PCB board process defect tracing system based on image recognition

Abstract

This application relates to the technical field of manufacturing processes and discloses a PCB board process defect tracing system based on image recognition. The system includes an image acquisition module for real-time acquisition of PCB board surface image data after key process steps; a defect detection module for receiving preprocessed image data from the image acquisition module, extracting feature regions on the PCB board surface, identifying these feature regions, marking feature regions that meet defect characteristics as suspected defect regions, identifying and classifying defect types for PCB boards with process defects; a tracing association module for calculating the probability of defect types occurring in each process step based on defect types and historical data; a tracing result output module for generating a defect tracing report based on defect probabilities and visually displaying the defect source; and a data storage and management module for storing PCB board inspection data and supporting data query and retrospective analysis.

Description

Technical Field

[0001] This application relates to the technical field of manufacturing processes, and in particular to a PCB board process defect tracing system based on image recognition. Background Technology

[0002] PCB boards are core components of electronic products, and their manufacturing process is complex and lengthy, involving multiple key steps such as drilling, electroplating, photolithography, etching, and solder mask application. During production, due to equipment malfunctions, environmental changes, or improper operation, PCB boards are highly susceptible to various process defects. These include poor conductivity caused by misaligned holes or inconsistent hole diameters, uneven electroplating, soldering reliability issues due to uneven pads, short circuits or open circuits caused by incomplete etching, and aesthetic and protective defects such as blurred characters and damaged solder mask. These defects not only affect the performance and aesthetics of the PCB board but can also lead to electronic product failures in practical applications, increasing manufacturing costs and rework rates. Therefore, controlling PCB board quality is crucial.

[0003] Traditional inspection methods rely heavily on manual visual inspection or offline sampling inspection. These two methods have revealed significant limitations when facing the demands of modern large-scale, high-precision production. Manual visual inspection is not only slow and difficult to match the large-volume inspection needs of production lines, but also prone to visual fatigue from prolonged operation, leading to the overlooking of minute and hidden defects, resulting in missed or incorrect detections and inaccurate feedback. Furthermore, different inspectors may have different judgments about the same defect, affecting the consistency and reliability of the inspection results. In addition, while offline sampling inspection improves accuracy with the help of testing equipment, it can only cover a small number of samples and cannot comprehensively reflect the quality status of the entire batch of products. Especially for occasional, randomly distributed defects, the risk of missed detection is extremely high, and comprehensive defect detection is impossible.

[0004] In addition, traditional detection methods lack a standardized defect feature identification system, making it impossible to accurately classify defects of different regions and types, and thus failing to improve the reliability of detection results. Therefore, there is an urgent need for an intelligent PCB board defect detection system to improve the accuracy and efficiency of detection. Summary of the Invention

[0005] To address at least one of the above problems, this application provides a PCB board process defect tracing system based on image recognition.

[0006] This application provides a PCB board process defect tracing system based on image recognition, including:

[0007] The image acquisition module is used to acquire surface image data of the PCB board in real time after key process steps in the PCB production line.

[0008] The defect detection module is used to perform defect analysis on the surface image data, perform secondary screening on areas with suspected defects, verify the defects based on their continuity, and classify the defect types.

[0009] The source tracing and association module is used to calculate the probability of the defect type occurring in each process step based on the defect type and historical data.

[0010] The source tracing result output module is used to generate a defect source tracing report based on the defect probability and to visually display the source of the defect;

[0011] The data storage and management module is used to store the PCB board's testing data and supports data querying and backtracking analysis.

[0012] As a further technical solution, the method for performing defect analysis includes:

[0013] Receive preprocessed image data output from the image acquisition module and extract feature regions on the PCB board surface;

[0014] Based on the feature regions, the feature regions that meet the defect characteristics are marked as suspected defect regions;

[0015] The feature areas include the circuit area, pad area, solder mask area, hole area, character area and board surface area of ​​the PCB board;

[0016] The defect features include:

[0017] Characteristics of defects in circuit lines, pads, solder mask, holes, characters, and board surfaces;

[0018] A secondary screening is performed on the suspected defect areas to eliminate false defects and obtain the real defect areas.

[0019] Based on the actual defect area data, the defects are classified and defect type labels are output.

[0020] As a further technical solution, the method for secondary screening of suspected defective areas includes:

[0021] The system calls a preset PCB board standard image feature library, compares the feature parameters of the suspected defect area with the parameters of the corresponding area in the standard image, and calculates the feature similarity.

[0022] A similarity threshold is set. When the similarity between a suspected defect area and a standard area is higher than the similarity threshold, it is determined to be a false defect and removed.

[0023] When the similarity is below the threshold, further verification is performed by combining the continuity of defects to confirm that the area is a real defect area.

[0024] Simultaneously, the characteristic parameters during the screening process are recorded for subsequent source tracing analysis.

[0025] As a further technical solution, the method for verifying by combining defect continuity includes:

[0026] Extract the edge contour features and pixel distribution features of the suspected defective region;

[0027] Multiple and densely packed pixels are set along the PCB production transport direction and the board outline, and the pixel connectivity and outline integrity of the defect features are judged based on the collected pixels.

[0028] If the collected pixel results show that the defect features have continuous pixel connectivity and the outline is uninterrupted, then the defect is determined to have defect continuity and is confirmed as a real defect area.

[0029] If the collected pixel results show that the defect features are discrete points, broken contours and no effective extension, they are judged as pseudo defects or defects to be verified based on the similarity value range.

[0030] For the defects to be reviewed, the manual review process is automatically triggered, and the manual review results are transmitted to the PCB board standard image feature library for updating and calibration.

[0031] As a further technical solution, the method for calculating the probability of the defect type occurring in each process stage includes:

[0032] It retrieves historical inspection data, process parameter records for each process step, and defect statistics stored in the data storage and management module;

[0033] Extract historical defect cases that are consistent with the current defect type, and count the number of times the defect type has occurred in each key process step and the total number of defects.

[0034] Based on the similarity matching value, a preset weight is assigned to the historical statistical proportion and the similarity matching value;

[0035] The defect probability of each process step is calculated based on the preset weights to obtain the defect probability of each process step and the current defect, and then sorted from high to low probability.

[0036] As a further technical solution, the source tracing and association module specifically includes:

[0037] The defect feature extraction unit is used to extract the shape features and location distribution features of the defects from the real defect area;

[0038] A process feature library, which pre-stores standard process parameters and tolerance ranges for each key process step, as well as feature templates of defects that are prone to occur in each step;

[0039] The probability calculation unit is used to perform similarity matching between the defect features of the current defect and the defect feature templates of each process step, and to calculate the probability of the defect originating from each process step by combining the statistical distribution of the defect type in the historical data of the batch of products.

[0040] As a further technical solution, the source tracing and association module also includes an association analysis unit;

[0041] The correlation analysis unit is used to statistically analyze the number of defective PCBs produced in each process step and the overall defect rate in real time.

[0042] When the number of defective PCBs or the overall defect rate exceeds a preset qualified threshold, the production line self-inspection is automatically triggered.

[0043] The production line self-inspection includes retrieving real-time operating parameters, equipment status data, and defect tracing data for each process step;

[0044] By comparing the tolerance range of standard process parameters with the normal operating threshold of the equipment, the fault node of the production line that causes too many defects is located, and the fault node location result and parameter deviation details are output.

[0045] As a further technical solution, the specific working steps of the system include:

[0046] S1. Start the system. The image acquisition module collects real-time image data of the PCB board surface at each key process stage of the PCB production line and transmits it to the defect detection module.

[0047] S2. Perform defect analysis based on the PCB board surface image data and identify the defect type. If no defect is detected, store the detection result in the data storage and management module; if a defect is detected, transmit the defect type information to the source tracing and association module.

[0048] S3. Calculate the defect probability of each process step according to the defect type, and generate source tracing and correlation data;

[0049] S4. Generate a visual traceability report based on the traceability association data, showing the defect type, suspected defect source process, and defect probability ranking;

[0050] S5. Store all test data, traceability data, and traceability reports.

[0051] In summary, this application includes at least one of the following beneficial technical effects:

[0052] 1. This invention identifies defects using surface image data of PCB boards. For suspected defect data, a multi-level defect identification mechanism is constructed. Secondary screening is performed based on similarity comparison with a standard image feature library and defect continuity verification, effectively eliminating interference within the allowable range of normal production errors and reducing false positive and false negative rates. Simultaneously, this invention implements a dynamic update mechanism during the detection process, continuously optimizing the identification model to ensure accurate and reliable detection results, meeting the quality control needs of large-scale production lines.

[0053] 2. This invention extracts the shape and location features of defects, matches the defects with common defect feature templates from each process stage, and statistically evaluates the probability of defects originating from each stage based on historical defect data. Finally, it outputs suspected defect sources in probability-sorted order. This achieves defect traceability in PCB manufacturing processes, providing accurate data support for process optimization and troubleshooting.

[0054] 3. This invention sets up a correlation analysis unit to count the number of defects and the overall defect rate in each process step in real time. When the defect rate exceeds a preset threshold, it automatically triggers the production line self-inspection, quickly locates the fault node and outputs parameter deviation details, avoids the generation of batch defective products, and significantly improves production efficiency and product qualification rate. Attached Figure Description

[0055] Figure 1 This is a structural diagram of a PCB board process defect tracing system based on image recognition. Detailed Implementation

[0056] The embodiments of this application are described in detail below, and examples of the embodiments are shown in the accompanying drawings.

[0057] In the description of this specification, the references to "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples" refer to specific features, structures, materials, or characteristics described in connection with the described embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0058] This application discloses an image recognition-based PCB board process defect tracing system. This embodiment is applied to a PCB mass production line in an electronics manufacturing plant. The core processes of the production line include drilling → electroplating → photolithography → etching → solder masking. Each batch produces 5000 PCB boards. To achieve accurate defect tracing and production line anomaly early warning, an image recognition-based PCB board process defect tracing system is adopted. Figure 1 As shown, it includes:

[0059] The image acquisition module is used to acquire surface image data of PCB boards in real time after key process steps in the PCB production line, providing basic data for subsequent defect detection.

[0060] The defect detection module is used to perform defect analysis on the surface image data, and to identify and classify the defect types of PCB boards with process defects.

[0061] The source tracing and association module is used to calculate the probability of the defect type occurring in each process step based on the defect type and historical data.

[0062] The source tracing result output module is used to generate a defect source tracing report based on the defect probability and to visually display the source of the defect;

[0063] The data storage and management module is used to store the PCB board's testing data and supports data querying and backtracking analysis.

[0064] In this embodiment, based on the above modules, the working process of the system includes the following steps:

[0065] S1. Start the system. The image acquisition module collects real-time image data of the PCB board surface at each key process stage of the PCB production line and transmits it to the defect detection module.

[0066] In key processes of the PCB production line, such as drilling, electroplating, photolithography, etching, and solder masking, one primary and one backup image acquisition point are set up. When the primary image acquisition point fails, the backup image acquisition point is automatically activated to avoid image acquisition omissions caused by single point failure. Each image acquisition point is equipped with a high-resolution industrial camera and an adjustable-angle LED illumination source. By adjusting the light source angle and illumination intensity, it is ensured that the illumination is uniform, without shadows or reflections, and the detection response time is ≤0.1ms, which can detect the transmission position of the PCB board in real time.

[0067] The PCB board is detected by a photoelectric sensor. When the PCB board reaches the image acquisition point, the industrial camera is immediately triggered to take real-time pictures and collect high-definition image data of the PCB board surface. After the acquisition is completed, the acquired surface image data is sent to the defect detection module.

[0068] S2. Perform defect analysis based on the PCB board surface image data and identify the defect type. If no defect is detected, store the detection result in the data storage and management module; if a defect is detected, transmit the defect type information to the source tracing and association module.

[0069] The received image is preprocessed, including grayscale conversion, geometric correction, and noise reduction filtering to remove invalid image information;

[0070] The preprocessed image is subjected to contour detection using an edge detection algorithm to accurately extract feature regions on the PCB board, including circuit regions, pad regions, solder mask regions, hole regions, character regions, and board surface regions. The matching degree between the extracted region contours and the actual regions is ≥99%.

[0071] For each feature region, a pre-defined classification model is used to identify defects, and regions that meet the defect characteristics are marked as suspected defect regions.

[0072] In this embodiment, the method for performing defect analysis includes:

[0073] Receive preprocessed image data output from the image acquisition module and extract feature regions on the PCB board surface;

[0074] Based on the feature regions, the feature regions that meet the defect characteristics are marked as suspected defect regions;

[0075] The feature areas include the circuit area, pad area, solder mask area, hole area, character area and board surface area of ​​the PCB board;

[0076] The defect features include:

[0077] Characteristics of defects in circuit lines, pads, solder mask, holes, characters, and board surfaces;

[0078] A secondary screening is performed on the suspected defect areas to eliminate false defects and obtain the real defect areas.

[0079] If no defects are detected, the test results will be stored in the data storage and management module.

[0080] If a defect is detected, the information of the actual defect area is transmitted to the source tracing and association module for defect source tracing.

[0081] Based on the actual defect area data, the defects are classified and defect type labels are output;

[0082] In this embodiment, the method for secondary screening of suspected defective areas includes:

[0083] The system calls a preset PCB board standard image feature library, compares the feature parameters of the suspected defect area with the parameters of the corresponding area in the standard image, and calculates the feature similarity.

[0084] The method for calculating feature similarity includes:

[0085]

[0086] in, For feature similarity, The first of the suspected defective areas Each feature parameter value, The first region corresponding to the standard image Each feature parameter value, The total number of types of feature parameters;

[0087] In this embodiment, five feature parameters are selected as references: contour perimeter, region area, pixel grayscale mean, grayscale variance, and aspect ratio.

[0088] In this embodiment, the similarity threshold is set to 95%, that is... ≥0.95, when the similarity between the suspected defect area and the standard area is higher than the similarity threshold of 95%, it is judged as a false defect and removed;

[0089] When the similarity between the suspected defect area and the standard area is less than 95% of the similarity threshold, further verification is performed by combining the defect continuity to confirm that the area is a real defect area.

[0090] In this embodiment, taking a suspected defect area in the hole location region as an example, five feature parameters are selected. The parameter values ​​and the parameter values ​​of the corresponding areas in the standard image are shown in Table 1. Substituting them into the formula, the feature similarity is calculated. The feature parameters of the suspected defect in the hole location region and the standard region are shown in the following table:

[0091] Feature parameters Outline perimeter Area Pixel grayscale average Gray variance Aspect Ratio Suspected defects 8.2 5.1 180 35 1.1 Standard Area 6.0 4.0 200 20 1.0

[0092] Calculate the feature similarity of the suspected defective regions. =0.68, which is less than the similarity threshold of 95%, therefore further verification of defect continuity is required;

[0093] In this embodiment, the method for verifying by combining defect continuity includes:

[0094] Extract the edge contour features and pixel distribution features of the suspected defective region;

[0095] Multiple dense pixels are set along the PCB production transport direction and the board outline, with a spacing of 0.01mm between each pixel. The pixel connectivity and outline integrity of the defect features are judged based on the collected pixels.

[0096] The verification rules include:

[0097] If the collected pixel results show that the defect features have continuous pixel connectivity and the outline is uninterrupted, then the defect is determined to have defect continuity and is confirmed as a real defect area.

[0098] If the collected pixel results show that the defect features are discrete points, broken contours and no effective extension, they are judged as pseudo defects or defects to be verified based on the similarity value range.

[0099] In this embodiment, the similarity value range is:

[0100] When feature similarity 0.8 ≤ When the value is less than 0.95, the defect is a defect that needs to be reviewed.

[0101] When feature similarity When the value is less than 0.8, the defect is considered a true defect.

[0102] For the defects to be reviewed, a manual review process is automatically triggered, where quality inspection personnel make manual judgments through a visual interface. The results of the manual review are transmitted to the PCB board standard image feature library in real time to update and calibrate the data in the library, thereby improving the accuracy of subsequent model recognition.

[0103] Simultaneously record the characteristic parameters during the screening process for subsequent source tracing analysis;

[0104] Defect identification is performed using surface image data of PCB boards. Suspected defect areas undergo secondary screening to effectively eliminate interference within the allowable range of normal production errors, reducing false positive and false negative rates. Simultaneously, a dynamic update mechanism is implemented during the inspection process to continuously optimize the identification model, ensuring accurate and reliable inspection results and meeting the quality control needs of large-scale production lines.

[0105] S3. Calculate the defect probability of each process step according to the defect type and generate source tracing association data; obtain historical detection data, process parameter records of each process step and defect statistics data from the data storage and management module, and use them to calculate the defect probability of the defect type in each process step according to the defect type and historical data.

[0106] In this embodiment, the method for calculating the probability of the defect type occurring in each process stage includes:

[0107] It retrieves historical inspection data, process parameter records for each process step, and defect statistics stored in the data storage and management module;

[0108] Extract historical defect cases that are consistent with the current defect type, and count the number of times the defect type has occurred in each key process step and the total number of defects.

[0109] Based on the similarity matching value, a preset weight is assigned to the historical statistical proportion and the similarity matching value;

[0110]

[0111] in, This represents the defect probability. Defect types in historical data summary In the process Number of times it appears Indicates the defect type in the historical data summary Total number of occurrences , These are the weighting coefficients;

[0112] In this embodiment, taking a defect in the hole location area as an example, historical data is retrieved from the data storage and management module to show the total number of times this defect type has occurred in historical production. =1000 times; =0.6, =0.4;

[0113] The probability of this defect in each process stage is as follows:

[0114] The defect probability in the drilling process is: =0.96;

[0115] The defect probability in the electroplating process is =0.054;

[0116] The defect probability in the photolithography process is =0.035;

[0117] The defect probability in the etching process is =0.046;

[0118] The defect probability in the solder resist process is: =0.095;

[0119] The source tracing and association module specifically includes:

[0120] The defect feature extraction unit is used to extract the shape features of defects from the real defect area using image feature extraction algorithms, such as the contour, area, aspect ratio and location distribution features of the defects. The accuracy of the extracted feature parameters is consistent with the image acquisition accuracy.

[0121] The process feature library pre-stores the standard process parameters and tolerance ranges of the five key process steps: drilling, electroplating, photolithography, etching, and solder resist, as well as feature templates of defects that are prone to occur in each step.

[0122] The standard process parameters and their tolerance ranges for key process steps are shown in the table below:

[0123] Process Standard process parameters Tolerance range drilling Drilling speed 30000 r / min; ±5% plating Current density 2A / dm²; electroplating time 20min ±3% Photolithography Exposure energy: 800 mJ / cm²; Development time: 60 s ±2% Etching Etching temperature: 45℃; Etching time: 90s ±2% Solder mask Coating thickness 30μm; curing temperature 150℃ ±5%

[0124] The probability calculation unit, as the core module of the defect tracing system, mainly functions to match the currently detected defect features, such as hole offset and uneven hole size, with the pre-built defect feature templates of each process step, and calculate the probability of the defect originating from each process step by combining the statistical distribution of the defect type in the historical data of the batch of products.

[0125] The defect probability of each process step is calculated according to the preset weight, and the defect probability of each process step and the current defect is obtained and sorted from high to low probability.

[0126] Based on the above calculations, the defect probability ranking for each process step is as follows: drilling (96.2%) > solder resist (9.5%) > electroplating (5.4%) > etching (4.6%) > photolithography (3.5%).

[0127] Therefore, it can be determined that the suspected source of the hole position deviation defect is the drilling process;

[0128] By matching defects with the characteristic templates of common defects in each process step, and combining the statistical evaluation of historical defect data to determine the probability of defects originating from each step, the system finally outputs the suspected defect sources in a probability-sorted manner. This enables the PCB manufacturing process to trace defects, providing accurate data support for process optimization and troubleshooting.

[0129] In actual production, when the number of defects in a batch of products remains consistently high, it will directly lead to a decrease in the overall product pass rate and affect the smooth execution of the production plan. If defect tracing is carried out frequently, multiple possible process steps need to be checked and verified each time, which will significantly lengthen the production cycle and thus affect the normal production rhythm and progress.

[0130] Therefore, when faced with an abnormal situation with an excessive number of defects, we should not only focus on the source analysis of PCB board defects, but also consider whether there is a problem with the production line equipment, such as equipment aging, parameter errors, or other abnormalities.

[0131] To verify this issue, in this embodiment, the source tracing and association module further includes an association analysis unit;

[0132] The correlation analysis unit is used to statistically analyze the number of defective PCBs produced in each process step and the overall defect rate in real time.

[0133] Methods for calculating defect rates include:

[0134]

[0135] in, For defect rate, This represents the number of defective PCB boards. This refers to the total number of PCB boards produced.

[0136] In this embodiment, the defect rate pass threshold for each process step is set at 1%, and the overall defect rate pass threshold for a single batch is set at 1%.

[0137] When the number of defective PCBs or the overall defect rate exceeds a preset qualified threshold, the production line self-inspection is automatically triggered.

[0138] During the production line self-inspection process, the system will retrieve the real-time operating parameters, equipment status data and defect source tracing data of each process link. By comparing the standard process parameter tolerance range in the process link feature library with the normal operating threshold of the equipment, the system will locate the production line fault node that causes too many defects and output the fault node location result and parameter deviation details.

[0139] By statistically analyzing the number of defects and the overall defect rate in each process step in real time, the production line self-inspection is automatically triggered when the defect rate exceeds the preset threshold, quickly locating the fault node, greatly avoiding the generation of batches of defective products, and significantly improving production efficiency and product qualification rate.

[0140] When multiple defects of different types point to the same equipment or the same time period, the correlation analysis unit can issue an early warning, indicating that there may be a equipment failure. This guides staff to prioritize checking production equipment and avoids the inefficient pitfall of checking each single defect one by one.

[0141] S4. Generate a visual traceability report based on the traceability association data, showing the defect type, suspected defect source process, and defect probability ranking;

[0142] A defect source tracing report is generated based on the defect probability, and the source of the defect is visualized.

[0143] The report presents the defect types, possible source processes, and their probability ranking in chart form. The report can be displayed in real time on the workshop management screen via a web interface, or exported as a PDF or Excel file for quality management personnel to view and archive.

[0144] S5. Classify and store the entire process data of this testing and traceability, and support data retrieval, backup, update, and deletion. Specific data types stored include:

[0145] Image acquisition data, such as high-resolution images of the PCB board surface, acquisition time, acquisition points, and device parameters;

[0146] Defect detection data, such as preprocessed image data, feature region parameters, suspected defect region data, actual defect region data, and defect type labels;

[0147] Tracing back to related data, such as defect characteristic parameters, the calculation process and results of defect probability in each process step, production line process parameter records, equipment status data, and fault node location results;

[0148] Source tracing results data, such as the original file, exported file, and screenshots of the visualized source tracing report;

[0149] Historical statistical data, such as the defect rate of each batch, the historical occurrence frequency of each defect type, and the failure records of each process step.

[0150] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A PCB board process defect tracing system based on image recognition, characterized in that, include: The image acquisition module is used to acquire surface image data of the PCB board in real time after key process steps in the PCB production line. The defect detection module is used to perform defect analysis on the surface image data, perform secondary screening on areas with suspected defects, verify the defects based on their continuity, and classify the defect types. The source tracing and association module is used to calculate the probability of the defect type occurring in each process step based on the defect type and historical data. The source tracing result output module is used to generate a defect source tracing report based on the defect probability and to visually display the source of the defect; The data storage and management module is used to store the PCB board's testing data and supports data querying and backtracking analysis.

2. The PCB board process defect tracing system based on image recognition according to claim 1, characterized in that, The method for performing defect analysis includes: Receive preprocessed image data output from the image acquisition module and extract feature regions on the PCB board surface; Based on the feature regions, the feature regions that meet the defect characteristics are marked as suspected defect regions; The feature areas include the circuit area, pad area, solder mask area, hole area, character area and board surface area of ​​the PCB board; The defect features include: Characteristics of defects in circuit lines, pads, solder mask, holes, characters, and board surfaces; A secondary screening is performed on the suspected defect areas to eliminate false defects and obtain the real defect areas. Based on the actual defect area data, the defects are classified and defect type labels are output.

3. The PCB board process defect tracing system based on image recognition according to claim 2, characterized in that, The method for secondary screening of suspected defective areas includes: The system calls a preset PCB board standard image feature library, compares the feature parameters of the suspected defect area with the parameters of the corresponding area in the standard image, and calculates the feature similarity. A similarity threshold is set. When the similarity between a suspected defect area and a standard area is higher than the similarity threshold, it is determined to be a false defect and removed. When the similarity is below the threshold, further verification is performed by combining the continuity of defects to confirm that the area is a real defect area. Simultaneously, the characteristic parameters during the screening process are recorded for subsequent source tracing analysis.

4. The PCB board process defect tracing system based on image recognition according to claim 3, characterized in that, The method for verification based on defect continuity includes: Extract the edge contour features and pixel distribution features of the suspected defective region; Multiple and densely packed pixels are set along the PCB production transport direction and the board outline, and the pixel connectivity and outline integrity of the defect features are judged based on the collected pixels. If the collected pixel results show that the defect features have continuous pixel connectivity and the outline is uninterrupted, then the defect is determined to have defect continuity and is confirmed as a real defect area. If the collected pixel results show that the defect features are discrete points, broken contours and no effective extension, they are judged as pseudo defects or defects to be verified based on the similarity value range. For the defects to be reviewed, the manual review process is automatically triggered, and the manual review results are transmitted to the PCB board standard image feature library for updating and calibration.

5. The PCB board process defect tracing system based on image recognition according to claim 1, characterized in that, The method for calculating the probability of the defect type occurring in each process stage includes: It retrieves historical inspection data, process parameter records for each process step, and defect statistics stored in the data storage and management module; Extract historical defect cases that are consistent with the current defect type, and count the number of times the defect type has occurred in each key process step and the total number of defects. Based on the similarity matching value, a preset weight is assigned to the historical statistical proportion and the similarity matching value; The defect probability of each process step is calculated based on the preset weights to obtain the defect probability of each process step and the current defect, and then sorted from high to low probability.

6. The PCB board process defect tracing system based on image recognition according to claim 1, characterized in that, The source tracing and association module specifically includes: The defect feature extraction unit is used to extract the shape features and location distribution features of the defects from the real defect area; A process feature library, which pre-stores standard process parameters and tolerance ranges for each key process step, as well as feature templates of defects that are prone to occur in each step; The probability calculation unit is used to perform similarity matching between the defect features of the current defect and the defect feature templates of each process step, and to calculate the probability of the defect originating from each process step by combining the statistical distribution of the defect type in the historical data of the batch of products.

7. A PCB board process defect tracing system based on image recognition according to claim 6, characterized in that, The source tracing and association module also includes an association analysis unit; The correlation analysis unit is used to statistically analyze the number of defective PCBs produced in each process step and the overall defect rate in real time. When the number of defective PCBs or the overall defect rate exceeds a preset qualified threshold, the production line self-inspection is automatically triggered. The production line self-inspection includes retrieving real-time operating parameters, equipment status data, and defect tracing data for each process step; By comparing the tolerance range of standard process parameters with the normal operating threshold of the equipment, the fault node of the production line that causes too many defects is located, and the fault node location result and parameter deviation details are output.

8. The PCB board process defect tracing system based on image recognition according to claim 1, characterized in that, The specific working steps of the system include: S1. Start the system. The image acquisition module collects real-time image data of the PCB board surface at each key process stage of the PCB production line and transmits it to the defect detection module. S2. Perform defect analysis based on the PCB board surface image data and identify the defect type. If no defect is detected, store the detection result in the data storage and management module; if a defect is detected, transmit the defect type information to the source tracing and association module. S3. Calculate the defect probability of each process step according to the defect type, and generate source tracing and correlation data; S4. Generate a visual traceability report based on the traceability association data, showing the defect type, suspected defect source process, and defect probability ranking; S5. Store all test data, traceability data, and traceability reports.

Images