Automobile circuit board drilling burr detection method and system based on region growth

Abstract

The invention discloses an automobile circuit board drilling burr detection method and system based on area growth, and aims to improve the quality control efficiency and accuracy of automobile circuit board manufacturing. The system is used for quickly and accurately identifying and evaluating burr defects in the automobile circuit board drilling process; a high-definition image and three-dimensional scanning data of an automobile circuit board drilling area are obtained through a high-precision image acquisition device, and image preprocessing including noise reduction, enhancement and graying processing is carried out. An intelligent seed point selection strategy is adopted, three-dimensional data are combined, deep analysis is carried out on the areas, and various features including the shape, the size, the brightness and the like are extracted. According to the method, the extracted features are classified by training a primitive machine learning model, the burrs are effectively identified and classified, the positions of the burrs are accurately positioned, the potential influence on the performance of the automobile circuit board is evaluated according to the features, a detailed detection report is generated, and a scientific basis is provided for quality control and production optimization.

Description

Technical Field

[0001] The present invention relates to the field of drilling detection, and more particularly to a method and system for detecting drilling burrs on automobile circuit boards based on region growth. Background Art

[0002] Mechanical drilling utilizes a high-speed rotating spindle to rapidly penetrate automotive circuit boards, achieving the desired hole diameter. This mechanical motion can lead to burrs and burrs at the hole opening due to various factors. These burrs and burrs can grow and thicken after electroplating, severely impacting the product's appearance and hole functionality. Conventional inspection techniques rely on drill tape to produce inspection data, resulting in ineffective electroplating inspection data: burrs and burrs are lost to the customer, leading to difficulties with insertion and copper breakage, resulting in quality risks and financial losses. Traditional manual inspection methods rely on operator experience, are subjective, and susceptible to fatigue, making it difficult to maintain high accuracy and consistency over time. Furthermore, with the miniaturization and increased integration of electronic devices, quality requirements for automotive circuit boards are becoming increasingly stringent, and even tiny burrs can lead to product failure. The increasingly complex structures of modern automotive circuit boards, coupled with the widespread use of multilayer and high-density interconnect (HDI) boards, complicate burr detection. In summary, the automotive circuit board drilling burr detection method and system based on regional growth effectively addresses the challenges of quality control in automotive circuit board manufacturing through intelligent technology. Summary of the Invention

[0003] A method for detecting drilling burrs on automobile circuit boards based on region growing comprises the following steps: S1. Requirements Analysis and Data Collection: Clarify the inspection objectives, accuracy requirements, and application scenarios. Use a high-resolution industrial camera to capture clear images of the automotive PCB drilling area. Collect samples of various types and quality levels of automotive PCB drilling, and carefully label them, including normal drilling, mild burrs, and severe burrs. S2. Data Preprocessing: All collected images are formatted, resized, and brightness balanced. Color images are converted to grayscale to reduce color information interference. Median filtering is applied to remove image noise, and edge detection algorithms are used to enhance drill hole edges and burr features. S3. Region Segmentation and Seed Point Selection: Based on preset conditions such as brightness thresholds and shape features, the burr-free portion of the drilled hole is automatically selected as the seed point for the "normal" region. Starting from this seed point, the region is gradually expanded according to a specific growth rule based on a threshold of pixel similarity. This process distinguishes the normal surface around the drilled hole from the abnormal burr region. S4. 3D Scanning and Analysis: 3D scan the automotive circuit board to obtain an accurate 3D model of the drilled area. De-noise and segment the point cloud data to identify the 3D features of the drilled holes and burrs. For suspected burr areas separated by region growing, further extract their shape, size, and brightness features. Apply a support vector machine (SVM) to determine whether these features match those of a burr based on a trained classifier, thereby determining whether they are true burrs. S5. Burr Location and Assessment: Identify and mark the locations of all identified burr areas in the image. Based on the burr size and distribution, assess the potential impact on the functionality of the automotive PCB. Combined with 3D scanning data, perform multi-dimensional verification to improve detection accuracy and completeness. S6. Result Output and Feedback: Test results are output in the form of graphics and reports, indicating the test status of each drill hole, including the presence and location of burrs. These results are then fed back to the production system, allowing unqualified automotive PCBs to be removed and repaired. S7. Report generation and decision support: Summarize test data, generate detailed test reports, provide visual data analysis, and provide a scientific basis for production management and quality improvement.

[0004] Furthermore, a method for detecting drilling burrs on automotive circuit boards based on region growing is proposed. In step S3, the burr-free portion within the drilling area is automatically selected as the seed point of the "normal" area based on the preset conditions: brightness threshold and shape features, as follows; Automatic seed point selection strategy: Brightness threshold method: Based on the image grayscale histogram, an appropriate threshold is selected to distinguish between dark areas (normal surface) and bright areas (possible burrs or edges) in the drilled area. Based on this, the center point of the burr-free area is found as the seed point, usually the darker and more uniform part inside the drilled hole. Shape feature analysis: In addition to brightness, shape features can also be considered to assist in seed point selection. For example, using the image contour after edge detection, the circular outline of the drilled hole is identified, and the center point or a point close to the center without burr interference is selected as the seed point. Region growing algorithm: Similarity metric: The key to the region growing process is to define an appropriate pixel similarity metric. Commonly used distance metrics include grayscale difference, texture similarity, or color histogram. The algorithm starts from a seed point and gradually adds adjacent pixels that meet the similarity threshold to the current region until no new pixels can be added or the preset stop condition is reached. Multi-scale and multi-directional expansion: In order to adapt to burrs of different sizes and directions, the region growing algorithm can be designed to expand at multiple scales or in multiple directions, that is, starting at different scales or from multiple initial seed points to adapt to complex drilling edges and burr shapes; Abnormal region identification: Distinguishing between normal and abnormal regions. Regions are obtained through region growing. Normal regions should appear as continuous, large areas with relatively consistent grayscale, while burr regions appear as small, high-brightness, or irregularly shaped areas that are significantly different from normal regions. By comparing regional features (such as area, boundary roughness, grayscale variance, etc.), normal drilling surfaces can be effectively distinguished from burr regions. Real-time feedback and adjustment: The region segmentation and seed point selection processes need to provide feedback to preprocessing steps (such as edge enhancement) and subsequent 3D analysis. Parameters (such as similarity threshold and seed point location) can be dynamically adjusted to accommodate variations in automotive PCB materials and drilling processes, ensuring robustness and accuracy of detection.

[0005] Furthermore, a method for detecting drilling burrs on automotive circuit boards based on region growing is proposed. In step S4, a three-dimensional scan is performed on the automobile circuit board to obtain an accurate three-dimensional model of the drilling area. The specific steps are as follows: Application of 3D scanning technology: First, 3D scanning technology is used to perform a full-scale scan of the drilling area of the automotive circuit board. 3D scanning emits a laser light source, captures the reflected light information, and uses triangulation principles or structured light technology to reconstruct a 3D point cloud model of the drilling area. This accurately captures subtle changes in the drilled surface, including the raised shape of burrs. Point cloud data processing: Denoising: The original point cloud data often contains a large number of noisy points: ambient light interference, sensor misreading, these irrelevant points are filtered out through the algorithm, retaining the information reflecting the real physical structure, Segmentation: In order to analyze each drill hole and the burrs around it separately, the point cloud is segmented. This is achieved through the distance-based DBSCAN clustering algorithm, which divides the point cloud into different clusters according to the spatial distance between points. Each cluster represents an independent drill hole area. 3D feature recognition: For each drilling area, the 3D features of the burr are further extracted, including the burr's location, size, shape (height, width, volume, and direction). These features are obtained by analyzing the local geometric properties of the point cloud data. Application of shape, size, and brightness features: In addition to geometric features, considering that the brightness of burrs is different from that of the surrounding normal surfaces, brightness information is also extracted from the scanning data as an auxiliary feature. SVM classification: After feature extraction, support vector machines are used to identify and classify burrs. The optimal hyperplane is searched to maximize the spacing between feature vectors of different categories. A large number of labeled samples, including normal drilling, mild burrs, and severe burrs, are used to train the SVM in advance. For suspected burr areas initially identified through region growing, their 3D features are input into the trained SVM model. The model then determines whether the area is indeed a burr based on these features.

[0006] Furthermore, a region growing-based automotive circuit board drilling burr detection system is provided, wherein the region growing-based automotive circuit board drilling burr detection system is used to implement any of the region growing-based automotive circuit board drilling burr detection methods described above; the region growing-based automotive circuit board drilling burr detection system comprises: an image acquisition and processing module, a data management and annotation module, a region segmentation and feature extraction module, a burr recognition and classification module, a burr location and evaluation module, and a result output and feedback module; The image acquisition and processing module, which includes high-resolution industrial cameras and 3D scanning equipment, is responsible for capturing high-definition images and 3D point cloud data of the drilling area of the automotive circuit board. It also performs image format conversion, size adjustment, brightness balance, grayscale conversion, median filtering to remove noise, and edge enhancement operations, providing high-quality image input for subsequent analysis. Data management and annotation module: stores various annotated drilling image data, including normal drilling, light burr, and heavy burr samples, for detailed annotation of new samples, supporting manual and semi-automatic annotation; Region Segmentation and Feature Extraction Module: This module automatically identifies and selects burr-free areas within the drilled area as seed points based on brightness thresholds and shape features. Using these seed points, it gradually expands the area using pixel similarity thresholds to distinguish between normal surfaces and burred areas. It then denoises and segments the 3D scan data, identifies the 3D features of the drilled holes and burrs, and fuses them with the 2D image analysis results. Burr recognition and classification module: Extracts shape, size, and brightness features from suspected burr areas and applies support vector machines (SVM) to classify the extracted features to determine whether they are true burrs; Burr Location and Evaluation Module: Accurately locates the positions of all identified burr areas in the image and evaluates their potential impact on the function of the automotive circuit board based on information such as burr size and distribution; Result output and feedback module: organizes the test results into graphic and text reports, including detailed information on drilling status and burr location, and feeds the test results back to the production management system in real time to guide the removal and repair of unqualified products.

[0007] Beneficial effects of the present invention: By combining high-precision image acquisition with three-dimensional scanning technology, as well as advanced region growing algorithms, the system can accurately identify tiny burrs, improve the sensitivity and accuracy of detection, and reduce the missed detection rate. The automated process reduces manual intervention, speeds up detection, improves production efficiency, and reduces labor costs. Automated processing is achieved from image acquisition to defect classification, positioning, and even evaluation report generation. The application of machine learning models for intelligent classification of features reduces the subjectivity of human judgment and ensures the consistency and reliability of detection results. The detection data collected by the system helps to analyze the causes of burrs, and then guides the adjustment of drilling parameters, material selection or post-processing processes, continuously optimizes the production process, and reduces burrs. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a flowchart of a method for detecting drilling burrs on automotive circuit boards based on region growing; DETAILED DESCRIPTION

[0009] A method for detecting burrs in automotive circuit boards based on region growing, the process is as follows Figure 1 As shown, the following steps are included: S1. Requirements Analysis and Data Collection: Clarify the inspection objectives, accuracy requirements, and application scenarios. Use a high-resolution industrial camera to capture clear images of the automotive PCB drilling area. Collect samples of various types and quality levels of automotive PCB drilling, and carefully label them, including normal drilling, mild burrs, and severe burrs. S2. Data Preprocessing: All collected images are formatted, resized, and brightness balanced. Color images are converted to grayscale to reduce color information interference. Median filtering is applied to remove image noise, and edge detection algorithms are used to enhance drill hole edges and burr features. S3. Region Segmentation and Seed Point Selection: Based on preset conditions such as brightness thresholds and shape features, the burr-free portion of the drilled hole is automatically selected as the seed point for the "normal" region. Starting from this seed point, the region is gradually expanded according to a specific growth rule based on a threshold of pixel similarity. This process distinguishes the normal surface around the drilled hole from the abnormal burr region. S4. 3D Scanning and Analysis: 3D scan the automotive circuit board to obtain an accurate 3D model of the drilled area. De-noise and segment the point cloud data to identify the 3D features of the drilled holes and burrs. For suspected burr areas separated by region growing, further extract their shape, size, and brightness features. Apply a support vector machine (SVM) to determine whether these features match those of a burr based on a trained classifier, thereby determining whether they are true burrs. S5. Burr Location and Assessment: Identify and mark the locations of all identified burr areas in the image. Based on the burr size and distribution, assess the potential impact on the functionality of the automotive PCB. Combined with 3D scanning data, perform multi-dimensional verification to improve detection accuracy and completeness. S6. Result Output and Feedback: Test results are output in the form of graphics and reports, indicating the test status of each drill hole, including the presence and location of burrs. These results are then fed back to the production system, allowing unqualified automotive PCBs to be removed and repaired. S7. Report generation and decision support: Summarize test data, generate detailed test reports, provide visual data analysis, and provide a scientific basis for production management and quality improvement.

[0010] Furthermore, a method for detecting drilling burrs on automotive circuit boards based on region growing is proposed. In step S3, the burr-free portion within the drilling area is automatically selected as the seed point of the "normal" area based on the preset conditions: brightness threshold and shape features, as follows; Automatic seed point selection strategy: Brightness threshold method: Based on the image grayscale histogram, an appropriate threshold is selected to distinguish between dark areas (normal surface) and bright areas (possible burrs or edges) in the drilled area. Based on this, the center point of the burr-free area is found as the seed point, usually the darker and more uniform part inside the drilled hole. Shape feature analysis: In addition to brightness, shape features can also be considered to assist in seed point selection. For example, using the image contour after edge detection, the circular outline of the drilled hole is identified, and the center point or a point close to the center without burr interference is selected as the seed point. Region growing algorithm: Similarity metric: The key to the region growing process is to define an appropriate pixel similarity metric. Commonly used distance metrics include grayscale difference, texture similarity, or color histogram. The algorithm starts from a seed point and gradually adds adjacent pixels that meet the similarity threshold to the current region until no new pixels can be added or the preset stop condition is reached. Multi-scale and multi-directional expansion: In order to adapt to burrs of different sizes and directions, the region growing algorithm can be designed to expand at multiple scales or in multiple directions, that is, starting at different scales or from multiple initial seed points to adapt to complex drilling edges and burr shapes; Abnormal region identification: Distinguishing between normal and abnormal regions. Regions are obtained through region growing. Normal regions should appear as continuous, large areas with relatively consistent grayscale, while burr regions appear as small, high-brightness, or irregularly shaped areas that are significantly different from normal regions. By comparing regional features (such as area, boundary roughness, grayscale variance, etc.), normal drilling surfaces can be effectively distinguished from burr regions. Real-time feedback and adjustment: The region segmentation and seed point selection processes need to provide feedback to preprocessing steps (such as edge enhancement) and subsequent 3D analysis. Parameters (such as similarity threshold and seed point location) can be dynamically adjusted to accommodate variations in automotive PCB materials and drilling processes, ensuring robustness and accuracy of detection.

[0011] Furthermore, a method for detecting drilling burrs on automotive circuit boards based on region growing is proposed. In step S4, a three-dimensional scan is performed on the automobile circuit board to obtain an accurate three-dimensional model of the drilling area. The specific steps are as follows: Application of 3D scanning technology: First, 3D scanning technology is used to perform a full-scale scan of the drilling area of the automotive circuit board. 3D scanning emits a laser light source, captures the reflected light information, and uses triangulation principles or structured light technology to reconstruct a 3D point cloud model of the drilling area. This accurately captures subtle changes in the drilled surface, including the raised shape of burrs. Point cloud data processing: Denoising: The original point cloud data often contains a large number of noisy points: ambient light interference, sensor misreading, these irrelevant points are filtered out through the algorithm, retaining the information reflecting the real physical structure, Segmentation: In order to analyze each drill hole and the burrs around it separately, the point cloud is segmented. This is achieved through the distance-based DBSCAN clustering algorithm, which divides the point cloud into different clusters according to the spatial distance between points. Each cluster represents an independent drill hole area. 3D feature recognition: For each drilling area, the 3D features of the burr are further extracted, including the burr's location, size, shape (height, width, volume, and direction). These features are obtained by analyzing the local geometric properties of the point cloud data. Application of shape, size, and brightness features: In addition to geometric features, considering that the brightness of burrs is different from that of the surrounding normal surfaces, brightness information is also extracted from the scanning data as an auxiliary feature. SVM classification: After feature extraction, support vector machines are used to identify and classify burrs. The optimal hyperplane is searched to maximize the spacing between feature vectors of different categories. A large number of labeled samples, including normal drilling, mild burrs, and severe burrs, are used to train the SVM in advance. For suspected burr areas initially identified through region growing, their 3D features are input into the trained SVM model. The model then determines whether the area is indeed a burr based on these features.

[0012] Furthermore, a region growing-based automotive circuit board drilling burr detection system is provided, wherein the region growing-based automotive circuit board drilling burr detection system is used to implement any of the region growing-based automotive circuit board drilling burr detection methods described above; the region growing-based automotive circuit board drilling burr detection system comprises: an image acquisition and processing module, a data management and annotation module, a region segmentation and feature extraction module, a burr recognition and classification module, a burr location and evaluation module, and a result output and feedback module; The image acquisition and processing module, which includes high-resolution industrial cameras and 3D scanning equipment, is responsible for capturing high-definition images and 3D point cloud data of the drilling area of the automotive circuit board. It also performs image format conversion, size adjustment, brightness balance, grayscale conversion, median filtering to remove noise, and edge enhancement operations, providing high-quality image input for subsequent analysis. Data management and annotation module: stores various annotated drilling image data, including normal drilling, light burr, and heavy burr samples, for detailed annotation of new samples, supporting manual and semi-automatic annotation; Region Segmentation and Feature Extraction Module: This module automatically identifies and selects burr-free areas within the drilled area as seed points based on brightness thresholds and shape features. Using these seed points, it gradually expands the area using pixel similarity thresholds to distinguish between normal surfaces and burred areas. It then denoises and segments the 3D scan data, identifies the 3D features of the drilled holes and burrs, and fuses them with the 2D image analysis results. Burr recognition and classification module: Extracts shape, size, and brightness features from suspected burr areas and applies support vector machines (SVM) to classify the extracted features to determine whether they are true burrs; Burr Location and Evaluation Module: Accurately locates the positions of all identified burr areas in the image and evaluates their potential impact on the function of the automotive circuit board based on information such as burr size and distribution; Result output and feedback module: organizes the test results into graphic and text reports, including detailed information on drilling status and burr location, and feeds the test results back to the production management system in real time to guide the removal and repair of unqualified products.

Claims

1. A method for detecting drilling burrs on automobile circuit boards based on region growing, characterized in that: The following steps are involved: S1. Requirements Analysis and Data Collection: Clarify the inspection objectives, accuracy requirements, and application scenarios. Use a high-resolution industrial camera to capture clear images of the automotive PCB drilling area. Collect samples of various types and quality levels of automotive PCB drilling, and carefully label them, including normal drilling, mild burrs, and severe burrs. S2. Data Preprocessing: All collected images are formatted, resized, and brightness balanced. Color images are converted to grayscale to reduce color information interference. Median filtering is applied to remove image noise, and edge detection algorithms are used to enhance drill hole edges and burr features. S3. Region Segmentation and Seed Point Selection: Based on preset conditions (brightness threshold and shape features), the burr-free portion of the drilled hole is automatically selected as the seed point for the "normal" region. Starting from this seed point, the region is gradually expanded based on a pixel similarity threshold according to a specific growth rule to distinguish the normal surface around the drilled hole from the abnormal burr region. S4. 3D Scanning and Analysis: 3D scan the automotive PCB to obtain an accurate 3D model of the drilled area. De-noise and segment the point cloud data to identify the 3D features of the drilled holes and burrs. For suspected burr areas separated by region growing, the shape, size, and brightness features are further extracted. A support vector machine (SVM) classifier is then used to determine whether these features match those of a burr, thereby confirming whether they are genuine burrs. S5. Burr Location and Assessment: Identify and mark the locations of all identified burr areas in the image. Based on the burr size and distribution, assess the potential impact on the functionality of the automotive PCB. Combined with 3D scanning data, perform multi-dimensional verification to improve detection accuracy and completeness. S6. Result Output and Feedback: Test results are output in the form of graphics and reports, indicating the test status of each drill hole, including the presence and location of burrs. These results are then fed back to the production system, allowing unqualified automotive PCBs to be removed and repaired. S7. Report generation and decision support: Summarize test data, generate detailed test reports, provide visual data analysis, and provide a scientific basis for production management and quality improvement.

2. The method for detecting burrs on automobile circuit boards based on region growing according to claim 1, characterized in that ； In step S3, based on the preset conditions: brightness threshold and shape features, the burr-free part in the drilling area is automatically selected as the seed point of the "normal" area, as follows; Automatic seed point selection strategy: Brightness threshold method: Based on the image grayscale histogram, an appropriate threshold is selected to distinguish between dark and bright areas in the drill hole area. Based on this, the center point of the burr-free area is found as the seed point, usually the dark and uniform part inside the drill hole. Shape feature analysis: In addition to brightness, shape features can also be considered to assist in seed point selection. For example, using the image contour after edge detection, the circular outline of the drill hole is identified, and the center point and points close to the center without burr interference are selected as seeds. Region growing algorithm: Similarity measurement: The key to the region growing process is to define an appropriate pixel similarity metric. Commonly used distance metrics include grayscale difference, texture similarity, or color histogram. The algorithm starts from a seed point and gradually adds adjacent pixels that meet the similarity threshold to the current region until no new pixels can be added or the preset stop condition is reached. Multi-scale and multi-directional expansion: To adapt to burrs of different sizes and directions, the region growing algorithm can be designed to expand at multiple scales or in multiple directions, that is, starting at different scales or from multiple initial seed points to adapt to complex drill hole edges and burr shapes; Abnormal region identification: Distinguishing between normal and abnormal regions. The regions obtained through region growing are characterized by a continuous, large area with relatively consistent grayscale. The burr region is characterized by a small area, high brightness, or irregular shape that is significantly different from the normal region. By comparing the regional features: area, boundary roughness, and grayscale variance, the normal drilling surface and the burr region can be effectively distinguished. Real-time feedback and adjustment: The region segmentation and seed point selection processes need to provide feedback to the preprocessing steps: edge enhancement and subsequent 3D analysis. Parameters such as similarity threshold and seed point location can be dynamically adjusted to accommodate changes in different automotive PCB materials and drilling processes, ensuring robustness and accuracy of detection.

3. The method for detecting burrs on automobile circuit boards based on region growing according to claim 1, characterized in that ； In step S4, a three-dimensional scan is performed on the automobile circuit board to obtain an accurate three-dimensional model of the drilling area. The specific steps are as follows: Application of 3D scanning technology: First, 3D scanning technology is used to perform a full-scale scan of the drilling area of the automotive circuit board. 3D scanning emits a laser light source, captures the reflected light information, and uses triangulation principles or structured light technology to reconstruct a 3D point cloud model of the drilling area. This accurately captures subtle changes in the drilled surface, including the raised shape of burrs. Point cloud data processing: The original point cloud data contains a large number of noisy points: ambient light interference, sensor misreading, etc. The algorithm filters out these irrelevant points and retains the information reflecting the actual physical structure. Segmentation: In order to analyze each drill hole and the burrs around it separately, the point cloud is segmented. This is achieved by the distance-based DBSCAN clustering algorithm. The point cloud is divided into different clusters according to the spatial distance between points. Each cluster represents an independent drill hole area. 3D feature recognition: For each drilling area, the 3D features of the burr are further extracted, including the burr's location, size, shape (height, width, volume, and direction). These features are obtained by analyzing the local geometric properties of the point cloud data. Application of shape, size, and brightness features: In addition to geometric features, considering that the brightness of burrs is different from that of the surrounding normal surfaces, brightness information is also extracted from the scanning data as an auxiliary feature. SVM classification: After feature extraction, support vector machines are used to identify and classify burrs. The optimal hyperplane is searched to maximize the spacing between feature vectors of different categories. A large number of labeled samples, including normal drilling, mild burrs, and severe burrs, are used to train the SVM in advance. For suspected burr areas initially identified through region growing, their 3D features are input into the trained SVM model. The model then determines whether the area is indeed a burr based on these features.

4. A system for detecting burrs in automobile circuit boards based on region growing, characterized in that: The automotive circuit board drilling burr detection system based on region growing is used to implement the automotive circuit board drilling burr detection method based on region growing according to any one of claims 1 to 3; the automotive circuit board drilling burr detection system based on region growing comprises: an image acquisition and processing module, a data management and annotation module, a region segmentation and feature extraction module, a burr recognition and classification module, a burr location and evaluation module, and a result output and feedback module; The image acquisition and processing module, which includes high-resolution industrial cameras and 3D scanning equipment, is responsible for capturing high-definition images and 3D point cloud data of the drilling area of the automotive circuit board. It also performs image format conversion, size adjustment, brightness balance, grayscale conversion, median filtering to remove noise, and edge enhancement operations, providing high-quality image input for subsequent analysis. Data management and annotation module: stores various annotated drilling image data, including normal drilling, light burr, and heavy burr samples, for detailed annotation of new samples, supporting manual and semi-automatic annotation; Region Segmentation and Feature Extraction Module: This module automatically identifies and selects burr-free areas within the drilled area as seed points based on brightness thresholds and shape features. Using these seed points, it gradually expands the area using pixel similarity thresholds to distinguish between normal surfaces and burred areas. It then denoises and segments the 3D scan data, identifies the 3D features of the drilled holes and burrs, and fuses them with the 2D image analysis results. Burr recognition and classification module: extracts shape, size, and brightness features from suspected burr areas, and applies support vector machines (SVM) to classify the extracted features to determine whether they are true burrs; Burr Location and Evaluation Module: Accurately locates the positions of all identified burr areas in the image and evaluates their potential impact on the function of the automotive circuit board based on the size and distribution of the burrs; Result output and feedback module: organizes the test results into graphic and text reports, including detailed information on drilling status and burr location, and feeds the test results back to the production management system in real time to guide the removal and repair of unqualified products.

Images