PCB back-drilling hole defect detection method, detection device and evaluation method

Abstract

The present application belongs to the technical field of PCB defect detection, and particularly relates to a PCB back-drilling defect detection method, a detection device and an evaluation method. The method first identifies and preliminarily screens the holes to be rechecked through the two-dimensional image of the whole board; then performs three-dimensional topography scanning on the holes to be rechecked, and performs spatial registration on the two-dimensional image subgraph and the three-dimensional point cloud data matrix of the same hole through a unified coordinate system, so that the two-dimensional texture and the three-dimensional topography information are accurately corresponded; based on the fused data after registration, residual stubs, copper wire residues, dirt, microcavities, depressions, scratches and hole wall roughness are identified and classified. The present application corresponds the two-dimensional preliminary screening result and the three-dimensional measurement data through spatial registration, so that the copper wire and the dirt, the microcavity and the scratch can be distinguished based on the three-dimensional topography difference, and the size parameters such as the residual stub length and the hole wall roughness are included in the quality evaluation.

Description

Technical Field

[0001] This invention belongs to the field of PCB defect detection technology, specifically relating to PCB back-drill hole defect detection methods, detection devices, and evaluation methods. Background Technology

[0002] Back drilling is a critical process in high-speed, high-frequency printed circuit board (PCB) manufacturing. It is used to remove unwanted copper foil pillars (i.e., "residual studs") from through-holes to reduce signal reflection and attenuation, thereby improving signal integrity. The quality of back drilling directly affects the electrical performance and reliability of the final product. Its core quality indicators include the accuracy of the residual stud length, hole wall quality (such as roughness and copper wire residue), the integrity of the inner copper ring, and the presence of internal defects (such as voids, dents, and contaminants).

[0003] Currently, the detection of back-drilled holes mainly relies on the following two methods: 2D Automated Optical Inspection (2D AOI): This method acquires planar images of back-drilled holes using area array or line scan cameras and identifies defects based on the image's grayscale, color, and contour features. Its advantages include high speed, full coverage inspection capability, and the ability to effectively detect defects such as missing hole rings and obvious foreign objects. However, 2D AOI has fundamental limitations: First, it cannot measure key three-dimensional dimensions such as residual pile length and hole wall roughness; second, for defects with similar features in the two-dimensional image (such as copper wires attached to the hole wall, resin dirt of similar color, tiny voids, and scratches / depressions), it is highly prone to misjudgment or missed detection, increasing the cost and uncertainty of manual re-inspection.

[0004] Manual microscope inspection: Operators use microscopes for random inspection. This method is extremely inefficient, highly subjective, has poor repeatability, and cannot be quantitatively recorded or statistically analyzed, making it difficult to meet the needs of modern large-scale, high-quality production management.

[0005] As PCBs evolve towards higher density, higher frequency, and higher speed, the precision requirements for back-drilled holes are becoming increasingly stringent (e.g., residual hole control requirements are within ±25 micrometers or even smaller). Therefore, the industry urgently needs a fully automated inspection solution that can achieve precise measurement of the three-dimensional morphology of back-drilled holes, automatic and accurate defect classification, and quantitative analysis while ensuring inspection efficiency. Existing single 2D or traditional measurement methods can no longer meet this requirement. Summary of the Invention

[0006] The purpose of this invention is to provide a method, device and evaluation method for detecting defects in PCB back-drilled holes, so as to establish a spatial registration relationship between the two-dimensional preliminary screening results and the three-dimensional measurement data of PCB back-drilled holes, so as to realize the fusion detection of two-dimensional texture and three-dimensional morphology of the same hole position.

[0007] The present invention achieves the above objectives through the following technical solutions: Firstly, this invention proposes a method for detecting defects in PCB back-drilled holes, the method comprising: The PCB board under test is scanned by a two-dimensional imaging module to generate a two-dimensional image. All back drill holes in the two-dimensional image are identified and their coordinate positions are determined. Based on the features of the two-dimensional image, a preliminary screening is performed, and a list of coordinates of holes to be verified is output. Based on the coordinate list, each hole to be verified is sequentially moved to the 3D measurement module for scanning via a motion platform to obtain the 3D point cloud data of the hole area; the 2D image and 3D point cloud data of the same hole are spatially registered to obtain the registered fused data; The fused data includes a two-dimensional image sub-image of the hole to be verified, a three-dimensional point cloud data matrix, and the spatial correspondence between the two. The registered fused data is subjected to defect detection and classification to identify at least one defect type among residual piles exceeding tolerance, copper wire residue, dirt, micro-voids, dents, scratches, and excessive roughness.

[0008] Furthermore, the two-dimensional image features include at least one of the following: continuity of the hole ring contour, grayscale deviation value of the area inside the hole, and proportion of foreign matter area inside the hole.

[0009] Furthermore, the list of coordinates of the holes to be verified is obtained through the following steps: In the two-dimensional image, all back drill holes are identified by a circle recognition algorithm, and the center pixel coordinates of each back drill hole are determined. For each identified back-drilled hole, extract the two-dimensional image features of its hole ring region and hole interior region, and compare them with a preset threshold. A hole is marked as a verification hole if it meets at least one of the following conditions: If the continuity of the hole ring profile is lower than a preset threshold, it is determined to be a hole ring defect; If the absolute value of the difference between the gray value of the area inside the hole and the average gray value of the standard hole is greater than the preset gray value threshold, it is determined to be a foreign object inside the hole; The elliptic fit of the hole area exceeds the preset range, which is determined to be hole deformation. By using a pre-calibrated transformation matrix between pixel coordinates and platform physical coordinates, the center pixel coordinates of each hole to be verified are converted into physical coordinates in the motion platform coordinate system, thereby generating a list of coordinates for the holes to be verified.

[0010] Furthermore, the registered fused data is obtained through the following steps: A world coordinate system is established with the preset fixed point of the motion platform as the origin. The first position coordinates of the optical axis center of the two-dimensional imaging module and the second position coordinates of the optical axis center of the three-dimensional measurement module in the world coordinate system are obtained. For any hole to be verified, based on its physical coordinates in the world coordinate system and the offset between the first position coordinates and the second position coordinates, the motion platform is controlled to move the hole to be verified to the center of the measurement field of view of the three-dimensional measurement module. The two-dimensional image sub-image of the same hole to be verified is associated with the three-dimensional point cloud data to obtain the registered fused data; wherein, the two-dimensional image sub-image is obtained by cropping the two-dimensional image with the physical coordinates of the hole to be verified as the center, and the three-dimensional point cloud data is a three-dimensional data matrix generated by the three-dimensional measurement module after scanning the hole to be verified; the central element of the three-dimensional data matrix corresponds to the physical coordinate position of the hole to be verified.

[0011] Furthermore, the identification of the out-of-tolerance residual pile specifically includes: In the registered fused data, a vertical section passing through the center of the hole is cut along the axial direction of the back drill hole, and the profile curve of the depth changing with the horizontal position is extracted. On the profile curve, identify the platform area corresponding to the surface of the target layer copper ring, calculate the average depth value in the platform area as the target layer surface height; identify the minimum depth value in the bottom area of ​​the borehole as the borehole bottom height; The difference between the surface height of the target layer and the bottom height of the borehole is calculated as the residual pile length; when the residual pile length exceeds a preset threshold range, it is determined to be an out-of-tolerance defect in the residual pile.

[0012] Furthermore, the identification of copper wire residue and dirt specifically includes: Extract the three-dimensional point cloud of the region corresponding to the two-dimensional abnormal spots, calculate the standard deviation of the height values ​​in the region as the local flatness, and calculate the average height difference of the region relative to the surrounding hole walls. If the average height difference is positive and greater than the preset height threshold, it is determined to be a copper wire residual defect; If the absolute value of the average height difference is less than a preset height threshold, and the two-dimensional abnormal spot has a difference in gray level or texture with the surrounding area, it is determined to be dirty.

[0013] Furthermore, the identification of microcavities, dents, and scratches specifically includes: For the pitted areas detected in the 3D point cloud, the least squares method is used to fit a spherical model to obtain the coordinates of the center and radius of the fitted sphere, and the goodness of fit is calculated. If the goodness of fit is greater than or equal to the preset fitting threshold, it is determined to be a micro-void defect; If the goodness of fit is less than the preset fitting threshold, the ratio of the major and minor axes of the smallest circumscribed ellipse of the pit region in the two-dimensional image is calculated; if the ratio of the major and minor axes is greater than the preset shape threshold, it is determined to be a scratch defect; otherwise, it is determined to be a pit defect.

[0014] Furthermore, the identification of excessive roughness specifically includes: The three-dimensional point cloud of the hole wall region is extracted from the registered fused data, and the three-dimensional point cloud of the hole wall region is unfolded into a two-dimensional depth image by cylindrical coordinate transformation. The two-dimensional depth image is filtered to obtain the filtered surface contour. A contour line is intercepted along a preset direction on the filtered surface contour, and the surface roughness parameter is calculated; when the surface roughness parameter exceeds a preset roughness threshold, it is determined to be a roughness over-standard defect; The surface roughness parameters include the arithmetic mean deviation Ra and / or the maximum profile height Rz; the preset direction includes the axial direction and / or the circumferential direction of the back drill hole.

[0015] Secondly, the present invention proposes a PCB back-drill hole defect detection device for implementing the back-drill hole defect detection method described above, comprising: The two-dimensional imaging module is used to scan the PCB board under test and generate a two-dimensional image; The 3D measurement module is used to scan the hole to be checked and obtain the 3D point cloud data of the hole area; A motion platform is used to support and drive the PCB board under test to move relative to the two-dimensional imaging module and the three-dimensional measurement module. The data processing module is used for: All back-drilled holes in the two-dimensional image are identified and their coordinate positions are determined. Based on the features of the two-dimensional image, a preliminary screening is performed, and a list of coordinates of holes to be verified is output. The motion platform is controlled to move each hole to be verified to the measurement field of view of the three-dimensional measurement module in sequence according to the coordinate list; The three-dimensional measurement module is controlled to scan each hole to be verified and obtain the three-dimensional point cloud data of the hole area; By using a unified coordinate system, the two-dimensional image and three-dimensional point cloud data of the same hole are spatially registered to obtain the registered fused data; the fused data includes the two-dimensional image sub-image of the hole to be verified, the three-dimensional point cloud data matrix, and the spatial correspondence between the two. The registered fused data is subjected to defect detection and classification to identify at least one defect type among residual piles exceeding tolerance, copper wire residue, dirt, micro-voids, dents, scratches, and excessive roughness.

[0016] Furthermore, the three-dimensional measurement module is a 3D white light interferometer, including a white light interferometer objective lens, a vertical scanning drive mechanism, and an interferometric image acquisition camera; the motion platform is an XYZ three-axis motion platform, and the two-dimensional imaging module and the three-dimensional measurement module are arranged side by side along the Y-axis.

[0017] Thirdly, this invention proposes a method for evaluating the quality of PCB back-drilled holes, the method comprising: Obtain the defect detection classification results output by the above detection method, as well as the detection data of normal holes; The defect detection and classification results are statistically analyzed at different levels to generate a board-level defect distribution report; Extract the residual pile length, hole diameter, and hole wall roughness data of holes that are determined to be normal, calculate the process capability index, analyze the distribution of each parameter in the plate area, and generate regional analysis results. The regional analysis results include information on whether each grid area is a parameter abnormal area. The proportions of each level of defects in the board-level defect distribution report, the process capability index, and the abnormal area information in the area analysis results are weighted and calculated to output a comprehensive quality score and a quality analysis report.

[0018] Furthermore, the analysis of the distribution of each parameter on the board surface includes: dividing the PCB board surface into multiple grid regions, calculating the mean value of at least one parameter among the residual length, hole diameter, and hole wall roughness of normal holes in each grid region; if the absolute value of the difference between the mean value of this parameter in a certain grid region and the mean value of this parameter in the whole board is greater than a preset multiple of the standard deviation of the whole board, then the grid region is determined to be a parameter abnormal region; the region analysis results include the determination information of whether each grid region is a parameter abnormal region.

[0019] The beneficial effects of this invention are as follows: This invention spatially registers two-dimensional image sub-images of the same back-drilled hole with a three-dimensional point cloud data matrix, enabling precise physical correspondence between two-dimensional texture features and three-dimensional morphological features. This allows for the differentiation of copper wire residue and contaminants based on the three-dimensional protrusion height, and the differentiation of microcavities and depressions or scratches based on the spherical fit goodness of the three-dimensional pits combined with two-dimensional shape features. The length of the residual pile is directly calculated from the height difference between the target layer copper ring surface platform and the bottom of the drill hole in the three-dimensional depth profile. The hole wall roughness is calculated by expanding the three-dimensional point cloud of the hole wall in cylindrical coordinates, giving the above three-dimensional dimensions quantifiable evaluation indicators. The quality evaluation comprehensively considers the defect classification ratio, process capability index, and abnormal parameter judgment results of each area of ​​the board, reflecting both the single board quality level and indicating abnormal process areas. Attached Figure Description

[0020] Figure 1 A schematic flowchart of a PCB back-drill hole defect detection method provided in an embodiment of the present invention; Figure 2 This is another flowchart illustrating the PCB back-drill hole defect detection method provided in this embodiment of the invention; Figure 3 This is a structural block diagram of a PCB back-drill hole defect detection device provided in an embodiment of the present invention. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0022] The PCB back-drilling defect detection method, detection device, and evaluation method provided by this invention are applicable to the automated detection of back-drilling processes in high-end PCB products such as high-density interconnect boards and packaging carrier boards. The following detailed description of each technical aspect is provided in conjunction with specific embodiments. Example 1

[0023] Please see Figure 1 and Figure 2 This embodiment proposes a method for detecting defects in PCB back-drilled holes, the method including: S1. Scan the PCB board to be tested through the two-dimensional imaging module to generate a two-dimensional image, namely a 2D panoramic image of the whole board. Identify all back drill holes in the two-dimensional image and determine their coordinate positions. Perform preliminary screening based on the features of the two-dimensional image and output a list of coordinates of holes to be verified. S2. Based on the coordinate list, each hole to be verified is moved to the 3D measurement module for scanning via the motion platform to obtain the 3D point cloud data of the hole area; the 2D image and 3D point cloud data of the same hole are spatially registered to obtain the registered fused data; wherein, the fused data includes the 2D image sub-image of the hole to be verified, the 3D point cloud data matrix, and the spatial correspondence between the two. S3. Perform defect detection and classification on the registered fused data to identify at least one type of defect, including residual piles exceeding tolerance, copper wire residue, dirt, micro-voids, dents, scratches, and excessive roughness.

[0024] The specific implementation methods for each step are explained in detail below.

[0025] I. Initial screening of two-dimensional images and positioning of wells to be verified The two-dimensional imaging module uses a line scan camera. The PCB board under test is driven by a motion platform to pass through the camera's field of view at a constant speed. The line scan camera continuously acquires and generates a high-resolution two-dimensional panoramic image of the entire board.

[0026] In a preferred embodiment, the two-dimensional image features include at least one of the following: continuity of the hole ring contour, grayscale deviation value of the area inside the hole, and proportion of foreign matter area inside the hole.

[0027] In a preferred embodiment, the list of coordinates of the wells to be verified is obtained through the following steps: S11. In the 2D image, all back-drilled holes are identified using a circle recognition algorithm to determine the center pixel coordinates of each hole. The circle recognition algorithm can employ either Hough circle transform or a gradient-direction-based circle detection method. Hough circle transform searches for the center coordinates and radius in the parameter space by accumulating votes on edge points; template matching algorithms pre-establish standard back-drilled hole image templates and slide to match them in the 2D image to locate the holes. Both methods can be used individually or in combination to adapt to different lighting conditions and hole shape variations.

[0028] S12. For each identified back-drilled hole, extract the two-dimensional image features of its hole ring region and hole inner region, and compare them with a preset threshold. If at least one of the following conditions is met, mark it as a hole to be reviewed: If the continuity of the hole ring outline is lower than a preset threshold, it is judged as a hole ring defect. The continuity of the hole ring outline is defined as the proportion of continuous pixels in the hole ring outline to the total number of pixels in the outline. For example, if the theoretical circumference of the hole ring is 1000 pixels, and the cumulative number of broken parts exceeds 50 pixels (i.e., the continuity is lower than 95%), it is marked as a suspected hole ring defect.

[0029] If the absolute value of the difference between the grayscale value of a region within a hole and the average grayscale value of a standard hole exceeds a preset grayscale threshold, it is determined to be a foreign object within the hole. The average grayscale value of a standard hole is obtained by statistically averaging the grayscale values ​​of the inner regions of multiple known normal holes. For example, the preset grayscale threshold is typically set to 20 (8-bit grayscale image, grayscale range 0-255). When the deviation of the grayscale value of a continuous region within a hole from the standard value exceeds this threshold, it indicates that there may be a foreign object or dirt in that region.

[0030] If the elliptic fit of the hole area exceeds the preset range, it is determined to be hole deformation. The elliptic fit is determined by calculating the minimum circumscribed ellipse of the hole profile and taking the ratio of its major and minor axes as the criterion. For example, if the ratio of the major and minor axes exceeds 1.1, it indicates that there is obvious elliptic deformation at the hole opening, which may be caused by drilling misalignment or stress on the plate.

[0031] It should be noted that "holes to be verified" refers to back-drilled holes that have been initially screened using 2D image features and deemed potentially defective. These back-drilled holes exhibit at least one of the following abnormal characteristics in the 2D image: hole ring contour continuity is below a preset threshold, grayscale deviation value within the hole area is above a preset threshold, and ellipse fitting degree of the hole area exceeds a preset range. In contrast, back-drilled holes not marked as "holes to be verified" are considered normal holes and can directly proceed to the subsequent process capability analysis stage or be exempted from 3D measurement. Through the above initial screening mechanism, the object of 3D precision measurement is limited to the holes to be verified, avoiding 3D scanning of all back-drilled holes on the entire board.

[0032] S13. Using a pre-calibrated transformation matrix between pixel coordinates and platform physical coordinates, the center pixel coordinates of each hole to be verified are converted into physical coordinates in the motion platform coordinate system, generating a list of coordinates for the holes to be verified. The calibration process uses a standard calibration board to extract the pixel coordinates of marker points with known spacing from the 2D image, while simultaneously measuring their physical coordinates in the motion platform coordinate system. The transformation matrix parameters are then solved using the least squares method. Thus, the hole positions identified in the 2D image can be accurately mapped to the target motion positions of the motion platform.

[0033] As an example, the list of coordinates of the holes to be reviewed can be a data structure (such as a CSV file or array), with each row corresponding to a suspected defective hole, containing a unique ID and physical coordinates, as shown in Table 1.

[0034] Table 1. List of coordinates of holes to be verified Through the aforementioned two-dimensional initial screening step, all back-drilled holes on the entire board are quickly scanned, and only the holes that meet the initial screening criteria are output for verification in the subsequent three-dimensional measurement stage. For example, for a typical high-speed backplane containing approximately 5,000 back-drilled holes, the number of holes to be verified after initial screening is usually between 100 and 200, significantly reducing the workload of three-dimensional measurement.

[0035] II. Three-dimensional topography scanning and data registration The 3D measurement module performs a vertical scan of the hole to be verified, acquiring the 3D surface topography point cloud data of the hole area. The 3D measurement module can use a white light interferometer, whose measurement principle is as follows: the white light source is split into a measurement beam and a reference beam by a beam splitter. After the measurement beam illuminates the sample surface, it is reflected and interferes with the reference beam. The optical path difference is changed by the vertical scanning mechanism, and the changes in the interference fringes are recorded, thereby reconstructing the 3D topography of the sample surface.

[0036] The vertical resolution of a white light interferometer can reach the nanometer level, while the lateral resolution is determined by the objective lens numerical aperture and the camera pixel size, with a typical sampling interval of 0.5 μm-2 μm. For a measurement field of view of 0.5 mm × 0.5 mm, sampling at 1 μm intervals yields a 501 × 501 three-dimensional data matrix, where each element represents the height value of the sampling point (in μm); this matrix represents the 3D surface topography point cloud data of that aperture.

[0037] In a preferred embodiment, the registered fused data is obtained through the following steps: S21. Establish a world coordinate system with the preset fixed point of the motion platform as the origin, and obtain the first position coordinates of the optical axis center of the 2D imaging module and the second position coordinates of the optical axis center of the 3D measurement module in the world coordinate system. The 2D imaging module and the 3D measurement module are arranged side by side on the motion platform, and the physical offset of their optical axis centers is obtained through high-precision calibration.

[0038] S22. For any hole to be verified, based on its physical coordinates in the world coordinate system and the offset between the first and second position coordinates, control the motion platform to move the hole to the center of the measurement field of view of the 3D measurement module. The specific motion control logic is as follows: if the physical coordinates of the hole to be verified are (X_board, Y_board), and the center coordinates of the optical axis of the 3D measurement module are (X_int, Y_int), then the target position that the motion platform needs to move to is (X_board-X_int, Y_board-Y_int), so that the hole to be verified is exactly located at the center of the measurement field of view.

[0039] S23. Associate the two-dimensional image sub-images of the same aperture to be verified with the three-dimensional point cloud data to obtain the registered fused data. The two-dimensional image sub-image is obtained by cropping the two-dimensional image with the physical coordinates of the aperture to be verified as the center. The cropping size corresponds to the size of the three-dimensional measurement field of view. For example, if the three-dimensional measurement field of view is 0.5mm × 0.5mm, the two-dimensional image sub-image is also cropped to the corresponding physical size. The three-dimensional point cloud data is a three-dimensional data matrix generated by the three-dimensional measurement module after scanning the aperture to be verified. The central element of the three-dimensional data matrix corresponds to the physical coordinate position of the aperture to be verified. Since the motion platform has moved the aperture to be verified to the center of the measurement field of view, the geometric center sampling point of the three-dimensional data matrix corresponds exactly to the physical coordinates of the aperture to be verified, thus establishing a spatial correspondence between the pixels of the two-dimensional image sub-image and the elements of the three-dimensional point cloud matrix.

[0040] III. Defect Detection Classification Based on Fuded Data The registered fused data includes both two-dimensional image sub-images and a three-dimensional point cloud matrix, with a spatial correspondence established between the two through physical coordinates. Based on this fused data, various defect types can be identified and classified.

[0041] In a preferred embodiment, the identification of out-of-tolerance residual piles specifically includes: S31. In the registered fused data, a vertical profile passing through the center of the hole is cut along the axis of the back borehole, and the profile curve showing the depth variation with the horizontal position is extracted. The profile cutting direction is parallel to the axis of the back borehole, and the cutting position passes through the center element of the three-dimensional data matrix (i.e., the physical coordinate position of the hole).

[0042] S32. On the profile curve, identify the plateau region corresponding to the surface of the target layer copper ring, and calculate the average depth value within this plateau region as the target layer surface height; identify the minimum depth value within the borehole bottom region as the borehole bottom height. The surface of the target layer copper ring appears as a relatively flat area on the profile curve, which can be located by finding a continuous interval where the first derivative of the curve is zero and the second derivative is greater than a certain positive threshold; the borehole bottom is the lowest point of the profile curve below the target layer, which can be determined by finding the location where the gradient change approaches zero and the depth value is maximum through a sliding window.

[0043] S33. Calculate the difference between the target layer surface height and the bottom height of the drill hole as the residual pile length; when the residual pile length exceeds the preset threshold range, it is judged as an out-of-tolerance defect. The specification requirements for the residual pile length are usually determined by the PCB design. The measured residual pile length can be directly compared with the upper and lower limits of the specification; if it exceeds the range, it is judged as an out-of-tolerance defect.

[0044] In a preferred embodiment, the identification of copper wire residue and dirt specifically includes: S41. Extract the 3D point cloud of the region corresponding to the 2D anomalous spot, calculate the standard deviation of the height values ​​within the region as the local flatness, and calculate the average height difference of the region relative to the surrounding hole walls. The 2D anomalous spot is located by the detection of foreign objects inside the hole in the 2D primary screening step, and the boundary of the spot can be determined by region growing or connected component analysis. Map the spot region onto the 3D point cloud and extract all height values ​​of the corresponding region.

[0045] S42. If the average height difference is positive and greater than the preset height threshold, it is determined to be a copper wire residue defect. Copper wire residue refers to copper foil that was not completely removed during drilling, forming a protrusion on the hole wall, with a height typically between 5μm and 50μm. The preset height threshold can be set to 5μm; when the average protrusion height exceeds this value, it is determined to be copper wire residue. Furthermore, the protrusion height and volume of copper wire residue can be calculated based on three-dimensional point computing to provide a quantitative basis for process improvement.

[0046] S43. If the absolute value of the average height difference is less than the preset height threshold, and there is a difference in grayscale or texture between the two-dimensional abnormal spot and the surrounding area, it is judged as dirt. Dirt is usually resin residue or contaminant, which is very thin and does not produce obvious protrusions in the three-dimensional morphology, but appears as dark spots or bright spots in the two-dimensional image due to color differences.

[0047] Using the above method, copper wire residue and dirt, two types of defects with similar features in two-dimensional images, can be effectively distinguished based on their differences in three-dimensional morphology, thus reducing the false alarm rate of traditional 2D AOI.

[0048] In a preferred embodiment, the identification of microcavities, dents, and scratches specifically includes: S51. For the pitted regions detected in the 3D point cloud, a spherical model is fitted using the least squares method to obtain the coordinates of the center and radius of the fitted sphere, and the goodness of fit is calculated. The pitted regions are detected through depth threshold segmentation of the 3D point cloud, extracting continuous regions with depths lower than the surrounding reference plane. The goodness of fit is calculated using the coefficient of determination R0. 2 R 2 =1-SS_res / SS_tot, where SS_res is the sum of squared residuals and SS_tot is the total sum of squares. R 2 The closer it is to 1, the closer the shape of the pit is to a part of the sphere.

[0049] S52. If the goodness of fit is greater than or equal to the preset fitting threshold, it is determined to be a micro-void defect. The preset fitting threshold can be set to 0.90. Micro-voids are usually formed by bubbles or material defects during the drilling process, and are regular in shape, approximately spherical. After being determined to be a micro-void, the radius and maximum depression depth of the void can be calculated based on the parameters of the fitted sphere. For example, for the radius R and center position of the fitted sphere, the maximum depression depth is R minus the distance from the center of the sphere to the reference surface of the hole wall.

[0050] S53. If the goodness of fit is less than the preset fitting threshold, calculate the ratio of the major and minor axes of the smallest circumscribed ellipse of the pit region in the 2D image; if the ratio is greater than the preset shape threshold, it is determined to be a scratch defect; otherwise, it is determined to be a dent defect. The preset shape threshold can be set to 2.0. Scratches are usually caused by mechanical scratches, appearing as thin strips with a large ratio of major and minor axes; dents are irregular in shape and may be caused by impacts or indentations. The depth of dents and scratches can be directly obtained from the height difference between the lowest point of the region and the surrounding normal surface in the 3D point cloud.

[0051] The above method combines the geometric features of three-dimensional morphology with the shape features of two-dimensional images to enable the accurate differentiation of microcavities, depressions and scratches.

[0052] In a preferred embodiment, the identification of excessive roughness specifically includes: S61. Extract the 3D point cloud of the borehole wall region from the registered fused data, and unfold the 3D point cloud of the borehole wall region into a 2D depth image through cylindrical coordinate transformation. The cylindrical coordinate transformation takes the back-drilled borehole axis as the Z-axis and transforms the 3D points (x, y, z) on the borehole wall into (…). (z) Two-dimensional coordinates, generate depth z with respect to angle Two-dimensional image I ( x, y); where z is the position (axial depth) along the axis of the back borehole, and x and y are the positions in the plane perpendicular to the axis. This represents the circumferential angular position of the sampling point on the borehole wall relative to the borehole's central axis. The pixel value corresponds to the radial deviation or height value of that point. The unfolded 2D depth image visually reflects the topographic undulations of the borehole wall surface.

[0053] S62, for two-dimensional image I ( The surface profile is obtained by filtering the sample interval (z) to obtain the filtered surface profile. Gaussian filtering or median filtering can be used to remove measurement noise and high-frequency burrs while preserving the true surface morphology. The kernel size and standard deviation of the Gaussian filter are set according to the sampling interval and the required roughness evaluation wavelength.

[0054] S63. Extract the contour line along the preset direction on the filtered surface contour and calculate the surface roughness parameter; when the surface roughness parameter exceeds the preset roughness threshold, it is judged as a roughness over-standard defect. The surface roughness parameters include the arithmetic mean deviation Ra and / or the maximum profile height Rz; the preset directions include the axial and / or circumferential directions of the back drill hole. The formula for calculating Ra is Ra = (1 / L)∫|y(x)| dx, where L is the sampling length, y(x) is the distance of a point on the profile line from the center line, and Rz is the distance between the peak line and the valley line within the sampling length. Example 2

[0055] Please see Figure 3 This embodiment proposes a PCB back-drill hole defect detection device to implement the back-drill hole defect detection method as described in Embodiment 1 above. The device includes: The two-dimensional imaging module is used to scan the PCB board under test and generate a two-dimensional image; The 3D measurement module is used to scan the hole to be checked and obtain the 3D point cloud data of the hole area; A motion platform is used to support and drive the PCB board under test to move relative to the two-dimensional imaging module and the three-dimensional measurement module. The data processing module is used for: identifying all back-drilled holes in the 2D image and determining their coordinate positions; performing initial screening based on 2D image features and outputting a list of coordinates for holes to be verified; controlling the motion platform to move each hole to be verified sequentially to the measurement field of view of the 3D measurement module according to the coordinate list; controlling the 3D measurement module to scan each hole to be verified and acquire 3D point cloud data of the hole area; registering the 2D image and 3D point cloud data of the same hole in a unified coordinate system to obtain the registered fused data; the fused data includes the 2D image sub-image of the hole to be verified, the 3D point cloud data matrix, and the spatial correspondence between the two; and performing defect detection and classification on the registered fused data to identify at least one defect type among residual pile deviation, copper wire residue, dirt, micro-voids, dents, scratches, and excessive roughness.

[0056] The specific configuration and algorithm implementation of the data processing module are consistent with the description of the corresponding steps in Example 1, and will not be repeated here.

[0057] Preferably, the three-dimensional measurement module is a 3D white light interferometer, including a white light interferometer objective, a vertical scanning drive mechanism, and an interferometric image acquisition camera; the motion platform is an XYZ three-axis motion platform, with the two-dimensional imaging module and the three-dimensional measurement module arranged side-by-side along the Y-axis. The white light interferometer objective can be of the Mirau or Michelson type, and the vertical scanning drive mechanism uses a piezoelectric ceramic actuator or a precision ball screw motor. The interferometric image acquisition camera is a CCD or CMOS camera, with the frame rate matched to the scanning speed.

[0058] In addition, the data processing module is also used to: randomly select a certain proportion of normal holes that are not marked as holes to be reviewed, and control the motion platform and the three-dimensional measurement module to scan them to obtain three-dimensional point cloud data of the normal holes for process capability analysis. Example 3

[0059] This embodiment proposes a method for evaluating the quality of PCB back-drilled holes, the method including: Obtain the defect detection classification results output by the detection method in Example 1 above, as well as the detection data of normal holes; The defect detection and classification results are statistically analyzed at different levels to generate a board-level defect distribution report. Extract the residual pile length, hole diameter, and hole wall roughness data of holes that are determined to be normal, calculate the process capability index, analyze the distribution of each parameter in the plate area, and generate regional analysis results. The regional analysis results include information on whether each grid area is a parameter abnormal area. The system performs weighted calculations on the proportion of defects at each level, the process capability index, and the abnormal area information in the area analysis results from the board-level defect distribution report, and outputs a comprehensive quality score and quality analysis report.

[0060] Preferably, analyzing the distribution of each parameter on the board surface includes: dividing the PCB board surface into multiple grid areas, calculating the mean value of at least one parameter among the residual length, hole diameter, and hole wall roughness of normal holes in each grid area, and if the absolute value of the difference between the mean value of this parameter in a certain grid area and the mean value of this parameter in the whole board is greater than a preset multiple of the standard deviation of the whole board, then the grid area is determined to be a parameter abnormal area; the area analysis results include the determination information of whether each grid area is a parameter abnormal area.

[0061] The following section provides a detailed explanation of the specific implementation of the quality evaluation method.

[0062] I. Defect Classification Statistics The defect grading statistics classify various defects in the inspection and classification results into three levels: severe, moderate, and minor, based on their severity. Grading standards can be formulated with reference to industry specifications or customer requirements. An example grading rule is as follows: Serious defects: The length of the residual pile exceeds the upper limit of the specification by more than 50%, the height of the residual copper wire protrusion exceeds 30μm, the diameter of the micro-void exceeds 80μm, the depth of the scratch exceeds 15μm, etc. Common defects include: residual pile length exceeding the specification but within 50%; residual copper wire height of 10-30μm; micro-void diameter of 40-80μm; and depression depth of 5-15μm. Minor defects include: residual pile length being at the edge of the specification, borehole wall roughness Ra being between 4-5 μm, and the dirty area being less than 5% of the borehole area.

[0063] Calculate the percentage of each defect level relative to the total number of back drill holes on the entire board, and generate a board-level defect distribution report. This report visually reflects the quality level of a single board; for example, a board might have a severe defect rate of 0.1%, a general defect rate of 0.5%, a minor defect rate of 1.2%, and an overall defect rate of 1.8%.

[0064] II. Process Capability Analysis For back-drilled holes determined to be normal, datasets of three key parameters—residual pile length, hole diameter, and hole wall roughness—are extracted. The process capability indices Cp and Cpk are calculated for each parameter.

[0065] Preferably, the analysis of the distribution of various parameters on the board surface includes: dividing the PCB board surface into multiple grid regions, the grid size of which can be set according to the board size and hole density, for example, dividing it into 10mm×10mm grids; calculating the mean value of at least one parameter among the residual length, hole diameter, and hole wall roughness of normal holes in each grid region; if the absolute value of the difference between the mean value of this parameter in a certain grid region and the mean value of this parameter in the whole board is greater than a preset multiple of the standard deviation of the whole board, then the grid region is determined to be a parameter abnormality region; the region analysis results include the determination information of whether each grid region is a parameter abnormality region. The preset multiple is typically set to 3 times, that is, the 3σ criterion is used to determine significant deviations. For example, if the mean residual length of the whole board is 32μm and the standard deviation is 4μm, and the mean of a certain grid region is 45μm, then |45-32|=13>3×4=12, and this grid is determined to be a residual length abnormality region and marked as a process warning area.

[0066] III. Overall Quality Score The overall quality score is calculated using a weighted method. Input items include: the proportion of each level of defects in the board-level defect distribution report, the process capability index Cpk value, and the number or area percentage of abnormal areas in the area analysis results.

[0067] An example weighted formula is: Comprehensive score = 100 - (W1 × Critical defect rate × 100 + W2 × General defect rate × 100 + W3 × Minor defect rate × 100) - W4 × Abnormal area ratio × 100 - W5 × max(0, 1.33 - Cpk) × 100 (when Cpk < 1.33).

[0068] W1 to W5 are preset weighting coefficients that can be adjusted according to the focus of quality control, for example, W1=0.5, W2=0.3, W3=0.1, W4=0.05, W5=0.05. The scoring results correspond to the board-level quality grades, such as 90 points or above being excellent, 80-89 points being good, 70-79 points being acceptable, and below 70 points being unacceptable.

[0069] In addition to the overall score, the quality analysis report also includes a defect distribution map (marking the location of each defect type), a key parameter trend map (such as a histogram of the distribution of the length of the remaining piles in the whole slab), a process capability report (including Cp / Cpk values ​​and judgment conclusions), and an area anomaly judgment map (marking the location of process warning areas).

[0070] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0071] In addition, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0072] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting defects in PCB back-drilled holes, characterized in that, The methods include: The PCB board under test is scanned by a two-dimensional imaging module to generate a two-dimensional image. All back drill holes in the two-dimensional image are identified and their coordinate positions are determined. Based on the features of the two-dimensional image, a preliminary screening is performed, and a list of coordinates of holes to be verified is output. Based on the coordinate list, each hole to be verified is sequentially moved to the 3D measurement module for scanning via a motion platform to obtain the 3D point cloud data of the hole area; the 2D image and 3D point cloud data of the same hole are spatially registered to obtain the registered fused data; The fused data includes a two-dimensional image sub-image of the hole to be verified, a three-dimensional point cloud data matrix, and the spatial correspondence between the two. The registered fused data is subjected to defect detection and classification to identify at least one defect type among residual piles exceeding tolerance, copper wire residue, dirt, micro-voids, dents, scratches, and excessive roughness.

2. The PCB back-drill hole defect detection method according to claim 1, characterized in that, The two-dimensional image features include at least one of the following: continuity of the hole ring contour, grayscale deviation value of the area inside the hole, and proportion of foreign matter area inside the hole.

3. The PCB back-drill hole defect detection method according to claim 2, characterized in that, The list of coordinates of the boreholes to be verified is obtained through the following steps: In the two-dimensional image, all back drill holes are identified by a circle recognition algorithm, and the center pixel coordinates of each back drill hole are determined. For each identified back-drilled hole, extract the two-dimensional image features of its hole ring region and hole interior region, and compare them with a preset threshold. A hole is marked as a verification hole if it meets at least one of the following conditions: If the continuity of the hole ring profile is lower than a preset threshold, it is determined to be a hole ring defect; If the absolute value of the difference between the gray value of the area inside the hole and the average gray value of the standard hole is greater than the preset gray value threshold, it is determined to be a foreign object inside the hole; The elliptic fit of the hole area exceeds the preset range, which is determined to be hole deformation. By using a pre-calibrated transformation matrix between pixel coordinates and platform physical coordinates, the center pixel coordinates of each hole to be verified are converted into physical coordinates in the motion platform coordinate system, thereby generating a list of coordinates for the holes to be verified.

4. The PCB back-drill hole defect detection method according to claim 1, characterized in that, The registered and fused data is obtained through the following steps: A world coordinate system is established with the preset fixed point of the motion platform as the origin. The first position coordinates of the optical axis center of the two-dimensional imaging module and the second position coordinates of the optical axis center of the three-dimensional measurement module in the world coordinate system are obtained. For any hole to be verified, based on its physical coordinates in the world coordinate system and the offset between the first position coordinates and the second position coordinates, the motion platform is controlled to move the hole to be verified to the center of the measurement field of view of the three-dimensional measurement module. The two-dimensional image sub-image of the same hole to be verified is associated with the three-dimensional point cloud data to obtain the registered fused data; wherein, the two-dimensional image sub-image is obtained by cropping the two-dimensional image with the physical coordinates of the hole to be verified as the center, and the three-dimensional point cloud data is a three-dimensional data matrix generated by the three-dimensional measurement module after scanning the hole to be verified; the central element of the three-dimensional data matrix corresponds to the physical coordinate position of the hole to be verified.

5. The PCB back-drill hole defect detection method according to claim 1, characterized in that, The identification of the out-of-tolerance residual piles specifically includes: In the registered fused data, a vertical section passing through the center of the hole is cut along the axial direction of the back drill hole, and the profile curve of the depth changing with the horizontal position is extracted. On the profile curve, identify the platform area corresponding to the surface of the target layer copper ring, calculate the average depth value in the platform area as the target layer surface height; identify the minimum depth value in the bottom area of ​​the borehole as the borehole bottom height; The difference between the surface height of the target layer and the bottom height of the borehole is calculated as the residual pile length; when the residual pile length exceeds a preset threshold range, it is determined to be an out-of-tolerance defect in the residual pile.

6. The PCB back-drill hole defect detection method according to claim 1, characterized in that, The identification of copper wire residue and dirt specifically includes: Extract the three-dimensional point cloud of the region corresponding to the two-dimensional abnormal spots, calculate the standard deviation of the height values ​​in the region as the local flatness, and calculate the average height difference of the region relative to the surrounding hole walls. If the average height difference is positive and greater than the preset height threshold, it is determined to be a copper wire residual defect; If the absolute value of the average height difference is less than a preset height threshold, and the two-dimensional abnormal spot has a difference in gray level or texture with the surrounding area, it is determined to be dirty.

7. The PCB back-drill hole defect detection method according to claim 1, characterized in that, The identification of microcavities, dents, and scratches specifically includes: For the pitted areas detected in the 3D point cloud, the least squares method is used to fit a spherical model to obtain the coordinates of the center and radius of the fitted sphere, and the goodness of fit is calculated. If the goodness of fit is greater than or equal to the preset fitting threshold, it is determined to be a micro-void defect; If the goodness of fit is less than the preset fitting threshold, the ratio of the major and minor axes of the smallest circumscribed ellipse of the pit region in the two-dimensional image is calculated; if the ratio of the major and minor axes is greater than the preset shape threshold, it is determined to be a scratch defect; otherwise, it is determined to be a pit defect.

8. The PCB back-drill hole defect detection method according to claim 1, characterized in that, The identification of excessive roughness specifically includes: The three-dimensional point cloud of the hole wall region is extracted from the registered fused data, and the three-dimensional point cloud of the hole wall region is unfolded into a two-dimensional depth image by cylindrical coordinate transformation. The two-dimensional depth image is filtered to obtain the filtered surface contour. A contour line is intercepted along a preset direction on the filtered surface contour, and the surface roughness parameter is calculated; when the surface roughness parameter exceeds a preset roughness threshold, it is determined to be a roughness over-standard defect; The surface roughness parameters include the arithmetic mean deviation Ra and / or the maximum profile height Rz; the preset direction includes the axial direction and / or the circumferential direction of the back drill hole.

9. A PCB back-drill hole defect detection device, used to implement the back-drill hole defect detection method as described in claim 1, characterized in that, include: The two-dimensional imaging module is used to scan the PCB board under test and generate a two-dimensional image; The 3D measurement module is used to scan the hole to be checked and obtain the 3D point cloud data of the hole area; A motion platform is used to support and drive the PCB board under test to move relative to the two-dimensional imaging module and the three-dimensional measurement module. The data processing module is used for: All back-drilled holes in the two-dimensional image are identified and their coordinate positions are determined. Based on the features of the two-dimensional image, a preliminary screening is performed, and a list of coordinates of holes to be verified is output. The motion platform is controlled to move each hole to be verified to the measurement field of view of the three-dimensional measurement module in sequence according to the coordinate list; The three-dimensional measurement module is controlled to scan each hole to be verified and obtain the three-dimensional point cloud data of the hole area; By using a unified coordinate system, the two-dimensional image and three-dimensional point cloud data of the same hole are spatially registered to obtain the registered fused data; the fused data includes the two-dimensional image sub-image of the hole to be verified, the three-dimensional point cloud data matrix, and the spatial correspondence between the two. The registered fused data is subjected to defect detection and classification to identify at least one defect type among residual piles exceeding tolerance, copper wire residue, dirt, micro-voids, dents, scratches, and excessive roughness.

10. The PCB back-drill hole defect detection device according to claim 9, characterized in that, The three-dimensional measurement module is a 3D white light interferometer, including a white light interferometer objective, a vertical scanning drive mechanism, and an interferometric image acquisition camera; the motion platform is an XYZ three-axis motion platform, and the two-dimensional imaging module and the three-dimensional measurement module are arranged side by side along the Y-axis.

11. A method for evaluating the quality of PCB back-drilled holes, characterized in that, The methods include: Obtain the defect detection classification results output by the detection method described in claim 1, as well as the detection data of normal holes; The defect detection and classification results are statistically analyzed at different levels to generate a board-level defect distribution report; Extract the residual pile length, hole diameter, and hole wall roughness data of holes that are determined to be normal, calculate the process capability index, analyze the distribution of each parameter in the plate area, and generate regional analysis results. The regional analysis results include information on whether each grid area is a parameter abnormal area. The proportions of each level of defects in the board-level defect distribution report, the process capability index, and the abnormal area information in the area analysis results are weighted and calculated to output a comprehensive quality score and a quality analysis report.

12. The PCB back-drilled hole quality evaluation method according to claim 11, characterized in that, The analysis of the distribution of various parameters on the board surface includes: dividing the PCB board surface into multiple grid areas, calculating the mean value of at least one parameter among the residual length, hole diameter, and hole wall roughness of normal holes in each grid area; if the absolute value of the difference between the mean value of this parameter in a certain grid area and the mean value of this parameter in the whole board is greater than a preset multiple of the standard deviation of the whole board, then the grid area is determined to be a parameter abnormal area; the regional analysis results include the determination information of whether each grid area is a parameter abnormal area.

Images