Defect Detection Method, Device, Equipment and Storage Medium for Flexible Printed Circuit Board

Abstract

The present invention relates to a method, device, equipment and storage medium for defect detection of a flexible printed circuit board. First, an image of the flexible printed circuit board after the conductive layer covers the insulating layer is obtained as a captured image, and then preprocessing is performed to obtain an optimized image. After that, based on a preset recognition model, the positions of a plurality of preset holes in the optimized image are recognized, and an image of each preset hole is intercepted to obtain a hole image. Finally, according to the area size relationship among the first color, the second color, the third color and the fourth color in the hole image, it is detected whether the conductive layer and the insulating layer are aligned. Compared with the prior art, the present invention realizes automatic detection of hole alignment based on image recognition technology, and when detecting alignment, it determines whether the holes are aligned by comparing the area size relationship among the first color, the second color, the third color and the fourth color, perfectly solving the problem of how to automatically detect whether the holes are aligned before bonding the insulating layer and the conductive layer.

Description

Technical Field

[0001] The present invention relates to the technical field of image defect detection, and particularly to a method, device, equipment and storage medium for defect detection of flexible printed circuit boards. Background Art

[0002] A flexible printed circuit board (FPC) is a special electronic component composed of multiple layers of materials. It mainly includes a substrate layer, usually a flexible insulating material such as polyester film or polyimide film; a conductive layer, using conductive copper foil to form the wires and welding areas of the circuit; and an insulating layer, used to isolate the conductive layer to prevent circuit short - circuit or other damages. These hierarchical structures enable the flexible printed circuit board to not only have good electrical performance, but also adapt to bending and space - limited application environments, such as mobile devices, medical devices and automotive electronic products.

[0003] Currently, when producing a flexible printed circuit board, it is necessary to bond the conductive layer and the insulating layer. However, when bonding, it is necessary to ensure that the holes opened on the conductive layer and the insulating layer are as aligned as possible, so as to ensure that in subsequent component installation and circuit connection, the position and size of the holes meet the design requirements, thereby avoiding circuit connection problems and functional failures caused by hole position deviation.

[0004] In the prior art, before bonding the insulating layer and the conductive layer, people usually use a magnifying glass manually to detect the defect that the holes are not aligned. Obviously, this method is time - consuming and laborious, and has a high error rate. Therefore, people need a solution that can automatically detect whether the holes are aligned before bonding the insulating layer and the conductive layer. Summary of the Invention

[0005] Therefore, the present invention provides a defect detection of flexible printed circuit boards to solve the problem of how to automatically detect whether the holes are aligned before bonding the insulating layer and the conductive layer in the prior art.

[0006] The present invention provides a method for defect detection of flexible printed circuit boards, including:

[0007] Obtain a captured image, which is an image of the flexible printed circuit board after the conductive layer covers the insulating layer, taken perpendicular to the plane where the flexible printed circuit board is located;

[0008] Pre - process the captured image to obtain an optimized image;

[0009] Based on a preset recognition model, identify the positions of multiple preset holes in the optimized image, and intercept the image of each preset hole to obtain a hole image;

[0010] Detect whether the conductive layer and the insulating layer are aligned according to the area size relationship among the first color, the second color, the third color, and the fourth color in the hole image, where the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color presented after the insulating layer covers the conductive layer, and the fourth color is the background color when the image is captured.

[0011] The present invention also provides a preferred solution: identify the positions of multiple preset holes in the optimized image based on a preset recognition model, and intercept the images of each preset hole to obtain hole images, including:

[0012] Divide multiple grids of preset specifications in the optimized image;

[0013] Identify the probability of the existence of a preset hole in each grid based on the preset recognition model;

[0014] Obtain the positions of multiple preset holes and intercept the images of each preset hole according to the probability of the existence of a preset hole in each grid.

[0015] The present invention also provides a preferred solution: the area of the grid is smaller than the area of the smallest preset hole.

[0016] The present invention also provides a preferred solution: the preset recognition model is a convolutional neural network model, and the preset recognition model includes an input layer, at least one convolutional layer, at least one fully connected layer, and an output layer connected in sequence, where the number of nodes in the input layer is equal to the number of pixels in the optimized image, each node in the input layer is respectively used to input the three color channel values of a pixel in the optimized image, the number of nodes in the output layer is equal to the number of grids in the optimized image, each node in the output layer is respectively used to output a probability description vector corresponding to a grid, and the elements of the probability description vector respectively include: the confidence probability of at least one detection frame where the preset hole exists in this grid, the center coordinates of at least one detection frame where the preset hole exists in this grid, and the size of at least one detection frame where the preset hole exists in this grid.

[0017] The present invention also provides a preferred solution: obtain the positions of multiple preset holes and intercept the images of each preset hole according to the probability of the existence of a preset hole in each grid, including:

[0018] Obtain multiple target detection frames with the highest confidence probabilities and no overlap according to the multiple probability description vectors output by the preset recognition model;

[0019] Intercept the positions where the multiple target detection frames are located from the optimized image to obtain the images of multiple preset holes.

[0020] The present invention also provides a preferred solution: detecting whether the conductive layer and the insulating layer are aligned according to the area size relationship among the first color, the second color, the third color, and the fourth color in the hole image, including:

[0021] According to the area size relationship between the first color and the second color, and the area ratio relationship between the third color and the fourth color in each hole image, obtain the hole overlap evaluation value of each hole image, where the hole overlap evaluation value is used to characterize the overlapping degree of the holes opened on the conductive layer and the overlapping holes opened on the insulating layer in the hole image;

[0022] If the hole overlap evaluation values corresponding to all hole images do not exceed the preset standard threshold, it is determined that the conductive layer and the insulating layer are aligned.

[0023] The present invention also provides a preferred solution: calculating the hole overlap evaluation value by the following formula:

[0024] ;

[0025] Wherein, represents the hole overlap evaluation value, , , and are respectively the areas of the first color, the second color, the third color, and the fourth color in the hole image, is the area of the preset hole corresponding to the type of the hole image, is the area of the target detection frame corresponding to the hole image, is a preset hyperparameter, and are respectively different preset weights.

[0026] The present invention also provides a defect detection device for a flexible printed circuit board, including:

[0027] An image acquisition module for acquiring an acquisition image, where the acquisition image is an image of the flexible printed circuit board after the conductive layer covers the insulating layer, taken perpendicular to the plane where the flexible printed circuit board is located;

[0028] An image optimization module for preprocessing the acquisition image to obtain an optimized image;

[0029] A hole recognition module for identifying the positions of a plurality of preset holes in the optimized image based on a preset recognition model, and intercepting the image of each preset hole to obtain a hole image;

[0030] A hole detection module for detecting whether the conductive layer and the insulating layer are aligned according to the area size relationship among the first color, the second color, the third color, and the fourth color in the hole image, where the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color presented after the insulating layer covers the conductive layer, and the fourth color is the background color when the acquisition image is captured.

[0031] The present invention also provides an electronic device, including:

[0032] A memory and a processor;

[0033] The memory is used to store a program, and the processor is used to execute the steps in the defect detection method of the flexible circuit board described in any one of the above when executing the program.

[0034] The present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, the steps in the defect detection method of the flexible circuit board described in any one of the above can be implemented.

[0035] The beneficial effects of adopting the above embodiments are:

[0036] The present invention provides a defect detection method for a flexible circuit board. First, an image of the flexible circuit board after the conductive layer covers the insulating layer is obtained as the acquisition image, then the acquisition image is preprocessed to obtain an optimized image. After that, the positions of multiple preset holes in the optimized image are identified based on a preset recognition model, and the image of each preset hole is intercepted to obtain a hole image. Finally, according to the area size relationship among the first color, the second color, the third color, and the fourth color in the hole image, it is detected whether the conductive layer and the insulating layer are aligned, where the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color presented after the insulating layer covers the conductive layer, and the fourth color is the background color when the acquisition image is captured. Compared with the prior art, the present invention realizes the automatic detection of hole alignment based on the idea of using image recognition technology to locate holes and obtain hole images, and then detecting the hole images. And when detecting alignment, it determines whether the holes are aligned by comparing the area size relationship among the first color, the second color, the third color, and the fourth color, achieving the purpose of rapid detection through simple comparison of color ratios rather than complex models. The entire defect detection process can not only run automatically but also does not affect the processing efficiency of the flexible circuit board, enabling it to be actually applied in production, perfectly solving the problem of how to automatically detect whether the holes are aligned before bonding the insulating layer and the conductive layer, and having good practicability. Description of the Drawings

[0037] Figure 1Flow chart of a method for defect detection of a flexible printed circuit board provided by the present invention;

[0038] Figure 2 Color schematic diagram of a preset hole in the present invention;

[0039] Figure 3 is Figure 1 Flow chart of step S103 in;

[0040] Figure 4 Structural schematic diagram of an embodiment of a defect detection device for a flexible printed circuit board provided by the present invention;

[0041] Figure 5 Structural schematic diagram of an electronic device provided by the present invention. Detailed implementation manners

[0042] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0043] Combined with Figure 1 As shown, a specific embodiment of the present invention discloses a method for defect detection of a flexible printed circuit board, including:

[0044] S101. Obtain a captured image, which is an image of the flexible printed circuit board after the conductive layer covers the insulating layer, captured perpendicular to the plane where the flexible printed circuit board is located;

[0045] S102. Preprocess the captured image to obtain an optimized image;

[0046] S103. Based on a preset recognition model, identify the positions of multiple preset holes in the optimized image, and intercept the image of each preset hole to obtain a hole image;

[0047] S104. According to the area size relationship among the first color, the second color, the third color, and the fourth color in the hole image, detect whether the conductive layer and the insulating layer are aligned, where the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color presented after the insulating layer covers the conductive layer, and the fourth color is the background color when the captured image is taken.

[0048] It can be understood that the conductive layer in the present invention can be understood as the conductive layer itself, or can be understood as the set of all layers below the insulating layer in the flexible printed circuit board including the conductive layer, that is, the conductive layer plus the substrate layer, etc. Which layers the conductive layer specifically refers to can be flexibly specified according to the actual process sequence, and only need to ensure that when the image is captured, the conductive layer is below the insulating layer (that is, on the side of the insulating layer away from the imaging device).

[0049] Compared with the prior art, the present invention realizes the automatic detection of hole alignment based on the idea of using image recognition technology to locate holes and obtain hole images, and then detecting the hole images. And when detecting the alignment, it determines whether the holes are aligned by comparing the area size relationship among the first color, the second color, the third color and the fourth color. By simply comparing the color ratios rather than a complex model, the purpose of rapid detection is achieved, making the entire defect detection process not only able to run automatically, but also not affect the processing efficiency of the flexible printed circuit board, enabling it to be actually applied in production, perfectly solving the problem of how to automatically detect whether the holes are aligned before bonding the insulating layer and the conductive layer, and having good practicality.

[0050] In practice, for the convenience of observing the circuit, the insulating layer generally has a certain transparency. It can be understood that when the insulating layer covers the conductive layer, if the preset holes opened on the insulating layer and the conductive layer are not aligned, then there will be a situation as Figure 2 shown. At this time, the conductive layer covering the insulating layer will present the third color, and the middle part of the preset hole will directly present the color of the background on which the conductive layer is placed, that is, the fourth color (such as the color of the workbench, or when the conductive layer is placed on a light-emitting platform, the fourth color will present the color of the light source, such as white). At the same time, there will be a dislocation at the unaligned position in the preset holes of the two layers. At this time, the insulating layer not covering the conductive layer will directly present its own color, that is, the first color. Similarly, the conductive layer not covered by the insulating layer will also present its own color, that is, the second color.

[0051] It can be understood that if the alignment effect of the preset holes is good, at this time, the area of the fourth color will be relatively close to the area of the preset hole itself, and the area of the third color will not show a large deviation. At the same time, the areas of the first color and the second color will be small and the areas of the two will be the same. Based on the above rules, it can be directly judged whether the preset holes are aligned. In the process of computer image processing, the above content can be realized by directly counting the number of pixels of each color and then making a simple comparison. Compared with other recognition algorithms or models, obviously the execution speed of the above process is extremely fast. During the actual production process, it can be detected in real time as the operator adjusts the position of the insulating layer, and a prompt is issued when the preset holes are aligned, so as to perform adhesion, improving the efficiency of defect detection and having high practical value.

[0052] For the above-mentioned step S101 of acquiring a captured image, the captured image is taken perpendicular to the plane where the flexible printed circuit board is located, and it is an image of the conductive layer covering the insulating layer in the flexible printed circuit board, which can be implemented in any way. For example, a high-precision camera can be added to the current workbench for laying the insulating layer to obtain the captured image, or a workbench that can emit light can also be designed, and a high-precision camera is also set above the workbench for alignment detection specifically.

[0053] Furthermore, for the above-mentioned step S102 of preprocessing the captured image to obtain an optimized image, it can also be implemented in any existing way. For example, the preprocessing methods include but are not limited to: grayscale conversion, filtering, image enhancement, geometric transformation, etc. Grayscale conversion is the process of converting a color image into a grayscale image, which simplifies image analysis by removing color information. Filtering techniques are used to remove noise or detail information in the image. Common filters include Gaussian filters and median filters. The Gaussian filter smooths the image by weighted averaging of neighboring pixels and is suitable for smoothing and removing high-frequency noise; while the median filter is used to remove low-frequency noise, etc. Image enhancement techniques include histogram equalization and contrast enhancement, which are used to enhance the visual effect and information content of the image. Histogram equalization enhances the image contrast and brightness by adjusting the distribution of image pixels. Geometric transformation includes operations such as rotation, scaling, and translation, which are used to correct the angle and size of the image to meet the subsequent analysis requirements in a specific embodiment of the present invention.

[0054] Furthermore, as shown in Figure 3 In a preferred embodiment, the above-mentioned step S103 of identifying the positions of multiple preset holes in the optimized image based on a preset recognition model and intercepting the image of each preset hole to obtain a hole image specifically includes:

[0055] S301: Divide multiple grids of preset specifications in the optimized image;

[0056] S302: Identify the probability of the existence of a preset hole in each grid based on the preset recognition model;

[0057] S303: Obtain the positions of multiple preset holes according to the probability of the existence of a preset hole in each grid and intercept the image of each preset hole.

[0058] The preset recognition model in the above process can be implemented by any existing model such as the R-CNN neural network model, the YOLO neural network model, etc. In this embodiment, dividing multiple grids of preset specifications in the optimized image has the advantage that it can judge each grid separately. On the one hand, it can reduce the complexity of the preset recognition model (because the number of input data for each recognition is significantly reduced), improve the efficiency and parallelism of the algorithm, and at the same time ensure the flexibility and accuracy of the algorithm.

[0059] Further, in a preferred embodiment, the area of the grid is smaller than the area of the smallest preset hole. After calculating the probability of the existence of a preset hole in each grid, grids with relatively high probabilities and adjacent to each other can be spliced to locate the preset hole.

[0060] The significance of the above limitation is to ensure that one grid is only used to identify one preset hole, thereby guaranteeing the accuracy of the algorithm and avoiding adverse phenomena such as calculation errors caused by the existence of two or more preset holes in one grid.

[0061] Further, in a preferred embodiment, the preset recognition model is a convolutional neural network model. The preset recognition model includes an input layer, at least one convolutional layer, at least one fully connected layer, and an output layer connected in sequence. The number of nodes in the input layer is equal to the number of pixels in the optimized image. Each node in the input layer is respectively used to input the three color channel values of one pixel in the optimized image. The number of nodes in the output layer is equal to the number of grids in the optimized image. Each node in the output layer is respectively used to output a probability description vector corresponding to a grid. The elements of the probability description vector respectively include: the confidence probability of at least one detection frame where the preset hole exists in this grid (equivalent to the probability of the existence of a certain specific type of preset hole in each grid), the center coordinates of at least one detection frame where the preset hole exists in this grid, and the size of at least one detection frame where the preset hole exists in this grid.

[0062] The structure of the above preset recognition model is actually a refined improvement of the existing YOLO neural network model. Specifically, because the application scenario in this embodiment is single, the input image is a simple flexible circuit board covered with an insulating layer, and the image content basically only includes a large number of colors and holes. Therefore, in this embodiment, the pooling layer of the existing YOLO neural network can be omitted while still ensuring the accuracy, and at the same time, the operation efficiency of the algorithm can be improved.

[0063] Meanwhile, and most importantly, this embodiment greatly omits the data content that the model needs to output. In the prior art, the vectors output by the YOLO neural network need to include the probabilities of various recognized objects. However, in the application scenario of this embodiment, since the number of types of preset holes in a flexible printed circuit board is limited, this embodiment omits the probabilities of recognized objects in the output vectors and represents the probabilities of predicted holes of different specifications only through detection boxes. In the YOLO neural network, a "bounding box" usually refers to the boundary box predicted by the neural network, which is used to identify the position and size of the target object detected in the input image. Therefore, the method of representing the probabilities of predicted holes of different specifications through detection boxes in this embodiment, on the one hand, omits the elements of the probabilities of various recognized objects in the probability description vector as described above, simplifies the vector length, reduces the amount of data that needs to be predicted, greatly reduces the amount of data that the preset recognition model needs to consider, and improves the efficiency. On the other hand, it also improves the recognition accuracy of the entire model due to the increase in detection boxes.

[0064] In addition, the structure of the preset recognition model in this embodiment, combined with the limitation that the area of the grid described above is smaller than the area of the smallest preset hole, further improves the accuracy, avoids inaccurate situations caused by the simultaneous presence of two holes in one grid, and overcomes the defects of the existing YOLO models (especially the YOLOv1 model).

[0065] Further, in a preferred embodiment, the above step S303, obtaining the positions of multiple preset holes and cropping the images of each preset hole according to the probabilities of the existence of preset holes in each grid, specifically includes:

[0066] Obtaining multiple target detection boxes with the highest confidence probabilities and no overlap according to the multiple probability description vectors output by the preset recognition model;

[0067] Cropping the positions where the multiple target detection boxes are located from the optimized image to obtain the images of multiple preset holes.

[0068] The above process can be implemented by using the existing non-maximum suppression method to obtain multiple target detection boxes. The advantage of the above process is that the target detection boxes are directly used as the cropping ranges of each recognized preset hole, omitting the process of determining the cropping range, and at the same time ensuring the flexibility and accuracy of cropping for preset holes of different specifications. In addition, the above target detection boxes will also be directly used for the calculation of the subsequent hole overlap evaluation value, providing a data reference basis for it.

[0069] Further, in a preferred embodiment, the above step S104, detecting whether the conductive layer and the insulating layer are aligned according to the area size relationship between the first color, the second color, the third color, and the fourth color in the hole image, specifically includes:

[0070] Based on the area size relationship between the first color and the second color, and the area ratio relationship between the third color and the fourth color in each hole image, a hole overlap evaluation value of each hole image is obtained, where the hole overlap evaluation value is used to characterize the overlap degree between the holes opened on the conductive layer and the overlap on the insulating layer in the hole image;

[0071] If the hole overlap evaluation values corresponding to all hole images do not exceed the preset standard threshold, it is determined that the conductive layer and the insulating layer are aligned.

[0072] The above process judges the hole overlap from two dimensions: the area size relationship between the first color and the second color, and the area ratio relationship between the third color and the fourth color, so as to ensure the recognition accuracy and avoid the interference caused by phenomena such as shadows.

[0073] Specifically, in a preferred embodiment, the hole overlap evaluation value is calculated by the following formula:

[0074] ;

[0075] Wherein, represents the hole overlap evaluation value, , , and are respectively the areas of the first color, the second color, the third color and the fourth color in the hole image, is the area of the preset hole corresponding to the type of the hole image, is the area of the target detection frame corresponding to the hole image, is a preset hyperparameter, and are different preset weights respectively.

[0076] The significance of the above formula is that the difference between the first color and the second color is measured by the first item, the ratio relationship between the third color and the fourth color is measured by the second item, and the area of the target detection frame and the area of the preset hole are introduced as references for standard data, making it more accurate. The result is ensured to be an integer through the preset hyperparameter, and the influence degree is controlled through the preset weight. The preset weight and the preset hyperparameter can be obtained by any existing method such as experiments.

[0077] Combined with Figure 4 shown, the present invention also provides a defect detection device for a flexible printed circuit board, including:

[0078] An image acquisition module 410, configured to acquire an acquisition image, where the acquisition image is an image of the flexible printed circuit board after the conductive layer covers the insulating layer, taken perpendicular to the plane where the flexible printed circuit board is located;

[0079] An image optimization module 420, configured to preprocess the acquired image to obtain an optimized image;

[0080] A hole recognition module 430, configured to recognize the positions of multiple preset holes in the optimized image based on a preset recognition model, and intercept the image of each preset hole to obtain a hole image;

[0081] A hole detection module 440, configured to detect whether the conductive layer and the insulating layer are aligned according to the area size relationship among a first color, a second color, a third color, and a fourth color in the hole image, where the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color presented after the insulating layer covers the conductive layer, and the fourth color is the background color when the acquired image is captured.

[0082] It should be noted here that: the corresponding system provided in the above embodiment can implement the technical solutions described in the above method embodiments. The specific implementation principles of the above modules or units can refer to the corresponding content in the above method embodiments, which will not be elaborated here.

[0083] Reference Figure 5 , which shows a schematic structural diagram of an electronic device provided in an embodiment of the present invention. In this embodiment, the electronic device includes:

[0084] A memory 510 and a processor 520;

[0085] The memory 510 is used to store a program, and the processor 520 is configured to execute the defect detection method of the flexible circuit board in the above embodiment when executing the program.

[0086] This embodiment also provides a computer-readable storage medium, on which a defect detection program of a flexible circuit board is stored. When the defect detection program of the flexible circuit board is executed by a processor, the steps in the above embodiment can be implemented.

[0087] The present invention provides a method for defect detection of a flexible printed circuit board. First, an image of the flexible printed circuit board after the conductive layer covers the insulating layer is obtained as a captured image. Then, the captured image is preprocessed to obtain an optimized image. After that, based on a preset recognition model, the positions of multiple preset holes in the optimized image are recognized, and an image of each preset hole is intercepted to obtain a hole image. Finally, according to the area size relationship among the first color, the second color, the third color, and the fourth color in the hole image, it is detected whether the conductive layer and the insulating layer are aligned, where the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color presented after the insulating layer covers the conductive layer, and the fourth color is the background color when the captured image is taken. Compared with the prior art, the present invention realizes the automatic detection of hole alignment based on the idea of using image recognition technology to locate holes and obtain hole images, and then detecting the hole images. When detecting alignment, it determines whether the holes are aligned by comparing the area size relationship among the first color, the second color, the third color, and the fourth color, and achieves the purpose of rapid detection through simple comparison of color ratios rather than complex models. The entire defect detection process can not only run automatically but also does not affect the processing efficiency of the flexible printed circuit board, enabling it to be actually applied in production, perfectly solving the problem of how to automatically detect whether the holes are aligned before bonding the insulating layer and the conductive layer, and having good practicability.

[0088] It should be noted that the embodiments in this specification are all described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other.

[0089] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A defect detection method for a flexible circuit board, characterized in that: include: Acquire a captured image, where the captured image is taken perpendicular to the plane where the flexible circuit board is located, and is an image of the flexible circuit board after the conductive layer covers the insulating layer; Preprocessing the collected image to obtain an optimized image; Based on a preset recognition model, positions of a plurality of preset holes in the optimization image are identified, and an image of each preset hole is intercepted to obtain a hole image; According to the area size relationship between the first color, the second color, the third color and the fourth color in the hole image, it is detected whether the conductive layer and the insulating layer are aligned, wherein the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color superimposed after the insulating layer covers the conductive layer, and the fourth color is the background color when the image is captured.

2. The defect detection method for a flexible circuit board according to claim 1, characterized in that: The positions of multiple preset holes in the optimized image are identified based on a preset recognition model, and an image of each preset hole is intercepted to obtain a hole image, including: Divide the optimized image into multiple grids of preset sizes; Identify the probability of a preset hole existing in each grid based on a preset recognition model; According to the probability of the existence of a preset hole in each grid, the positions of a plurality of preset holes are obtained and an image of each preset hole is captured.

3. The defect detection method of the flexible circuit board according to claim 2, characterized in that: The area of ​​the grid is smaller than the area of ​​the smallest preset hole.

4. The defect detection method of the flexible circuit board according to claim 3, characterized in that: The preset recognition model is a convolutional neural network model, which includes an input layer, at least one convolutional layer, at least one fully connected layer and an output layer connected in sequence, wherein the number of nodes in the input layer is equal to the number of pixels in the optimized image, and each node in the input layer is used to input the three color channel values ​​of a pixel in the optimized image, the number of nodes in the output layer is equal to the number of grids in the optimized image, and each node in the output layer is used to output a probability description vector corresponding to a grid, and the elements of the probability description vector include: the confidence probability of at least one detection box where a preset hole exists in the grid, the center coordinates of at least one detection box where a preset hole exists in the grid, and the size of at least one detection box where a preset hole exists in the grid.

5. The defect detection method for a flexible circuit board according to claim 4, characterized in that: According to the probability of the existence of a preset hole in each grid, the positions of multiple preset holes are obtained and the image of each preset hole is intercepted, including: According to the multiple probability description vectors output by the preset recognition model, multiple non-overlapping target detection frames with the highest confidence probability are obtained; The positions of multiple target detection frames are captured from the optimized image to obtain images of multiple preset holes.

6. The defect detection method of the flexible circuit board according to claim 5, characterized in that: According to the area size relationship among the first color, the second color, the third color and the fourth color in the hole image, detecting whether the conductive layer and the insulating layer are aligned includes: According to the area size relationship between the first color and the second color, and the area ratio relationship between the third color and the fourth color in each hole image, a hole overlap evaluation value of each hole image is obtained, wherein the hole overlap evaluation value is used to characterize the degree of overlap between the holes opened on the conductive layer and the holes opened on the insulating layer in the hole image; If the hole overlap evaluation values ​​corresponding to all hole images do not exceed the preset standard threshold, it is determined that the conductive layer and the insulating layer are aligned.

7. The defect detection method of the flexible circuit board according to claim 6, characterized in that: The hole overlap evaluation value is calculated by the following formula: ； in, represents the hole overlap evaluation value, , , and The areas of the first color, second color, third color and fourth color in the hole image respectively, is the area of ​​the preset hole corresponding to the type of hole image, is the area of ​​the target detection box corresponding to the hole image, To preset hyperparameters, and They have different preset weights.

8. A defect detection device for a flexible circuit board, characterized in that: include: An image acquisition module is used to acquire an image, where the image is taken perpendicular to the plane where the flexible circuit board is located, and is an image of the flexible circuit board after the conductive layer covers the insulating layer; An image optimization module is used to preprocess the collected image to obtain an optimized image; A hole recognition module, used to recognize the positions of multiple preset holes in the optimized image based on a preset recognition model, and intercept the image of each preset hole to obtain a hole image; The hole detection module is used to detect whether the conductive layer and the insulating layer are aligned according to the area size relationship between the first color, the second color, the third color and the fourth color in the hole image, wherein the first color is the color of the insulating layer, the second color is the color of the conductive layer, the third color is the color superimposed after the insulating layer covers the conductive layer, and the fourth color is the background color when the image is captured.

9. An electronic device, characterized in that: include: Memory and processor; The memory is used to store a program, and the processor is used to execute the steps of the defect detection method for a flexible circuit board according to any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps of the defect detection method for a flexible circuit board as described in any one of claims 1 to 7.

Images