Mating condition inspection device and mating condition inspection method

Abstract

The present invention provides a connector mating condition inspection device and method that can determine minute differences with high precision. [Solution] The mating state inspection device inspects the mating state of a connector that mats when a locking piece overcomes a claw and engages with the claw. The device comprises: an imaging means for photographing an area including the locking piece; an image cropping means for cropping an area including the tip side of the locking piece from the captured image as a first discrimination image and an area including the rear end side of the locking piece as a second discrimination image; a first determination means for determining the mating state from the first discrimination image and a first image learning model that has learned the shape characteristics of the tip side of the locking piece by machine learning; a second determination means for determining the mating state from the second discrimination image and a second image learning model that has learned the shape characteristics of the rear end side of the locking piece by machine learning; and an overall determination means for making an overall determination based on the determination results from the first determination means and the second determination means.

Description

Technical Field

[0001] The present invention relates to a fitting state inspection apparatus and a fitting state inspection method for inspecting the fitting state of a connector.

Background Art

[0002] In an assembly work process of a vehicle engine or the like, for example, a plurality of wiring connections are made by connectors. If the fitting state of the connector is incomplete, it may cause assembly defects and also lead to a deterioration in the quality of the product. Therefore, conventionally, when fitting the connector, the fitting sound, the pushing-in feeling, and visual inspection are used to confirm whether the connector is properly fitted. However, conventionally, since it was confirmed by the five senses of the operator, there was a problem that variations and oversights due to insufficient confirmation occurred.

[0003] In Patent Document 1, in an inspection method for inspecting the fitting state of a first connector having an engaging portion and a second connector having an elastically deformable engaging claw that engages with the engaging portion, a portion including the engaging claw is photographed by a camera, the edge of the engaging claw is detected from the image obtained by the photographing, the opening dimension of the engaging claw is measured, and the fitting state is determined by comparing it with a threshold value for determining an appropriate fitting state.

[0004] In addition, in Patent Document 2, in an inspection method for inspecting the fitting state of a connector including an insert having a claw portion protruding outward and a fitting body having a locking piece that engages with the claw portion to prevent the insert from coming out when the insert is inserted, light is irradiated onto the claw portion and the locking piece in a state where the insert and the fitting body are connected, and imaging is performed by an imaging device. A light and shade image pattern of the reflected light is created from the obtained image, the inclined state of the locking piece is detected from the light and shade image pattern, and the fitting state of the connector is determined.

[0005] However, the method described in Patent Document 1 detects the edge of each engaging claw and compares it to the correct dimensions, but there is a possibility of overlooking minute deformations. Furthermore, the method described in Patent Document 2 utilizes light reflection, making it susceptible to surface conditions and ambient light. Therefore, the methods in Patent Document 1 or Patent Document 2 have the problem of being unable to accurately determine the minute differences between properly fitted and improperly fitted cases. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] Japanese Patent Publication No. 2013-175347 [Patent Document 2] Japanese Patent Publication No. 2014-107177 [Overview of the project] [Problems that the invention aims to solve]

[0007] This invention was made based on these problems and aims to provide a connector mating state inspection device and mating state inspection method that can determine minute differences with high accuracy. [Means for solving the problem]

[0008] The fitting state inspection device of the present invention inspects the fitting state of a connector in which an insert having a claw portion is inserted into a fitting body having a locking piece that engages with the claw portion, and the insert and the fitting body are fitted together when the locking piece overcomes the claw portion and engages with the claw portion. The device includes an imaging means for imaging a region including the locking piece when the insert is inserted into the fitting body, an image cropping means for cropping a region including the tip side of the locking piece as a first discriminant image and a region including the rear end side of the locking piece as a second discriminant image from the image obtained by the imaging means, and the first discriminant image and machine learning to determine whether the connector is properly fitted or not. The system includes: a first determination means for determining whether the connector is properly fitted based on a first image learning model that has been trained to recognize the shape characteristics of the tip side of the locking piece in the case where it is not properly fitted; a second determination means for determining whether the connector is properly fitted based on a second determination image and a second image learning model that has been trained by machine learning to recognize the shape characteristics of the rear end side of the locking piece in the cases where the connector is properly fitted and when it is not properly fitted; and an overall determination means for determining whether the connector is properly fitted based on the first determination result from the first determination means and the second determination result from the second determination means.

[0009] The present invention provides a method for inspecting the mating state of a connector in which an insert having a claw portion is inserted into a mating body having a locking piece that engages with the claw portion, and the insert and mating body are mated when the locking piece overcomes the claw portion and engages with the claw portion. The method includes: an imaging procedure in which an image of the region including the locking piece is taken with the insert inserted into the mating body; an image cropping procedure in which a region including the tip side of the locking piece is cropped as a first discriminant image and a region including the rear end side of the locking piece is cropped as a second discriminant image from the image obtained in the imaging procedure; and the first discriminant image and machine learning to determine whether the connector is properly mated or not. The method includes: a first determination procedure for determining whether the connector is properly mated based on a first image learning model that has been trained to recognize the shape characteristics of the tip side of the locking piece in the case where it is not properly mated; a second determination procedure for determining whether the connector is properly mated based on a second discriminant image and a second image learning model that has been trained by machine learning to recognize the shape characteristics of the rear end side of the locking piece in the cases where the connector is properly mated and when it is not properly mated; and an overall determination procedure for determining whether the connector is properly mated based on the first determination result from the first determination procedure and the second determination result from the second determination procedure. [Effects of the Invention]

[0010] According to the present invention, the region including the tip side of the locking piece and the region including the rear end side of the locking piece are cut out from the captured image, and the determination is made using a first image learning model and a second learning model that have learned features by machine learning. In the captured image, the area where features appear in the case where the connector is properly mated and the case where it is not properly mated is cut out as a square, and the image can be used without compression. Therefore, the features of the image are not compressed, and even minute differences can be determined with high accuracy.

[0011] Furthermore, by positioning the camera to photograph the locking piece from an oblique angle, the difference in the shape of the protrusions and indentations between a properly fitted and a non-fitted piece becomes more pronounced, allowing for more accurate determination.

[0012] Furthermore, if the shooting means includes a brightness adjustment means that changes the brightness setting according to the color of the connector, it is possible to improve the situation where the uneven shape becomes unclear due to so-called overexposure or underexposure, and to make judgments with higher accuracy. [Brief explanation of the drawing]

[0013] [Figure 1] This figure shows the configuration of a fitting condition inspection device according to one embodiment of the present invention. [Figure 2] This figure shows an example of a connector configuration that is inspected by the mating condition inspection device shown in Figure 1. [Figure 3] This is another diagram showing an example of a connector configuration to be inspected by the mating condition inspection device shown in Figure 1. [Figure 4] This diagram simulates an image of a properly mated connector, taken from an oblique angle to the locking piece. [Figure 5] This diagram simulates an image of a partially mated connector, taken from an oblique angle with respect to the locking piece. [Figure 6] This figure shows the region in the captured image that includes the entire locking piece to be extracted. [Figure 7] This figure illustrates examples of the first and second discrimination images extracted from a captured image. [Figure 8] This figure shows an example of the hardware configuration of the image cropping means, first determination means, second determination means, and overall determination means shown in Figure 1. [Figure 9] This diagram shows the flow of a fitting condition inspection method according to one embodiment of the present invention. [Modes for carrying out the invention]

[0014] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0015] FIG. 1 shows the configuration of the fitting state inspection apparatus 1 according to an embodiment of the present invention. FIGS. 2 and 3 show configuration examples of the connector M inspected by the fitting state inspection apparatus 1. The fitting state inspection apparatus 1 inspects the fitting state of the connector M, that is, whether the connector M is properly fitted or not.

[0016] As the connector M inspected by the fitting state inspection apparatus 1, preferably, an insertion body M10 having a claw portion M11 and a fitting body M20 having a locking piece M21 engaged with the claw portion M11 are provided. For the connector M, for example, when the insertion body M10 is inserted into the fitting body M20 from the state shown in FIG. 3(A), as shown in FIG. 3(B), the locking piece M21 gets over the claw portion M11, and as shown in FIG. 3(C), the locking piece M21 is engaged with the claw portion M11, so that the insertion body M10 and the fitting body M20 are configured to be fitted. The locking piece M21 is, for example, arranged to be rotatable about a fulcrum M23 with respect to the fitting body main body M22. In FIG. 2, (A) represents the state where the insertion body M10 and the fitting body M20 are fitted, and (B) represents the cross-sectional configuration of the insertion body M10. In FIG. 3, in order to make the configuration of the connector M easy to understand, the claw portion M11 and the locking piece M21 are taken out and shown in cross section.

[0017] The fitting state inspection apparatus 1 includes, for example, a photographing means 10 for photographing a region including the locking piece M21 in a state where the insertion body M10 is inserted into the fitting body M20, an image cutting means 20 for cutting out a region including the tip side of the locking piece M21 as a first discrimination image and cutting out a region including the rear end side of the locking piece M21 as a second discrimination image from the photographed image obtained by the photographing means 10, a first determination means 30 for determining whether the connector M is properly fitted or not based on the first discrimination image, a second determination means 40 for determining whether the connector M is properly fitted or not based on the second discrimination image, and a comprehensive determination means 50 for determining whether the connector M is properly fitted or not based on the first determination result by the first determination means 30 and the second determination result by the second determination means 40. Further, the fitting state inspection apparatus 1 preferably includes, for example, a display means 60 for displaying the determination result of the comprehensive determination means 50.

[0018] The photographing means 10 is constituted by a camera such as a CCD camera, for example, and is arranged to photograph the region including the locking piece M21 after the insertion body M10 is inserted into the fitting body M20. This is because a difference appears in the shape of the locking piece M21 between the case where the connector M is normally fitted and the case where it is not normally fitted. When it is not normally fitted, the tip side of the locking piece M21 rises and the rear end side of the locking piece M sinks compared to the case where it is normally fitted (see FIG. 3), and characteristic features of the shape appear on the tip side and the rear end side of the locking piece M21. Note that the tip side of the locking piece M21 is the side that gets over the claw portion M11 and engages with the claw portion M11, and the rear end side of the locking piece M21 is the opposite side.

[0019] The photographing means 10 is preferably arranged to photograph the locking piece M21 from an oblique direction. This is because, by photographing from an oblique direction, the difference in the concavo-convex shape of the locking piece M21 appears more prominently between the case where it is normally fitted and the case where it is not normally fitted. FIG. 4 shows a diagram simulating a photographed image of the locking piece M21 taken from an oblique direction for the normally fitted connector M. Also, FIG. 5 shows a diagram simulating a photographed image of the locking piece M21 taken from an oblique direction for the semi-fitted connector M. As shown in FIGS. 4 and 5, in the case of semi-fitting, it can be seen that the tip side of the locking piece M21 rises and the rear end side of the locking piece M sinks, and the shape of the locking piece M21 in the photographed image looks significantly different compared to the case where it is normally fitted.

[0020] The oblique direction with respect to the locking piece M21 means an oblique direction with respect to the facing direction of the locking piece M21, that is, the protruding direction of the claw portion M11. Further, the photographing means 10 is more preferably arranged to photograph from a direction intersecting the plane including the protruding direction of the claw portion M11 and the insertion direction of the insertion body M10. This is because the difference in the concavo-convex shape appears more prominently.

[0021] Preferably, the shooting means 10 also has an adjustment means 11 that changes the brightness setting according to the color of the connector M. This is because it can improve the blurring of uneven shapes caused by so-called overexposure or underexposure. The adjustment means 11 is configured such that, for example, a gamma value is set as a correction value for each color of the connector M, and the correction value of the shooting means 10 is changed to a preset value according to the color of the connector M to be photographed. The color of the connector M to be photographed may be detected by the shooting means 10 from an image taken before shooting.

[0022] The image cropping means 20, the first determination means 30, the second determination means 40, and the comprehensive determination means 50 can be configured, for example, by a computer, and are configured to function as the determination means 20, the first determination means 30, the second determination means 40, or the comprehensive determination means 50 by executing a program.

[0023] The image cropping means 20 is connected to the shooting means 10, for example, and is configured to crop a first discrimination image and a second discrimination image from the captured image input from the shooting means 10. For example, as shown in Figure 6, the image cropping means 20 is preferably configured to extract a region including the entire locking piece M21 from the captured image, and from the extracted image, as shown in Figure 7, crop a region including the tip side of the locking piece M21 as the first discrimination image, and crop a region including the rear end side of the locking piece M21 as the second discrimination image. In Figure 6, the region including the entire locking piece M21 to be extracted is shown by a thick dashed line. Also, in Figure 7, (A) is a representation of the first discrimination image cropped with the region including the tip side of the locking piece M21, and (B) is a representation of the second discrimination image cropped with the region including the rear end side of the locking piece M21.

[0024] In other words, it is preferable that the image cropping means 20 includes, for example, a locking piece extraction means 21 that extracts a region including the entire locking piece M21 from a captured image, a first discriminant image cropping means 22 that crops a region including the tip side of the locking piece M21 from the image extracted by the locking piece extraction means 21 as a first discriminant image, and a second discriminant image cropping means 23 that crops a region including the rear end side of the locking piece M21 from the image extracted by the locking piece extraction means 21 as a second discriminant image.

[0025] When training features using machine learning, images are compressed into squares. However, the region containing the entire locking piece M21 is rectangular. Therefore, if an image is used that extracts the region containing the entire locking piece M21, the feature portion will be compressed and distorted, making it difficult to distinguish subtle differences. For this reason, it is preferable to crop the front and rear ends of the locking piece M21, where features appear in both the case where the connector M is properly mated and the case where it is not properly mated, into squares, so that the feature portion can be used as is without compressing the image.

[0026] For example, the region containing the entire locking piece M21 may be extracted from the captured image as a rectangle with a 1:2 aspect ratio, and then divided into two parts to obtain a first discriminant image and a second discriminant image. The detection of the locking piece M21 in the captured image can be performed using known object detection techniques, and is preferably done by image processing such as pattern matching.

[0027] The first determination means 30 determines whether the connector M is properly fitted or not, based on the first discrimination image and the first image learning model 31, which has been trained by machine learning to recognize the shape characteristics of the tip side of the locking piece M21 when the connector M is properly fitted and when it is not properly fitted. The first determination means 30 is connected to, for example, the image cropping means 20, and is configured to receive the first discrimination image from the image cropping means 20. The first determination means 30 is also connected to, for example, the comprehensive determination means 50, and is configured to output the first determination result determined by the first determination means 30 to the comprehensive determination means 50.

[0028] Preferably, the first determination means 30 includes, for example, a first image learning model 31, a first output value determination means 32 that inputs a first discriminant image to the first image learning model 31 and determines whether the connector M is properly mated based on the output value obtained thereby, and outputs this as a first determination result, and a first image learning model generation means 33 that generates the first image learning model 31. Preferably, the determination in the first output value determination means 32 is configured to determine, for example, whether the output value obtained from the first image learning model 31, i.e., the score value indicating whether it is properly mated or not, is greater than a preset threshold.

[0029] The first image learning model generation means 33 is preferably configured to prepare multiple first learning images, for example, as positive examples of the first discriminant image, for cases where the connector M is properly fitted and for cases where the connector M is not properly fitted, and to extract the shape features of the locking piece M21 in each first learning image using deep learning, and to learn them in association with the fitted state of the connector M. It is preferable to use a convolutional neural network (CNN) for deep learning.

[0030] Each first learning image can be obtained by, for example, taking pictures with the shooting means 10 in the case where the connector M is properly fitted and in the case where it is partially fitted (i.e., not properly fitted), and then cutting out the region including the tip side of the locking piece M21 from the obtained captured images in the same manner as the image cropping means 20 described above.

[0031] The second determination means 40 determines whether the connector M is properly fitted or not, based on the second discrimination image and a second image learning model 41 that has been trained by machine learning to recognize the shape characteristics of the rear end side of the locking piece M21 when the connector M is properly fitted and when it is not properly fitted. The second determination means 40 is connected to, for example, an image cropping means 20, and is configured to receive the second discrimination image from the image cropping means 20. The second determination means 40 is also connected to, for example, an overall determination means 50, and is configured to output the second determination result determined by the second determination means 40 to the overall determination means 50.

[0032] The second determination means 40 preferably includes, for example, a second image learning model 41, a second output value determination means 42 that inputs a second discriminant image to the second image learning model 41 and determines whether the connector M is properly mated based on the output value obtained thereby, and outputs it as a second determination result, and a second image learning model generation means 43 that generates the second image learning model 41. The determination in the second output value determination means 42 is preferably configured to determine, for example, whether the output value obtained from the second image learning model 41, i.e., the score value indicating whether it is properly mated or not, is greater than a preset threshold.

[0033] The second image learning model generation means 43 is preferably configured to prepare multiple second learning images, for example, as positive examples of the second discriminant image, for cases where the connector M is properly fitted and for cases where the connector M is not properly fitted, and to extract the shape features of the locking piece M21 in each second learning image using deep learning and to learn them in association with the fitting state of the connector M. It is preferable to use a convolutional neural network for deep learning.

[0034] Each second learning image can be obtained by, for example, taking pictures with the shooting means 10 in the case where the connector M is properly fitted and in the case where it is partially fitted (i.e., not properly fitted), and then cutting out the region including the rear end side of the locking piece M21 from the obtained captured images in the same manner as the image cropping means 20 described above.

[0035] The comprehensive determination means 50 is connected, for example, to the first determination means 30, the second determination means 40, and the display means 60, and is configured to perform a comprehensive determination from the first determination result input from the first determination means 30 and the second determination result input from the second determination means 40, and to output it to the display means 60. Preferably, the comprehensive determination is configured to determine that the parts are "fitted properly" if both the first and second determination results indicate "fitted properly," and to determine that the parts are "not fitted properly" if at least one of the first and second determination results indicates "not fitted properly."

[0036] Figure 8 shows an example of the hardware configuration of the image cropping means 20, the first determination means 30, the second determination means 40, and the comprehensive determination means 50. The image cropping means 20, the first determination means 30, the second determination means 40, and the comprehensive determination means 50 each include, for example, a CPU (Center Processing Unit) 71, a ROM (Read Only Memory) 72, a RAM (Random Access Memory) 73, an HDD (Hard Disk Drive) 74, and an operation interface (operation I / F) 75. The CPU 71 executes various processes according to various programs recorded in the ROM 72 or various programs loaded from the HDD 74 into the RAM 73. The RAM 73 also appropriately stores data necessary for the CPU 71 to execute various processes. The HDD 74 stores various data.

[0037] The mating state inspection device 1 is used, for example, as follows. Figure 9 shows the flow of the mating state inspection method using the mating state inspection device 1. First, as a preparation procedure, the first image learning model 31 is generated by the first image learning model generation means 33, and the second image learning model 41 is generated by the second image learning model generation means 43 (preparation procedure; step S110).

[0038] In the preparation procedure, first, for example, when connector M is properly mated and when connector M is partially mated (not properly mated), images are taken using the imaging means 10, and the front and rear ends of the locking piece M21 are cropped to prepare a first and second learning image, which will serve as positive examples. Next, for example, the first learning image is used to extract the shape features of the front end of locking piece M21 using deep learning and to learn them in association with the mating state of connector M. At the same time, the second learning image is used to extract the shape features of the rear end of locking piece M21 using deep learning and to learn them in association with the mating state of connector M.

[0039] After the preparation procedure (step S110), for example, it is checked whether the connector M is properly mated. First, for example, the imaging means 10 is used to photograph the area including the locking piece M21 with the insert M10 inserted into the mating body M20 (imaging procedure; step S120). At this time, it is preferable that the imaging means 10 photographs from an oblique direction with respect to the locking piece M21. Also, for example, it is preferable to change the brightness setting of the imaging means 10 using the adjustment means 11 according to the color of the connector M.

[0040] Next, the image cropping means 20 crops the region including the tip side of the locking piece M21 as a first discriminant image from the captured image obtained by the shooting means 10, and crops the region including the rear end side of the locking piece M21 as a second discriminant image (image cropping procedure; step S130). For example, the locking piece extraction means 21 extracts the region including the entire locking piece M21 from the captured image, and from the extracted image, the first discriminant image cropping means 22 crops the region including the tip side of the locking piece M21 as a first discriminant image, and the second discriminant image cropping means 23 crops the region including the rear end side of the locking piece M21 as a second discriminant image.

[0041] Next, the first determination means 30 uses the first discrimination image and the first image learning model 31 to determine whether or not the connector M is properly fitted and outputs the result as the first determination (first determination procedure; step S140). Specifically, for example, the first output value determination means 32 inputs the first discrimination image to the first image learning model 31 and determines whether or not the connector M is properly fitted based on the output value obtained therefrom.

[0042] Furthermore, the second determination means 40 uses the second discrimination image and the second image learning model 41 to determine whether or not the connector M is properly mated and outputs this as the second determination result (second determination procedure; step S150). Specifically, for example, the second output value determination means 42 inputs the second discrimination image to the second image learning model 41 and determines whether or not the connector M is properly mated based on the output value obtained therefrom.

[0043] Next, the comprehensive determination means 50 makes a comprehensive determination of whether the connector M is properly mated based on the first determination result from the first determination means 30 and the second determination result from the second determination means 40 (comprehensive determination procedure; step S160). Specifically, for example, if both the first determination result and the second determination result indicate "properly mated," it is determined that "properly mated," and if at least one of the first determination result and the second determination result indicates "not properly mated," it is determined that "not properly mated." After that, the display means 50 displays the comprehensive determination result from the comprehensive determination means 50 (display procedure; step S170).

[0044] As described above, according to this embodiment, the region including the tip side of the locking piece M21 and the region including the rear end side of the locking piece M21 are cut out from the captured image, and the first image learning model 31 and the second learning model 41, whose features have been learned by machine learning, are used for determination. In this way, the areas in the captured image where features appear in the case where the connector M is properly fitted and the case where it is not properly fitted are cut out as squares, and the image can be used without compression. Therefore, the features of the image are not compressed, and even minute differences can be determined with high accuracy.

[0045] Furthermore, by positioning the imaging device 10 to photograph the locking piece M21 from an oblique angle, the difference in the shape of the protrusions and indentations between a properly fitted and a non-fitted state becomes more pronounced, allowing for more accurate determination.

[0046] Furthermore, if the shooting means 10 is equipped with a brightness adjustment means 11 that changes the brightness setting according to the color of the connector M, it is possible to improve the problem of uneven shapes becoming unclear due to so-called overexposure or underexposure, and to make judgments with higher accuracy.

[0047] The present invention has been described above with reference to embodiments, but the present invention is not limited to the above embodiments and can be modified in various ways. For example, although each component was described in detail in the above embodiments, the specific structure and shape of each component may differ, and the present invention does not have to include all of the above-mentioned components, but may include other components as well. [Explanation of Symbols]

[0048] 1…Matching state inspection device, 10…Photography means, 11…Adjustment means, 20…Image cropping means, 21…Locking piece extraction means, 22…First discriminant image cropping means, 23…Second discriminant image cropping means, 30…First determination means, 31…First image learning model, 32…First output value determination means, 33…First image learning model generation means, 40…Second determination means, 41…Second image learning model, 42…Second output value determination means, 43…First image learning model generation means, 50…Comprehensive determination means, 60…Display means, 71…CPU, 72…ROM, 73…RAM, 74…HDD, 75…Operation interface, M…Connector, M10…Insertor, M11…Claw part, M20…Matching body, M21…Locking piece, M22…Matching body body, M23…Pivot point

Claims

1. A fitting state inspection device for inspecting the fitting state of a connector in which an insert having a claw portion is inserted into a fitting body having a locking piece that engages with the claw portion, and the locking piece overcomes the claw portion and engages with the claw portion, thereby fitting the insert and the fitting body together, A photographic means for photographing the region including the locking piece when the insert is inserted into the fitting body, Image cropping means crops the region including the tip side of the locking piece as a first discriminant image from the captured image obtained by the aforementioned shooting means, and crops the region including the rear end side of the locking piece as a second discriminant image. A first determination means for determining whether the connector is properly fitted or not, based on the first determination image and a first image learning model that has been trained by machine learning to recognize the shape characteristics of the tip side of the locking piece in cases where the connector is properly fitted and when it is not properly fitted. A second determination means for determining whether the connector is properly fitted or not, based on the second discrimination image and a second image learning model that has been trained by machine learning to recognize the shape characteristics of the rear end side of the locking piece in cases where the connector is properly fitted and when it is not properly fitted. A comprehensive determination means that determines whether the connector is properly fitted based on the first determination result from the first determination means and the second determination result from the second determination means. A fitting condition inspection device characterized by being equipped with the following features.

2. The fitting condition inspection device according to claim 1, characterized in that the photographing means is arranged to photograph the locking piece from an oblique direction.

3. The mating condition inspection device according to claim 1, characterized in that the imaging means has a brightness adjustment means that changes the brightness setting according to the color of the connector.

4. A fitting state inspection method for inspecting the fitting state of a connector in which an insert having a claw portion is inserted into a fitting body having a locking piece that engages with the claw portion, and the locking piece overcomes the claw portion and engages with the claw portion, thereby fitting the insert and the fitting body together, A photographic procedure for photographing the region including the locking piece while the insert is inserted into the fitting body, From the captured image obtained by the above shooting procedure, an image cropping procedure is performed in which the region including the tip side of the locking piece is cropped as a first discriminant image, and the region including the rear end side of the locking piece is cropped as a second discriminant image. A first determination procedure for determining whether the connector is properly fitted or not, based on the first determination image and a first image learning model that has been trained by machine learning to recognize the shape characteristics of the tip side of the locking piece when the connector is properly fitted and when it is not properly fitted, A second determination procedure for determining whether the connector is properly fitted or not, based on the second determination image and a second image learning model that has been trained by machine learning to recognize the shape characteristics of the rear end side of the locking piece in cases where the connector is properly fitted and when it is not properly fitted, A comprehensive determination procedure that determines whether the connector is properly mated based on the first determination result from the first determination procedure and the second determination result from the second determination procedure. A fitting condition inspection method characterized by including the following.

Images