PCB (Printed Circuit Board) back drilling defect detection process

Abstract

The invention discloses a PCB (Printed Circuit Board) back drilling defect detection process, which relates to the technical field of defect detection, and comprises the following steps: placing a PCB to be detected on a detection table; wherein the PCB to be detected is subjected to back drilling processing; obtaining distribution of through holes of the PCB to be detected and positions and directions of back drilling holes; according to the scanning direction, the scanning visual angle and the scanning base point of the laser scanner, point cloud data of the back drilling face of the PCB to be detected and point cloud data of the through hole face of the PCB to be detected are collected; performing depth map conversion on the point cloud data of the back drilling surface of the PCB to be detected and the point cloud data of the through hole surface of the PCB to be detected to obtain a back drilling depth image of the back drilling surface and a through hole depth image of the through hole surface; and according to the direction of the back drilling hole, the back drilling depth image of the back drilling surface and the through hole depth image of the through hole surface, identifying the defect of the back drilling hole of the PCB to be detected. According to the method and the device, the defect detection efficiency of the back drill of the PCB is improved.

Description

Technical Field

[0001] The present application relates to the technical field of defect detection, and in particular to a PCB back-drilling defect detection process. Background Art

[0002] After electronic equipment adopts PCB boards, due to the consistency of similar printed boards, manual wiring errors are avoided, and electronic components can be automatically inserted or mounted, automatically soldered, and automatically tested, ensuring the quality of electronic equipment, improving labor productivity, reducing costs, and facilitating maintenance. PCB boards have developed from single-layer to multi-layer, and still maintain their respective development trends. Due to the continuous development towards high precision, high density and high reliability, the continuous reduction in size, reduction in cost, and improvement in performance, the printed circuit board still maintains a strong vitality in the development of future electronic equipment projects.

[0003] In multi-layer PCBs, signals are usually transmitted from one layer to another through through holes. These through holes will form a copper layer on the hole wall during the electroplating process, forming a complete conductive path. When the signal is transmitted through the through hole, if the through hole is not properly cut off, the part of the through hole on the layer that does not need to be connected may cause signal reflection and interference, affecting the integrity of the signal. Back drilling is a secondary processing of the through hole that has been completed with the first drilling and electroplating during the PCB manufacturing process, drilling out the unnecessary part of the through hole from the back of the PCB board.

[0004] At present, the back drilling hole detection is usually carried out by measuring the back drilling hole slices. However, this method requires manual polishing and measurement. When processing multi-layer printed circuit boards with a large number of layers, there is a possibility of errors and omissions in manual layer counting, which affects the results of the back drilling hole detection.

[0005] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0006] The embodiment of the present application provides a PCB back drilling defect detection process to solve the above technical problems.

[0007] The present application provides a PCB back drilling defect detection process, including:

[0008] Placing the PCB board to be tested on the testing table; wherein the PCB board to be tested has been back-drilled;

[0009] Obtaining the distribution of through holes of the PCB board to be tested and the position and direction of the back drilling holes;

[0010] Determine the scanning strategy of the laser scanner according to the distribution of the through holes of the PCB board to be inspected and the position and direction of the back-drilled holes; wherein the scanning strategy includes a scanning direction, a scanning angle and a scanning base point;

[0011] According to the scanning direction, the scanning angle and the scanning base point of the laser scanner, the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected are collected;

[0012] Performing depth map conversion on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface;

[0013] According to the direction of the back-drilled hole, the back-drilled depth image of the back-drilled surface, and the through-hole depth image of the through-hole surface, defects of the back-drilled hole of the PCB board to be inspected are identified.

[0014] Further, the point cloud data of the back-drilled surface of the PCB board to be detected and the point cloud data of the through-hole surface of the PCB board to be detected are converted into a depth map to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface, including:

[0015] Density clustering is performed on the point cloud data of the through-hole surface to obtain a first point cloud cluster;

[0016] Performing density clustering on the point cloud data of the back-drilled surface to obtain a second point cloud cluster;

[0017] Correcting the scanning direction, the scanning angle, and the scanning base point according to the point cloud density of the first point cloud cluster;

[0018] Determine the elevation value plane based on the corrected scanning base point;

[0019] According to the corrected scanning direction, the corrected scanning angle and the corrected scanning base point, the corrected point cloud data of the back drilling surface under the elevation value plane is collected;

[0020] Performing density clustering on the corrected point cloud data to obtain a third point cloud cluster;

[0021] performing density correction on the second point cloud cluster based on the point cloud density of the third point cloud cluster and the point cloud density of the first point cloud cluster to obtain a density-corrected second point cloud cluster;

[0022] A back-drilling depth image of the back-drilling surface and a through-hole depth image of the through-hole surface are determined according to the first point cloud cluster and the density-corrected second point cloud cluster.

[0023] Furthermore, density correction is performed on the second point cloud cluster based on the point cloud density of the third point cloud cluster and the point cloud density of the first point cloud cluster to obtain a density-corrected second point cloud cluster.

[0024] Further, the determining the back-drilling depth image of the back-drilling surface and the through-hole depth image of the through-hole surface according to the first point cloud cluster and the density-corrected second point cloud cluster includes:

[0025] Performing plane fitting on each point cloud cluster in the first point cloud clusters to obtain a first plane curvature of the fitting plane of each point cloud cluster;

[0026] Performing plane fitting on each point cloud cluster in the density-corrected second point cloud cluster to obtain a second plane curvature of the fitting plane of each point cloud cluster;

[0027] Generate a first affine transformation matrix from the first point cloud cluster to the through-hole depth image according to the first plane curvature;

[0028] generating a second affine transformation matrix from a density-corrected second point cloud cluster to the backdrilling depth image according to the curvature of the second plane;

[0029] Based on the first affine transformation matrix and the second affine transformation matrix, the first point cloud cluster is mapped to the backdrilling depth image, and the density-corrected second point cloud cluster is mapped to the through-hole depth image.

[0030] Further, the defects of the back-drilled holes include hole eccentricity, back-drilled hole leakage, through-hole leakage, hole diameter abnormality, foreign matter in the hole, and hole intersection; wherein the hole eccentricity is used to characterize the deviation between the actual center position of the hole and the designed center position; the back-drilled hole leakage is used to characterize the phenomenon that the hole does not reach a predetermined depth; the through-hole leakage is used to characterize the phenomenon that the hole is not drilled through; the back-drilled hole defects of the PCB board to be inspected are identified according to the direction of the back-drilled hole, the back-drilled depth image of the back-drilled surface, and the through-hole depth image of the through-hole surface, including:

[0031] The back-drilling direction, the back-drilling depth image and the through-hole depth image are input into a trained and optimized back-drilling defect detection model to obtain an output of the back-drilling defect of the PCB board to be detected; wherein the back-drilling defect detection model is trained and optimized based on the following steps:

[0032] Acquire backdrill test depth image to through hole test depth image;

[0033] filtering the backdrill test depth image and the through-hole test depth image;

[0034] Performing pixel annotation on the filtered backdrill test depth image and the filtered through-hole test depth image; generating a bounding box and defect label attribute file containing defects contained in the image after pixel annotation on several of the images;

[0035] According to the defect label attribute file and the annotated bounding box, output the center point coordinates, length, width and rotation angle of the bounding box to construct a training set and a validation set;

[0036] Introduced an image gradient-based attention mechanism module and built a deep learning model based on the UNet model;

[0037] The constructed deep learning model is trained and verified using a training set and a validation set to obtain the back drilling defect detection model.

[0038] Furthermore, the first affine transformation matrix is ​​used to indicate the transformation of the first point cloud cluster to the through-hole depth image; and the second affine transformation matrix is ​​used to indicate the transformation of the density-corrected second point cloud cluster to the backdrilling depth image.

[0039] Furthermore, when collecting the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected, the scanning frequency, scanning angle resolution, number of scanning layers, laser wavelength and scanning speed of the laser scanner are adjusted according to the inspection environment conditions.

[0040] Furthermore, if the PCB board to be inspected has undergone back drilling processing on the upper surface and back drilling processing on the lower surface, the back drilled holes on the upper surface and the back drilled holes on the lower surface of the PCB board to be inspected are inspected simultaneously.

[0041] Based on the embodiment provided by the present application, a PCB board to be inspected is placed on a detection table; wherein the PCB board to be inspected is back-drilled; the distribution of through holes of the PCB board to be inspected and the position and direction of the back-drilled holes are obtained; according to the distribution of through holes of the PCB board to be inspected and the position and direction of the back-drilled holes, a scanning strategy of a laser scanner is determined; wherein the scanning strategy includes a scanning direction, a scanning angle and a scanning base point; according to the scanning direction, the scanning angle and the scanning base point of the laser scanner, point cloud data of a back-drilled surface of the PCB board to be inspected and point cloud data of a through-hole surface of the PCB board to be inspected are collected; depth map conversion is performed on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilling depth image of the back-drilled surface and a through-hole depth image of the through-hole surface; according to the direction of the back-drilling holes, the back-drilling depth image of the back-drilled surface and the through-hole depth image of the through-hole surface, defects of the back-drilled holes of the PCB board to be inspected are identified. Thus, it is possible to quickly and accurately detect the defects and performance indicators of the circuit board; there is no need to cut or damage the board to be tested, nor is there any need to manually count the layers, which is conducive to reducing the detection cost and improving the detection efficiency. In other words, this application realizes efficient and automated detection of the back drilling quality of the PCB board, reduces the labor intensity and subjective errors of manual detection, and improves production quality and efficiency. Through precise scanning strategies and deep image analysis, problems in the production process can be discovered and corrected in a timely manner to ensure the quality and reliability of the PCB board. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described herein are used to provide a further understanding of the embodiments of the present invention and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0043] Figure 1 It is a flow chart of an optional PCB back drilling defect detection process according to an embodiment of the present application;

[0044] Figure 2 The present invention is a flowchart of another optional PCB back drilling defect detection process according to an embodiment of the present application.

[0045] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0046] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0047] Alternatively, if Figure 1 As shown, the present application provides a PCB back drilling defect detection process, comprising:

[0048] S101, placing a PCB board to be tested on a testing table; wherein the PCB board to be tested is back-drilled;

[0049] S102, obtaining the distribution of through holes of the PCB board to be inspected and the position and direction of the back drilling holes;

[0050] S103, determining a scanning strategy of a laser scanner according to the distribution of through holes of the PCB board to be inspected and the position and direction of the back-drilled holes; wherein the scanning strategy includes a scanning direction, a scanning angle and a scanning base point;

[0051] S104, collecting point cloud data of the back-drilled surface of the PCB board to be inspected and point cloud data of the through-hole surface of the PCB board to be inspected according to the scanning direction, scanning angle and scanning base point of the laser scanner;

[0052] S105, performing depth map conversion on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface;

[0053] S106, identifying defects of the back-drilled holes of the PCB board to be inspected according to the direction of the back-drilled holes, the back-drilled depth image of the back-drilled surface, and the through-hole depth image of the through-hole surface.

[0054] Based on the embodiment provided by the present application, a PCB board to be inspected is placed on a detection table; wherein the PCB board to be inspected is back-drilled; the distribution of through holes of the PCB board to be inspected and the position and direction of the back-drilled holes are obtained; according to the distribution of through holes of the PCB board to be inspected and the position and direction of the back-drilled holes, a scanning strategy of a laser scanner is determined; wherein the scanning strategy includes a scanning direction, a scanning angle and a scanning base point; according to the scanning direction, the scanning angle and the scanning base point of the laser scanner, point cloud data of a back-drilled surface of the PCB board to be inspected and point cloud data of a through-hole surface of the PCB board to be inspected are collected; depth map conversion is performed on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilling depth image of the back-drilled surface and a through-hole depth image of the through-hole surface; according to the direction of the back-drilling holes, the back-drilling depth image of the back-drilled surface and the through-hole depth image of the through-hole surface, defects of the back-drilled holes of the PCB board to be inspected are identified. Thus, it is possible to quickly and accurately detect the defects and performance indicators of the circuit board; there is no need to cut or damage the board to be tested, nor is there any need to manually count the layers, which is conducive to reducing the detection cost and improving the detection efficiency. In other words, this application realizes efficient and automated detection of the back drilling quality of the PCB board, reduces the labor intensity and subjective errors of manual detection, and improves production quality and efficiency. Through precise scanning strategies and deep image analysis, problems in the production process can be discovered and corrected in a timely manner to ensure the quality and reliability of the PCB board.

[0055] Further, the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected are converted into a depth map to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface, including:

[0056] Density clustering is performed on the point cloud data of the through-hole surface to obtain the first point cloud cluster;

[0057] Density clustering is performed on the point cloud data of the back-drilled surface to obtain a second point cloud cluster;

[0058] According to the point cloud density of the first point cloud cluster, the scanning direction, scanning angle and scanning base point are corrected;

[0059] Determine the elevation value plane based on the corrected scanning base point;

[0060] According to the corrected scanning direction, the corrected scanning angle and the corrected scanning base point, the corrected point cloud data of the back drilling surface under the elevation value plane is collected;

[0061] Density clustering is performed on the corrected point cloud data to obtain the third point cloud cluster;

[0062] Density correction is performed on the second point cloud cluster based on the point cloud density of the third point cloud cluster and the point cloud density of the first point cloud cluster to obtain a density-corrected second point cloud cluster;

[0063] A back-drilling depth image of the back-drilling surface and a through-hole depth image of the through-hole surface are determined according to the first point cloud cluster and the density-corrected second point cloud cluster.

[0064] Based on the embodiments provided in the present application, through cluster analysis, the point cloud data of the through-hole surface can be divided into different clusters according to density, which is helpful to identify and distinguish different feature areas and provide an accurate data basis for subsequent depth image generation; by analyzing the point cloud density, the scanning strategy can be optimized to ensure that the scanning process more accurately captures the features of the through-hole surface and improves the quality and efficiency of data acquisition; collecting the corrected point cloud data can further improve the recognition accuracy of the back-drilled surface features and provide more accurate data for the generation of depth images; clustering the corrected point cloud data again can help further refine and optimize the feature recognition of the back-drilled surface and ensure the accuracy of the depth image; by combining the point cloud density information of the back-drilled surface and the through-hole surface, the second point cloud cluster can be accurately corrected for density, thereby improving the accuracy of defect detection.

[0065] Further, density correction is performed on the second point cloud cluster based on the point cloud density of the third point cloud cluster and the point cloud density of the first point cloud cluster to obtain a density-corrected second point cloud cluster, including:

[0066]

[0067] in, is the point cloud density of the second point cloud cluster after density correction; is the point cloud density of the second point cloud cluster; is the point cloud density of the first point cloud cluster; is the point cloud density of the third point cloud cluster; h1 is the elevation value corresponding to the corrected point cloud data; h0 is the elevation value of the standard scanning area.

[0068] Alternatively, if Figure 2 As shown, according to the first point cloud cluster and the density-corrected second point cloud cluster, a back-drilling depth image of the back-drilling surface and a through-hole depth image of the through-hole surface are determined, including:

[0069] S201, performing plane fitting on each point cloud cluster in the first point cloud cluster to obtain a first plane curvature of the fitting plane of each point cloud cluster;

[0070] S202, performing plane fitting on each point cloud cluster in the density-corrected second point cloud cluster to obtain a second plane curvature of the fitting plane of each point cloud cluster;

[0071] S203, generating a first affine transformation matrix from the first point cloud cluster to the through-hole depth image according to the first plane curvature;

[0072] S204, generating a second affine transformation matrix from a density-corrected second point cloud cluster to a back-drilling depth image according to the second plane curvature;

[0073] S205 , based on the first affine transformation matrix and the second affine transformation matrix, mapping the first point cloud cluster to the backdrilling depth image, and mapping the density-corrected second point cloud cluster to the through-hole depth image.

[0074] Furthermore, the defects of back-drilled holes include hole eccentricity, back-drilled hole leakage, through-hole leakage, hole diameter abnormality, foreign matter in the hole, and hole intersection; wherein, hole eccentricity is used to characterize the deviation between the actual center position of the hole and the designed center position; back-drilled hole leakage is used to characterize the phenomenon that the hole does not reach the predetermined depth; through-hole leakage is used to characterize the phenomenon that the hole is not drilled through; according to the direction of the back-drilled hole, the back-drilled depth image of the back-drilled surface, and the through-hole depth image of the through-hole surface, the defects of the back-drilled holes of the PCB board to be inspected are identified, including:

[0075] The back drilling direction, the back drilling depth image and the through hole depth image are input into the trained and optimized back drilling defect detection model to obtain the output of the back drilling defect of the PCB board to be detected; wherein the back drilling defect detection model is trained and optimized based on the following steps:

[0076] Acquire backdrill test depth image to through hole test depth image;

[0077] Filtering the backdrill test depth image and the through-hole test depth image;

[0078]

[0079] Where H1(x1,y1) represents the filtered backdrill test depth image; m1 and n1 represent the size of the first filter; Center1 xy Indicates that the center of the first filter is at point (x1, y1); O1(s1, t1) represents the input backdrill test depth image;

[0080] Performing pixel annotation on the filtered backdrill test depth image and the filtered through-hole test depth image; generating a bounding box and defect label attribute file containing defects contained in the image after pixel annotation on several of the images;

[0081] According to the defect label attribute file and the annotated bounding box, the center point coordinates, length, width and rotation angle of the bounding box are output to build the training set and validation set;

[0082] Introduced an image gradient-based attention mechanism module and built a deep learning model based on the UNet model;

[0083] The constructed deep learning model is trained and verified using the training set and validation set to obtain the back drilling defect detection model.

[0084] Based on the embodiments provided in the present application, through a deep learning model, back-drilling defects can be automatically identified and classified to improve detection efficiency and accuracy; filtering helps to remove image noise, enhance image features, and provide clearer data for subsequent image analysis and defect identification; pixel annotation provides an accurate training target for the model, enabling the model to learn specific defect features; the attention mechanism can help the model better focus on key areas in the image and improve the accuracy of defect detection; the encoder-decoder structure and jump connection of the UNet model help to accurately locate and identify defects.

[0085] Further,

[0086] Where H1(x1,y1) represents the filtered through-hole test depth image; m2, n2 represent the size of the second filter; Center2 xy It indicates that the center of the second filter is at the point (x2, y2); O2(s2, t2) represents the input through-hole test depth image.

[0087] Furthermore, the first affine transformation matrix is ​​used to indicate the transformation from the first point cloud cluster to the through-hole depth image; and the second affine transformation matrix is ​​used to indicate the transformation from the density-corrected second point cloud cluster to the back-drilling depth image.

[0088] Furthermore, when collecting point cloud data of the back-drilled surface of the PCB board to be inspected and point cloud data of the through-hole surface of the PCB board to be inspected, the scanning frequency, scanning angle resolution, number of scanning layers, laser wavelength and scanning speed of the laser scanner are adjusted according to the inspection environment.

[0089] Furthermore, if the PCB board to be inspected has undergone back drilling processing on the upper surface and back drilling processing on the lower surface, the back drilling holes on the upper surface and the back drilling holes on the lower surface of the PCB board to be inspected are inspected at the same time.

[0090] Further, the direction of the back-drilled hole is a direction from the side provided with the back-drilled hole to the side away from the back-drilled hole.

[0091] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A PCB back drilling defect detection process, characterized in that: include: Placing the PCB board to be tested on the testing table; wherein the PCB board to be tested has been back-drilled; Obtaining the distribution of through holes of the PCB board to be tested and the position and direction of the back drilling holes; Determine the scanning strategy of the laser scanner according to the distribution of the through holes of the PCB board to be inspected and the position and direction of the back-drilled holes; wherein the scanning strategy includes a scanning direction, a scanning angle and a scanning base point; According to the scanning direction, the scanning angle and the scanning base point of the laser scanner, the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected are collected; Performing depth map conversion on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface; According to the direction of the back-drilled hole, the back-drilled depth image of the back-drilled surface, and the through-hole depth image of the through-hole surface, defects of the back-drilled hole of the PCB board to be inspected are identified.

2. The PCB back drilling defect detection process according to claim 1, characterized in that: The step of performing depth map conversion on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface includes: Density clustering is performed on the point cloud data of the through-hole surface to obtain a first point cloud cluster; Performing density clustering on the point cloud data of the back-drilled surface to obtain a second point cloud cluster; The scanning direction, the scanning angle and the scanning base point are corrected according to the point cloud density of the first point cloud cluster.

3. The PCB back drilling defect detection process according to claim 2, characterized in that: The step of performing depth map conversion on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface further includes: Determine the elevation value plane based on the corrected scanning base point; According to the corrected scanning direction, the corrected scanning angle and the corrected scanning base point, the corrected point cloud data of the back drilling surface under the elevation value plane is collected; Density clustering is performed on the corrected point cloud data to obtain a third point cloud cluster.

4. The PCB back drilling defect detection process according to claim 3, characterized in that: The step of performing depth map conversion on the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected to obtain a back-drilled depth image of the back-drilled surface and a through-hole depth image of the through-hole surface further includes: performing density correction on the second point cloud cluster based on the point cloud density of the third point cloud cluster and the point cloud density of the first point cloud cluster to obtain a density-corrected second point cloud cluster; A back-drilling depth image of the back-drilling surface and a through-hole depth image of the through-hole surface are determined according to the first point cloud cluster and the density-corrected second point cloud cluster.

5. The PCB back drilling defect detection process according to claim 4, characterized in that: The step of determining the back-drilling depth image of the back-drilling surface and the through-hole depth image of the through-hole surface according to the first point cloud cluster and the density-corrected second point cloud cluster includes: Performing plane fitting on each point cloud cluster in the first point cloud clusters to obtain a first plane curvature of the fitting plane of each point cloud cluster; A plane fitting is performed on each point cloud cluster in the density-corrected second point cloud cluster to obtain a second plane curvature of the fitting plane of each point cloud cluster.

6. The PCB back drilling defect detection process according to claim 5, characterized in that: The determining, based on the first point cloud cluster and the density-corrected second point cloud cluster, the back-drilling depth image of the back-drilling surface and the through-hole depth image of the through-hole surface comprises: generating a first affine transformation matrix from the first point cloud cluster to the through-hole depth image based on the first plane curvature; generating a second affine transformation matrix from a density-corrected second point cloud cluster to the backdrilling depth image according to the curvature of the second plane; Based on the first affine transformation matrix and the second affine transformation matrix, the first point cloud cluster is mapped to the backdrilling depth image, and the density-corrected second point cloud cluster is mapped to the through-hole depth image.

7. The PCB back drilling defect detection process according to claim 1, characterized in that: The defects of back-drilled holes include hole eccentricity, missing back holes, missing through holes, abnormal hole diameter, foreign matter in the hole, and hole intersection; wherein the hole eccentricity is used to characterize the deviation between the actual center position of the hole and the designed center position; the missing back holes are used to characterize the phenomenon that the hole does not reach the predetermined depth; the missing through holes are used to characterize the phenomenon that the hole is not drilled through; the back-drilled hole defects of the PCB board to be inspected are identified according to the direction of the back-drilled hole, the back-drilled depth image of the back-drilled surface, and the through-hole depth image of the through-hole surface, including: The back-drilling direction, the back-drilling depth image and the through-hole depth image are input into a trained and optimized back-drilling defect detection model to obtain an output of the back-drilling defect of the PCB board to be detected; wherein the back-drilling defect detection model is trained and optimized based on the following steps: Acquire backdrill test depth image to through hole test depth image; filtering the backdrill test depth image and the through-hole test depth image; Performing pixel annotation on the filtered backdrill test depth image and the filtered through-hole test depth image; generating a bounding box and defect label attribute file containing defects contained in the image after pixel annotation on several of the images; According to the defect label attribute file and the annotated bounding box, output the center point coordinates, length, width and rotation angle of the bounding box to construct a training set and a validation set; Introduced an image gradient-based attention mechanism module and built a deep learning model based on the UNet model; The constructed deep learning model is trained and verified using a training set and a validation set to obtain the back drilling defect detection model.

8. The PCB back drilling defect detection process according to claim 7, characterized in that: The first affine transformation matrix is ​​used to indicate the transformation from the first point cloud cluster to the through-hole depth image; and the second affine transformation matrix is ​​used to indicate the transformation from the density-corrected second point cloud cluster to the back-drilling depth image.

9. The PCB back drilling defect detection process according to claim 1, characterized in that: When collecting the point cloud data of the back-drilled surface of the PCB board to be inspected and the point cloud data of the through-hole surface of the PCB board to be inspected, the scanning frequency, scanning angle resolution, number of scanning layers, laser wavelength and scanning speed of the laser scanner are adjusted according to the inspection environment.

10. The PCB back drilling defect detection process according to claim 1, characterized in that: If the PCB board to be inspected has undergone back drilling processing on the upper surface and back drilling processing on the lower surface, the back drilling holes on the upper surface and the back drilling holes on the lower surface of the PCB board to be inspected are inspected at the same time.

Images