evaluation result of the creator H1. It is assumed that the inspection support system 1 determines the inspection standard value P15 (fifth painting condition) of the evaluation image B17 (see FIG. 7) as the inspection standard. Then, the painting system 300 creates a painted object T1 (limit sample) according to the fifth painting condition, and generates a captured image X3 (see FIG. 7) by capturing the limit sample. The processing unit 10 compares the difference in color density between the evaluation image B17 and the captured image X3 of the (provisional) limit sample in the third reference determination process (the difference between the image data I5 and the image data I5') with a predetermined value.

[0102] In the example of Fig. 7, since the difference is smaller than the predetermined value, it is determined that the "error" has been eliminated in the third standard determination process, and the inspection support system 1 determines the inspection standard value P15 (fifth coating condition) as the true inspection standard in the third standard determination process. Then, the painted product under the fifth coating condition becomes the true limit sample.

[0103] In the above example, the inspection support system 1 automatically determines whether the "error" has been eliminated by comparing with a predetermined value, but the creator H1 may visually determine the same.

[0104] According to the configuration of this modified example, the accuracy of the inspection standard can be further improved. In addition, it is possible to identify the (unknown) solid curve Y1, which is the "true change characteristic," from the trajectory of the captured images X1, X2, X3, etc. As a result, it is possible to create an image generation model M1 that is closer to the actual object.

[0105] (3.2) Second Modification The test support system 1 of the second modified example will be described below with reference to Fig. 8. In the following description, components that are substantially the same as those in the basic example will be appropriately denoted by the same reference symbols, and descriptions thereof may be omitted as appropriate.

[0106] In the basic example, the image generation model M1 is a function model with the condition parameter P1 as a variable. In this modified example, the image generation model M1 is different from the basic example in that it is a model (trained model) that is machine-learned for images generated by changing the condition parameter P1.

[0107] For example, in the first modified example, the reference determination process is repeated to eliminate the "error", that is, to identify the (unknown) solid curve Y1. The image generation model M1 is machine-learned to minimize this "error". Specifically, a neural network that predicts the surface condition is configured as the image generation model M1. The neural network is trained and optimized to minimize the error between the image generated by the image generation model M1 and a sample image of an actual painted sample product. The neural network may be configured as a GAN (Generative Adversarial Network) that uses two networks, a generator and a discriminator, to learn while competing with each other.

[0108] In this case, for example, captured images X1, X2, X3, etc. may be used as training data. By applying the machine-learned image generation model M1 as in this modification, it is possible to make the change characteristic (dashed curve Y2) closer to the solid curve Y1, as shown in FIG.

[0109] (3.3) Third Modification The test support system 1 of the third modified example will be described below with reference to Fig. 9. In the following description, components that are substantially the same as those in the basic example will be appropriately denoted by the same reference symbols, and descriptions thereof may be omitted as appropriate.

[0110] In the basic example, the image data I was related to the change in RGB color density. In this modified example, the image data I is different from the basic example in that it is related to the change in the texture ratio. In FIG. 9, the granular ratio is illustrated as an example of the texture ratio, in which the granular feel is increased by paint containing a glittering material such as metal pieces in the surface coating, resulting in a unique texture. The first standard image A11 has a granular ratio β of 0 (zero). The second standard image A12 has a certain amount of granularity within a predetermined area. The granular ratio β is a normalized value of the painting parameter p, and can take a value from 0 to 1. In other words, the granular ratio β is a value obtained by scaling the values ​​that can be taken by the painting parameter p, such as the discharge amount, the number of paintings, and the atomization pressure, from the first standard image A11 to the second standard image A12, from 0 to 1.

[0111] In this modification, the processing unit 10 of the inspection support system 1 extracts only the granular texture from the second standard image A12 by image processing (see texture A2 in FIG. 9). The processing unit 10 defines the granular ratio β in the texture A2 as “1”.

[0112] In this modification, the image generating unit 12 varies the graininess ratio β (condition parameter P1) between 0 and 1 based on the image generation model M1. The image generating unit 12 generates an evaluation image B1 by adding (combining) a grainy texture A3 according to the graininess ratio β (for example, "0.5") to the first standard image A11. In this case, the image generation model M1 is a model for quantifying the deviation direction of the texture (graininess).

[0113] According to the configuration of this modified example, it is possible to generate an evaluation image B1 even when the texture (graininess) is changed.

[0114] (3.4) Other Modifications In the basic example, the evaluation result is two stages, "OK" and "NG". However, the evaluation result may be three stages or more, for example, as shown in FIG. 10A, three stages may be used, including "OK" and "NG" plus "gray" which is intermediate between them. The creator H1 visually evaluates each evaluation image B1 as shown in FIG. 10B, and inputs the evaluation result (OK, NG, or gray) for each evaluation image B1 using the operation unit 3. In the example of FIG. 10A, the evaluation results of evaluation images B13 and B14 are "gray", and a △ mark is printed near them. In this case, the limit sample may be created using painting conditions corresponding to at least one of the evaluation images B12 to B14 (for example, evaluation images B12 and B14). By providing an intermediate stage (gray) between OK and NG in this way, it is possible to create a more gradual inspection standard. In this case, a pass limit sample that is the limit for "OK" may be created using the painting conditions that correspond to the "OK" evaluation image B12, which is closest to the "gray" evaluation image B13, and a fail limit sample that is the limit for "NG" may be created using the painting conditions that correspond to the "NG" evaluation image B15, which is closest to the "gray" evaluation image B14.

[0115] In the basic example, the display unit 4 simultaneously displays the standard image A1 and each evaluation image B1 in a one-to-one relationship, but is not limited to such a display mode. The display unit 4 may simultaneously display the standard image A1 and the evaluation images B1 in a display mode in which all evaluation images B1 are arranged in a line, as shown in screen Z1 in Fig. 11 (display of a list of evaluation images). Alternatively, the display unit 4 may simultaneously display the standard image A1 and the evaluation images B1 in a display mode in which all evaluation images B1 surround the center of the standard image A1, as shown in screen Z2 in Fig. 11. In this case, the creator H1 can reduce the number of trials compared to paired comparison in which images are compared one by one as in the basic example.

[0116] In addition to displaying the standard image A1 and the evaluation image B1, the display unit 4 may also display the process and results up to the determination of the inspection criteria, as shown in screen Z3 in Fig. 12. In the example of Fig. 12, a graph including the results related to the inspection criteria shown in Fig. 10A is displayed on screen Z3. In this case, the creator H1 can easily understand the process by which the inspection criteria were determined.

[0117] In the basic example, the inspection support system 1 generates the evaluation image B1 by changing the condition parameter P1 in a direction increasing it from the origin (first standard value P11). However, the inspection support system 1 may generate the evaluation image B1 by changing the condition parameter P1 in a direction decreasing it from, for example, the first standard value P11 or the second standard value P12.

[0118] (4) Summary As described above, the inspection support system (1) according to the first aspect includes an image acquisition unit (11) and an image generation unit (12). The image acquisition unit (11) acquires a standard image (A1) of an object (T1), in which a condition parameter (P1) in a process condition related to a surface state of the object (T1) is set to a standard value. The image generation unit (12) generates a plurality of evaluation images (B1) of the object (T1) by changing the condition parameter (P1) based on a predetermined image generation model (M1) and the standard image (A1) with the standard value as a reference.

[0119] According to this embodiment, a plurality of evaluation images (B1) are generated using the condition parameters (P1) in the process conditions. Therefore, by using the evaluation images (B1), the inspection standard can be determined more easily than when the inspection standard is determined by designing complex appearance features, for example. As a result, the inspection support system (1) has the advantage that complex design is unlikely to be required.

[0120] The examination support system (1) according to the second aspect is the first aspect, and further includes a display unit (4) that displays a standard image (A1) and a plurality of evaluation images (B1).

[0121] According to this embodiment, the user can visually check the standard image (A1) and the multiple evaluation images (B1), making it easier to determine the inspection criteria.

[0122] The inspection support system (1) according to the third aspect is the first or second aspect, and further includes an evaluation acquisition unit (13) and a standard determination unit (14). The evaluation acquisition unit (13) acquires evaluation information related to two or more stages of evaluation results for a plurality of evaluation images (B1). The standard determination unit (14) determines an inspection standard for the surface condition of the object (T1) based on the evaluation information.

[0123] According to this aspect, a more accurate inspection standard is determined.

[0124] The test support system (1) according to the fourth aspect is the third aspect, and further includes an output unit (15) that outputs information regarding a condition parameter (P1) corresponding to the test standard.

[0125] According to this aspect, the output information is fed back to a process related to the surface condition, for example, so that an actual product (limit sample) that meets the inspection standard can be created and confirmed.

[0126] The inspection support system (1) according to the fifth aspect is the third or fourth aspect, and further includes a learning unit (7) and a pass / fail judgment unit (8). The learning unit (7) generates a trained model (M2) by using image data to which a pass / fail label regarding a surface condition is assigned, the label being based on the inspection standard determined by the standard determination unit (14), as training data. The pass / fail judgment unit (8) uses the trained model (M2) to make a pass / fail judgment regarding an inspection image (C1) of an object (T1).

[0127] According to this aspect, the accuracy of determining whether the surface condition is good or bad is improved.

[0128] The test support system (1) according to a sixth aspect is any one of the first to fifth aspects, and further includes a storage unit (first storage unit 5) and a condition determination unit (16). The storage unit (first storage unit 5) stores a plurality of candidate models (N1) corresponding to a plurality of standard values ​​in the condition parameter (P1). The condition determination unit (16) determines the similarity between the standard value and each of the plurality of standard values, and if a value having high similarity to the standard value exists among the plurality of standard values, selects the candidate model (N1) corresponding to the value from among the plurality of candidate models (N1) as the image generation model (M1).

[0129] According to this embodiment, the effort of creating or selecting a new image generation model (M1) is eliminated.

[0130] Regarding the examination support system (1) according to the seventh aspect, in any one of the first to sixth aspects, the standard image (A1) is a captured image of the object (T1) captured by the imaging unit (2).

[0131] According to this embodiment, the standard image (A1) can be prepared more easily than when the standard image (A1) is a CG image, and a more accurate inspection standard can be determined.

[0132] With respect to the examination support system (1) according to the eighth aspect, in any one of the first to seventh aspects, the image acquisition unit (11) further acquires a standard image (A1) in which a condition parameter (P1) is set to a second standard value (P12) different from a first standard value (P11) as a standard value. The image generation unit (12) generates a plurality of evaluation images (B1) in which the condition parameter (P1) is changed between the first standard value (P11) and the second standard value (P12).

[0133] According to this embodiment, the directionality of the change in the surface state of the object (T1) can be more clearly defined.

[0134] Regarding the examination support system (1) according to the ninth aspect, in any one of the first to eighth aspects, the image generation model (M1) is a function model with a condition parameter (P1) as a variable.

[0135] According to this aspect, the image generation model (M1) is easier to prepare and less likely to require a more complicated design than when the image generation model (M1) is, for example, a trained model that has been trained by machine learning.

[0136] Regarding the inspection support system (1) according to the tenth aspect, in any one of the first to eighth aspects, the image generation model (M1) is a model trained by machine learning on an image generated by changing the condition parameter (P1).

[0137] According to this embodiment, the accuracy of the image generation model (M1) is improved, and an evaluation image (B1) that is closer to the actual product is more likely to be generated.

[0138] Regarding the inspection support system (1) according to the eleventh aspect, in any one of the first to tenth aspects, the process conditions are painting conditions. The condition parameters (P1) relate to at least one of the paint discharge amount, the paint atomizing pressure, the paint distance to the target (T1) surface, the number of coats, and the paint drying speed.

[0139] According to this aspect, complicated design regarding painting is less necessary.

[0140] The inspection support method according to the twelfth aspect includes an image acquisition process and an image generation process. In the image acquisition process, a standard image (A1) of an object (T1) is acquired, in which a condition parameter (P1) in a process condition related to a surface state of the object (T1) is set to a standard value. In the image generation process, a plurality of evaluation images (B1) of the object (T1) are generated by changing the condition parameter (P1) based on a predetermined image generation model (M1) and the standard image (A1) with the standard value as a reference.

[0141] According to this aspect, it is possible to provide an inspection support method that does not require complex design.

[0142] A program according to a thirteenth aspect is a program for causing one or more processors to execute the inspection support method according to the twelfth aspect.

[0143] According to this embodiment, it is possible to provide a function that does not require a complex design.

[0144] The configurations according to the second to eleventh aspects are not essential for the examination support system (1) and may be omitted as appropriate. [Explanation of symbols]

[0145] 1. Inspection support system 11 Image acquisition section 12 Image Generation Unit 13 Evaluation Department 14 Criteria Determination Department 15 Output section 16 Condition judgment section 2. Imaging unit 4 Display 5 1st memory section (memory section) 7. Learning Department 8. Quality judgement section A1 Standard Image B1 Evaluation Image C1 Inspection image M1 Image generation model M2 pre-trained model N1 candidate model P1 Condition parameters P11 First standard value P12 Second standard value T1 Eligible

Claims

1. an image acquisition unit that acquires a standard image of an object, the standard image being one in which condition parameters in process conditions related to a surface state of the object are set to standard values; an image generation unit that generates a plurality of evaluation images of the object by changing the condition parameters based on a predetermined image generation model and the standard image, with the standard value as a reference; Equipped with the condition parameters are set so that the plurality of evaluation images become evaluation targets for subjective evaluation of appearance features of the surface of the object; Inspection support system.

2. Further comprising a display unit for displaying the standard image and the plurality of evaluation images. The inspection support system according to claim 1 .

3. An evaluation acquisition unit that acquires evaluation information regarding two or more stages of evaluation results for the plurality of evaluation images; and a standard determination unit that determines an inspection standard for a surface state of the object based on the evaluation information. The inspection support system according to claim 1 or 2.

4. Further comprising an output unit that outputs information regarding the condition parameter corresponding to the inspection standard. R, The inspection support system according to claim 3 .

5. A learning unit that generates a trained model by using image data to which the label of pass / fail regarding the surface condition is assigned, the label being based on the inspection standard determined by the standard determination unit, as training data; and A pass / fail determination unit that uses the trained model to determine pass / fail of the inspection image of the target; Further comprising: The inspection support system according to claim 3 or 4.

6. A storage unit that stores a plurality of candidate models respectively corresponding to a plurality of standard values ​​of the condition parameters; a condition determination unit that determines a similarity between the standard value and each of the plurality of standard values, and if a value having high similarity to the standard value exists among the plurality of standard values, selects a candidate model corresponding to the value among the plurality of candidate models as the image generation model; Further comprising: The inspection support system according to any one of claims 1 to 5.

7. The standard image is an image of the object captured by an imaging unit. The inspection support system according to any one of claims 1 to 6.

8. The image acquisition unit further acquires the standard image in which the condition parameter is set to a second standard value different from the first standard value as the standard value, the image generating unit generates the plurality of evaluation images in which the condition parameter is changed between the first standard value and the second standard value. The inspection support system according to any one of claims 1 to 7.

9. The image generation model is a function model with the condition parameters as variables. The inspection support system according to any one of claims 1 to 8.

10. The image generation model is a model trained by machine learning on images generated by changing the condition parameters. The inspection support system according to any one of claims 1 to 8.

11. The process conditions are painting conditions, The condition parameters relate to at least one of the amount of paint discharged, the pressure at which the paint is atomized, the distance to the target surface to be painted, the number of times the paint is applied, and the drying speed of the paint. The inspection support system according to any one of claims 1 to 10.

12. An image acquisition process for acquiring a standard image of an object, the standard image being obtained by setting condition parameters in process conditions relating to a surface state of the object to standard values; The standard value is used as a reference, and the condition pattern is calculated based on a predetermined image generation model and the standard image. an image generation process for generating a plurality of evaluation images of the object with different parameters; Including, the condition parameters are set so that the plurality of evaluation images become evaluation targets for subjective evaluation of appearance features of the surface of the object; Inspection support methods.

13. A program for causing one or more processors to execute the inspection support method described in claim 12.

Images