Circuit board processing abnormity detection system

Abstract

The invention relates to the technical field of anomaly detection, in particular to a circuit board processing anomaly detection system which comprises an image anchoring module, a gray scale comparison module, a candidate area screening module, a gradient closing verification module and a behavior track evolution module. According to the method, by setting the gray scale anchor points in the image frames and collecting the standard gray scale values, deviation accumulation caused by the overall illumination change of the image can be effectively avoided, the response delay area is recognized through inter-frame gray scale comparison, and early intervention on the etching abnormal process is achieved; a gray scale fluctuation low-amplitude segment in an image boundary region is extracted to recognize a texture missing region, the recognition accuracy of structural blur is enhanced, fine extraction of an abnormal structural form is realized in combination with gray scale gradient chain construction and a closed error judgment mechanism, region response change trajectory tracking in a frame sequence is superposed, and the recognition accuracy of structural blur is improved. Therefore, the judgment is not limited to single-frame anomaly detection and has a time sequence evolution recognition capability, an abnormal symptom region is stably captured, and a reliable detection result is generated.

Description

Technical Field

[0001] This invention relates to the field of anomaly detection technology, and in particular to a circuit board processing anomaly detection system. Background Technology

[0002] The field of anomaly detection technology involves the identification, analysis, and response to abnormal states in systems, equipment, or production processes. Core aspects include fault information acquisition methods, fault mode identification paths, and system response strategy settings. It is widely applied in various scenarios such as industrial manufacturing, equipment maintenance, and automation control. Among these, circuit board processing anomaly detection systems refer to systems that detect anomalies caused by equipment errors, process deviations, or material defects during the drilling, etching, and mounting processes of printed circuit boards. Typically, these systems acquire images using optical camera components installed on the processing equipment and use edge feature matching methods to determine anomalies such as processing position offsets or pattern missing parts. Some systems also employ threshold comparison methods based on statistical rules to continuously monitor specific process parameters such as temperature, voltage, and pressure to identify abnormal states exceeding set ranges.

[0003] In existing circuit board processing anomaly detection processes, edge feature matching is the primary method. However, this method is susceptible to variations in pattern complexity and edge clarity. In scenarios with uneven lighting or incomplete patterns, the matching accuracy drops significantly. Furthermore, threshold-based process parameter monitoring methods struggle to effectively respond to structural anomalies at the image level. For instance, when the overall pattern structure is blurred or etching is insufficient but does not exceed the parameter threshold, the system often fails to identify processing deviations in a timely manner. This results in defective circuit board products not being screened out and continuing to flow into the next process, reducing the overall process quality and stability. Summary of the Invention

[0004] To address the technical problems existing in the prior art, embodiments of the present invention provide a circuit board processing anomaly detection system. The technical solution is as follows: On the one hand, a circuit board processing anomaly detection system is provided, the system comprising: The image anchoring module acquires continuous process image frames during the circuit board etching stage, sets positioning anchor points in each frame, collects the standard grayscale values ​​of each anchor point as the first frame reference data, and statistically generates grayscale anchor point reference data. The grayscale comparison module obtains grayscale data at the same position in subsequent image frames based on the coordinates of each anchor point in the grayscale anchor point reference data, performs inter-frame grayscale value comparison operation, identifies anchor point sequences in consecutive frames with differences lower than the etching process monitoring threshold, and obtains the etching response sluggish region. The candidate region filtering module extracts the pixel grayscale sequence of the corresponding image boundary region based on the position of the etch response sluggish region in the image frame, identifies continuous pixel segments whose grayscale value changes are stable in the low dynamic range range, determines them as structurally blurred fragment regions, and obtains blurred structural candidate fragments. The gradient closure verification module extracts the grayscale difference between adjacent pixels based on the pixel grayscale sequence in the fuzzy structure candidate segment, performs absolute value extraction on the grayscale difference between the beginning and end of the gradient chain and performs error judgment with the gradient closure tolerance, records the segment region whose error exceeds the gradient closure tolerance as an abnormal structure region, and generates an abnormal region of structure chain closure.

[0005] As a further aspect of the present invention, the grayscale anchor reference data includes anchor position index, standard grayscale level, and pattern boundary mapping relationship; the etch response sluggish region includes inter-frame grayscale change sequence, low-variation anchor point set, and threshold trigger marker; the blurred structure candidate segment includes low dynamic range grayscale sequence, continuous pixel segment index, and image boundary region identifier; and the structural chain closure anomalous region includes gradient chain beginning and end grayscale difference anomalous segments, grayscale closure error record, and anomalous structural segment label.

[0006] As a further aspect of the present invention, the standard grayscale value is specifically the grayscale level defined by the international standard for electronic circuit board acceptance; the etching process monitoring threshold is specifically the allowable range of grayscale changes defined by the rigid printed circuit board qualification and performance specifications; and the low dynamic range interval is specifically the grayscale fluctuation range defined by the international standard for image coding.

[0007] As a further aspect of the present invention, the image anchoring module includes: The grayscale acquisition submodule acquires image frames of continuous processes during the circuit board etching stage. In each image frame, multiple pixels within the pattern boundary are set as positioning anchor points. The standard grayscale value of each anchor point is acquired, and interference pixels at the image edge are removed to generate a set of image anchor point grayscale values. The reference grayscale setting submodule selects the first frame data as the reference image frame according to the time series based on the grayscale values ​​of each frame anchor point in the image anchor grayscale value set, extracts the standard grayscale values ​​of all anchor points in the first frame, calls the standard color card grayscale level reference value for comparison and calibration, filters the effective reference anchor points, and obtains the effective anchor point grayscale reference value. The grayscale reference generation submodule calculates the grayscale change range in subsequent consecutive image frames based on the grayscale values ​​of each anchor point in the effective anchor point grayscale reference values, and statistically analyzes the grayscale mean, maximum deviation, and trend value of each anchor point in all image frames. It then classifies and groups the overall anchor points to generate grayscale anchor point reference data.

[0008] As a further aspect of the present invention, the grayscale contrast module includes: The image grayscale extraction submodule obtains the coordinate information of all anchor points in the grayscale anchor point reference data, collects the grayscale values ​​of the corresponding coordinate positions in subsequent image frames, selects continuous frame images with the same frame sequence length as the reference data, locates and extracts the grayscale values ​​of all anchor point coordinates in each frame image, and generates anchor point grayscale sequence values. The anchor point frame difference determination submodule calculates the gray level difference between each anchor point in the gray level sequence value of the anchor point in the adjacent frames. If the absolute value does not exceed the etching process monitoring threshold, the anchor point is marked as a continuous and consistent frame point. The set of anchor points that meet the continuous and consistent condition is extracted to obtain the gray level consistent anchor point sequence. The response anomaly identification submodule performs spatial clustering based on the coordinates of anchor point regions in the grayscale consistent anchor point sequence where the difference between consecutive frames is lower than the etching process monitoring threshold. It identifies continuous consistent anchor point groups that appear repeatedly in multiple frames, determines whether the proportion of the anchor point group in the total anchor points is higher than the etching uniformity threshold, and marks it as a response sluggish region if it is higher. It then establishes a sluggish region index map under the corresponding frame sequence to obtain the etching response sluggish region.

[0009] As a further aspect of the present invention, the candidate region filtering module includes: The image region extraction submodule obtains the region index and coordinate information of all corresponding image frames in the etch response sluggish region. In each frame image, it reads the image boundary range corresponding to the region marked by the index, extends outward according to the coordinates to construct the boundary pixel extraction band, traverses all pixels in the boundary band and records the gray level value, arranges the extracted gray level value in row and column order to form a one-dimensional gray level sequence, establishes the gray level boundary record set of all sluggish regions in the corresponding image frame, and generates the boundary gray level sequence set. The grayscale fluctuation detection submodule calculates the difference between the maximum and minimum grayscale values ​​of each pixel sequence in the boundary grayscale sequence set, filters out sequence segments with grayscale fluctuation less than or equal to the grayscale fluctuation judgment threshold, and records the pixel start position and length to generate low dynamic grayscale interval segments. The fuzzy structure recognition submodule, based on the starting coordinates and continuous pixel lengths of all grayscale sequences in the low dynamic grayscale interval, marks sequences with continuous pixel lengths greater than the grayscale fluctuation judgment threshold as candidate segments, calculates the spatial position index of each corresponding region in the image frame, and matches it with the coordinates of the initial stagnant region to obtain fuzzy structure candidate segments.

[0010] As a further aspect of the present invention, the gradient closure verification module includes: The grayscale difference extraction submodule obtains the grayscale value sequence of all pixels in the image frame of the candidate segment of the fuzzy structure, extracts the grayscale difference between two adjacent pixels according to the pixel arrangement order in each segment, records each difference in each gradient chain and the corresponding pixel position, and generates a continuous grayscale gradient chain. The gradient error calculation submodule extracts the gray level difference between the first and last points of each chain based on the continuous gray level gradient chain, calculates the absolute value and records it as the closure difference, filters all segments with closure differences greater than the gradient closure tolerance, records the segment number, the first and last pixel coordinates and the error value, and obtains the first and last gradient error values. The abnormal region determination submodule performs a region index extraction operation on all segments with error values ​​greater than the gradient closure tolerance based on the error data corresponding to each segment in the first and last gradient error values. It counts the coordinate range, number of pixels and gradient chain length of the segment in the image frame, marks the abnormal region number and archives it according to the corresponding frame number, and obtains the structural chain closure abnormal region.

[0011] As a further aspect of the present invention, the system further includes: The behavior trajectory evolution module marks whether the structural chain closure abnormal region appears in different frames according to the corresponding position of the abnormal region in consecutive image frames, tracks the response changes in the image frame index order, determines whether the corresponding region continues to show stable abnormal signs, and generates circuit board processing abnormality detection results.

[0012] As a further aspect of the present invention, the circuit board processing anomaly detection results include anomaly area trajectory index results, anomaly response continuous marking results, and stable anomaly determination identification records.

[0013] As a further aspect of the present invention, the behavior trajectory evolution module includes: The inter-frame marking and recording submodule obtains the frame number and coordinate index information of each abnormal segment in the structural chain closure abnormal region, traverses the continuous image frame sequence, marks whether there are abnormal regions with the same number and the same position in each frame image, constructs a binary sequence according to the frame order, and generates abnormal region frame sequence recording values. The continuous response tracking submodule calculates the length of the continuous existence segment of each abnormal region in the frame sequence based on the inter-frame changes of each record in the abnormal region frame sequence record value. If the length of the continuous existence segment is greater than or equal to the stable segment judgment benchmark length, it is determined to be a response continuous segment. All abnormal region numbers and start and end frame indices that meet the conditions are extracted to obtain the response continuous frame segment interval. The abnormal state determination submodule filters out the region numbers that meet the stable abnormal judgment threshold based on the abnormal region numbers and the length of the continuous frame segment in the response, integrates and statistically analyzes the corresponding coordinate range and response frequency in the image frame, and archives them by frame index to obtain the circuit board processing abnormality detection results.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By setting grayscale anchor points in image frames and collecting standard grayscale values, the accumulation of deviations caused by changes in overall image illumination can be effectively avoided. By comparing grayscale between frames to identify regions with slow response, early intervention in the etching anomaly process can be achieved. By extracting low-amplitude segments of grayscale fluctuations in the image boundary region to identify texture-deficient areas, the accuracy of identifying structural ambiguity can be enhanced. Combined with grayscale gradient chain construction and closure error judgment mechanism, the abnormal structural morphology can be refined. By superimposing the tracking of regional response change trajectories in the frame sequence, the judgment is not limited to single-frame anomaly detection but has the ability to identify temporal evolution. Thus, even under image noise interference and etching process fluctuations, it can still stably capture abnormal sign areas and generate reliable detection results. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of a circuit board processing anomaly detection system provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the image anchoring module of the present invention; Figure 4 This is a flowchart of the grayscale contrast module of the present invention; Figure 5 This is a flowchart of the candidate region screening module of the present invention; Figure 6 This is a flowchart of the gradient closure verification module of the present invention; Figure 7 This is a flowchart of the behavior trajectory evolution module of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a circuit board processing anomaly detection system, such as... Figures 1-2 The schematic diagram shown is of a circuit board processing anomaly detection system. The system includes: The image anchoring module acquires consecutive process image frames during the circuit board etching stage. In each image frame, multiple pixels within the pattern boundary are set as positioning anchor points. The standard grayscale value of each anchor point (grayscale level (0-255 levels) defined by the international standard for electronic circuit board acceptance, used to quantify the image features of the etched area. This standard is calibrated through a standard color chart, and the measurement standard is ΔE≤1.5 (color difference meter measurement value)) is used as the reference data for the first frame, and grayscale anchor point reference data is generated statistically. The grayscale comparison module obtains grayscale data at the same position in subsequent image frames based on the coordinates of each anchor point in the grayscale anchor point reference data. It then performs inter-frame grayscale value comparison operations sequentially to identify anchor point sequences in consecutive frames whose differences are lower than the etching process monitoring threshold (the allowable range of grayscale changes defined by the qualification and performance specifications of rigid printed circuit boards (grayscale difference between adjacent frames ≤ 15), exceeding this value is considered an abnormal response. This threshold is determined based on copper foil etching rate experiments, and the standard is etching uniformity ≥ 95%), thus obtaining the etching response sluggish region. The candidate region filtering module extracts the pixel grayscale sequence of the corresponding image boundary region based on the position of the etch response sluggish region in the image frame, identifies continuous pixel segments whose grayscale value changes are stable in the low dynamic range range (the grayscale fluctuation range (maximum-minimum grayscale value ≤20) defined by the international standard for image coding, which is used to identify regions with missing texture features), and determines them as structurally blurred fragment regions, thus obtaining blurred structural candidate fragments. The gradient closure verification module extracts the gray level difference between adjacent pixels based on the pixel gray level sequence in the fuzzy structure candidate segment, forming a continuous gray level gradient chain. It performs absolute value extraction on the gray level difference between the first and last gray level of the gradient chain and performs error judgment with the gradient closure tolerance (the gradient consistency allowable error defined by the standard (default ±5 gray levels), used to determine the contour closure quality). The segment region with the error exceeding the gradient closure tolerance is recorded as an abnormal structure region, generating an abnormal region of structural chain closure. The behavior trajectory evolution module marks whether the structural chain closure anomaly region appears in different frames based on the corresponding position of the region in consecutive image frames, tracks the response changes in the image frame index order, determines whether the corresponding region continues to show stable abnormal signs, and generates circuit board processing anomaly detection results.

[0023] The grayscale anchor reference data includes anchor position index, standard grayscale level, and pattern boundary mapping relationship. The etch response sluggish region includes inter-frame grayscale change sequence, low-variation anchor point set, and threshold trigger marker. The fuzzy structure candidate segment includes low dynamic range grayscale sequence, continuous pixel segment index, and image boundary region identifier. The structural chain closure abnormal region includes gradient chain start and end grayscale difference abnormal segment, grayscale closure error record, and abnormal structure segment label. The circuit board processing abnormality detection result includes abnormal region trajectory index result, abnormal response continuous marking result, and stable abnormality judgment mark record.

[0024] Specifically, such as Figure 2 , Figure 3 As shown, the image anchoring module includes: The grayscale acquisition submodule acquires image frames of continuous processes during the circuit board etching stage. In each image frame, multiple pixels within the pattern boundary are set as positioning anchor points. The standard grayscale value of each anchor point is acquired, and interference pixels at the image edge are removed to generate a set of image anchor point grayscale values. To acquire continuous process image frames during the circuit board etching stage, an automatic image acquisition device must first capture the image frame sequence on the circuit board process line in real time during the etching process. The image acquisition frequency is set to 10 frames per second, and the total acquisition time is set to 20 seconds, corresponding to acquiring 200 images. During acquisition, the time interval between image frames must be uniform. Each image frame is set to 2048 pixels × 2048 pixels, and the image format is grayscale for subsequent grayscale value extraction. Then, a pattern boundary range is set in each image frame. According to standard circuit diagram size specifications, this boundary can be set to an area within a radius of 500 pixels from the center of the circuit diagram. Within this area, 30 pixels are evenly selected as anchor points. Specifically, anchor points can be distributed by selecting points every 100 pixels. For example, in the 10th frame image, the coordinate positions are selected as (100, 100), (200, 100), (300, 100), and (300, 100). 100) and so on, sequentially incrementing the anchor point positions, and collecting the corresponding grayscale values ​​of each anchor point in the grayscale image. The grayscale values ​​are unsigned integers between 0 and 255. Then, edge interference removal processing is performed on all collected grayscale data. This is done by removing pixels whose distance from the image edge is less than 30 pixels from the boundary to eliminate interference data. The grayscale values ​​of the remaining anchor points are then summarized. For example, after removing 5 edge points in the 10th frame image, the remaining 25 anchor point grayscale values ​​are 220, 218, 215, 219, etc. By comparing these grayscale values ​​with the reference grayscale range under the standard color chart, the ΔE color difference calibration limit is set to ΔE≤1.5. A portable colorimeter is used to measure the color difference between the collected anchor point grayscale and the standard grayscale. If the ΔE value exceeds 1.5, the anchor point value is removed. Finally, the set of all anchor point grayscale values ​​that meet the ΔE value condition is retained as the valid anchor point dataset for that frame image. The formula for calculating ΔE is: ,in , , These are the tristimulus values ​​of the measured color values. , , As a reference value under the standard color chart, this formula can be used to verify the color difference of each anchor point grayscale data, and finally generate the image anchor point grayscale value set.

[0025] The reference grayscale setting submodule is based on the grayscale values ​​of each frame anchor point in the image anchor grayscale value set. It selects the first frame data as the reference image frame according to the time series, extracts the standard grayscale values ​​of all anchor points in the first frame, calls the standard color card grayscale level reference value for comparison and calibration, filters the effective reference anchor points, and obtains the effective anchor point grayscale reference value. Based on the anchor point data of each frame in the image anchor grayscale value set, all image frames must first be arranged in chronological order, for example, from frame 1 to frame 200. The earliest acquired frame 1 is used as the reference image frame. The standard grayscale values ​​of all 30 anchor points in that frame are read. Assuming the grayscale values ​​of each anchor point in frame 1 are 210, 212, 215, 213, 211, etc., the Lab* grayscale standard values ​​from the standard color chart are used to compare the grayscale values ​​of this frame to determine whether each anchor point falls within the reference range. The ΔE value of each anchor point in this frame is measured using a colorimeter. For example, the measured ΔE value for anchor point 1 is 1.2, for anchor point 2 it is 1.6, for anchor point 3 it is 1.3, etc. Values ​​with ΔE values ​​between 1.0 and 1.5 are selected. If there are 18 anchor points in the first frame that meet the requirements of this interval, such as 1, 3, 4, 5, 8, 10, etc., then the grayscale values ​​of these 18 anchor points are extracted to construct a reference anchor point set. At the same time, the judgment and filtering threshold range is set to ΔE from 1.0 to 1.5. This threshold comes from the international electronic image acceptance standard IEC61747 and is formulated according to the medium deviation level standard required by the standard. Based on the circuit board color card data, the grayscale data of the corresponding anchor points are finally extracted and a reference grayscale value sequence is formed. For example, the grayscale values ​​of anchor points 1, 3, and 4 are taken as 210, 215, and 213 respectively, forming an effective anchor point reference value matrix for subsequent reference operations, thus obtaining the effective anchor point grayscale reference values.

[0026] The grayscale reference generation submodule calculates the grayscale change range in subsequent consecutive image frames based on the grayscale values ​​of each anchor point in the effective anchor point grayscale reference values, and statistically analyzes the grayscale mean, maximum deviation, and trend value of each anchor point in all image frames. It then classifies and groups the overall anchor points to generate grayscale anchor point reference data. Based on the data of 18 reference anchor points from the effective anchor point grayscale reference values, the grayscale values ​​of the corresponding anchor point positions in frames 2 to 200 are retrieved, and their grayscale variation range is detected and recorded. The maximum, minimum, and average grayscale values ​​of each anchor point in the entire frame sequence are calculated. For example, the grayscale values ​​of anchor point 1 in frames 2 to 200 are 209, 208, 210, 211... The maximum value in frame 198 is 213, the minimum value is 207, and the average value is 210.2. The grayscale variation trend is calculated as follows: For anchor point 1, it is The grayscale trend values ​​of all 18 anchor points are calculated in this way. Then, the trend values ​​are grouped and divided into three groups: 0 to 0.02, 0.02 to 0.05, and above 0.05. The anchor points are divided into three categories: stable, slightly fluctuating, and significantly fluctuating. Stable anchor points are numbered 2, 5, and 9, and significantly fluctuating anchor points are numbered 1, 7, and 10, etc. Finally, grayscale standards are constructed for anchor points in different groups. By calculating the mean, standard deviation, and median trend value of the grayscale of anchor points in each group, a grayscale standard template for that group is formed. This template is used to organize the reference benchmark for the overall grayscale anchor point data. A mapping relationship table between all anchor points and their grayscale trends and standards is constructed, and finally, grayscale anchor point reference data is generated.

[0027] Specifically, such as Figure 2 , Figure 4 As shown, the grayscale contrast module includes: The image grayscale extraction submodule obtains the coordinate information of all anchor points in the grayscale anchor point reference data, collects the grayscale values ​​of the corresponding coordinate positions in subsequent image frames, selects continuous frame images with the same frame sequence length as the reference data, locates and extracts the grayscale values ​​of all anchor points in each frame image, and generates anchor point grayscale sequence values. To obtain the coordinate information of all anchor points in the grayscale anchor reference data, it is necessary to first read the anchor point index table stored in the reference data. This table contains the image frame number, anchor point number, and corresponding coordinate position. In specific implementations, for example, the first frame image contains anchor point numbers 001 to 030, corresponding to coordinate positions (150, 200), (250, 300), etc. Subsequently, the grayscale values ​​of the corresponding coordinate positions in subsequent image frames are collected. The image frames are continuously recorded at a rate of 10 frames per second for 20 seconds, for a total of 200 grayscale images. The grayscale range is set to integer values ​​between 0 and 255. The frame sequence length is selected to be consistent with the reference data, i.e., image frames 2 to 200. In each image frame, the grayscale values ​​of the aforementioned 30 anchor point coordinates are extracted. The grayscale extraction process is based on pixel grayscale reading, and each... The pixel values ​​of anchor points in a frame are read out to form a list. For example, in the 10th frame image, the grayscale value of anchor point 001 is 210, anchor point 002 is 215, and so on. A two-dimensional grayscale record table is established using the frame number as the horizontal axis and the anchor point number as the vertical axis. The elements in the table are the grayscale values ​​of the corresponding frames and anchor points. After the table is established, each row represents the complete time series grayscale change value of a certain anchor point, and each column represents the grayscale distribution of all anchor points in the same frame. For example, the grayscale values ​​of anchor point 005 in frames 2 to 200 are 208, 210, 209, 211, etc., forming a set of records representing the grayscale changes of anchor points over time. At the same time, the image frame number and the acquisition timestamp are bound one-to-one to establish a complete grayscale record system for all anchor points in the frame sequence and obtain the grayscale sequence values ​​of the anchor points.

[0028] The anchor point frame difference determination submodule calculates the gray level difference between each anchor point in the gray level sequence value of each anchor point in the adjacent frames. If the absolute value does not exceed the etching process monitoring threshold, the anchor point is marked as a continuous and consistent frame point. The set of anchor points that meet the continuous and consistent condition is extracted to obtain the gray level consistent anchor point sequence. To retrieve the grayscale difference between each anchor point in adjacent frames from the anchor point grayscale sequence, the grayscale sequence of the anchor points needs to be read row by row from the aforementioned record table, and the difference between adjacent frames needs to be calculated. For example, if the grayscale values ​​of anchor point 008 are 212 and 198 between frames 5 and 6, the difference is 14. The difference calculation method is as follows: ,in For the grayscale value of the nth frame, determine whether each difference does not exceed the etching process monitoring threshold of 15. If the result is less than or equal to 15, mark the anchor point and the frame pair as grayscale consistent frame points. Repeat this calculation to judge all anchor points across the entire frame range and establish a grayscale consistency marking matrix. The matrix elements are set to 1 to indicate consistency and 0 to indicate inconsistency. For example, anchor point 003 has 8 pairs of adjacent frames with grayscale differences less than 15 in frames 10 to 20, so its corresponding position is 1, and the rest are 0. Then, count the total number of consistent frame points for each anchor point in all frames. If an anchor point maintains a consistent state in more than 80% of the frame pairs, then the anchor point is included in the consistent anchor point set. For example, anchor point 012 has 182 differences less than or equal to 15 in 198 adjacent frame pairs, corresponding to a proportion of 91.9%, which meets the consistency condition. Finally, the anchor point numbers and coordinate information of all anchor points that meet the standard are extracted to form a dataset and obtain the grayscale consistent anchor point sequence.

[0029] The response anomaly identification submodule performs spatial clustering based on the coordinates of anchor point regions in the grayscale consistent anchor point sequence where the difference between consecutive frames is lower than the etching process monitoring threshold. It identifies the continuous consistent anchor point groups that appear repeatedly in multiple frames, determines whether the proportion of the anchor point group in the total anchor points is higher than the etching uniformity threshold, and marks it as a response sluggish region. It then establishes a sluggish region index map under the corresponding frame sequence to obtain the etching response sluggish region. Based on the anchor point regions in the grayscale consistent anchor point sequence where the consecutive frame difference is less than 15, it is necessary to first extract all coordinate information from the aforementioned consistent anchor point set, and then perform spatial clustering processing on their corresponding positions in the circuit board image. Euclidean distance is used as the spatial partitioning criterion, and the maximum clustering radius between anchor points is set to 100 pixels. Multiple consistent anchor points within this range are grouped into the same region. For example, anchor points 001, 002, 003, and 004 are located between coordinate regions (100, 200) and (190, 210), and the distance between any two of them does not exceed 100 pixels; therefore, they are grouped into the same region. Then, the number of anchor points in each cluster region is further... Statistical analysis was conducted, and an etching uniformity threshold of 95% was set. This threshold was derived from the image response judgment standard in the copper foil etching rate experiment. If the number of anchor points in a certain region divided by the total number of anchor points is greater than 0.95, that is, more than 28.5 anchor points are grayscale consistent points, then the region is judged to be a region with slow response. For example, if a region contains 29 consistent anchor points and the total number of anchor points is 30, then the proportion is 96.7%, which meets the set standard. The index information and frame number of the region are retained, and a slow response region index map is established. This index map uses the image frame number as the main index and the region number as the secondary index, and records the spatial clustering results of the corresponding frame segments and anchor points to obtain the etching response slow regions.

[0030] Specifically, such as Figure 2 , Figure 5 As shown, the candidate region filtering module includes: The image region extraction submodule obtains the region index and coordinate information of all corresponding image frames in the etch response sluggish region. In each frame image, it reads the image boundary range corresponding to the region marked by the index, extends outward according to the coordinates to construct the boundary pixel extraction band, traverses all pixels in the boundary band and records the gray level value, arranges the extracted gray level value in row and column order to form a one-dimensional gray level sequence, establishes the gray level boundary record set of all sluggish regions in the corresponding image frame, and generates the boundary gray level sequence set. To obtain the region index and coordinate information of all corresponding image frames in the etch response hysteresis region, it is necessary to first extract the frame number marked as having a hysteresis response and its corresponding region number and coordinate range information from the previous module. For example, in frame 18, the hysteresis region number is Z01, and its boundary coordinate range is (200, 150) at the top left corner and (300, 250) at the bottom right corner. Then, based on this coordinate range, a boundary pixel extraction band is constructed by extending 50 pixels outward along the four boundaries in each frame image. The extended new boundary range becomes (150, 100) to (350, 300). Subsequently, the grayscale values ​​of all pixels are read row by row within this range, and the grayscale values ​​are stitched together row by row according to the image scanning order to form a one-dimensional grayscale sequence. The length of this sequence is... The length of each grayscale sequence is equal to the region width multiplied by the boundary band width. For example, if the extraction bandwidth is 50 pixels and the region width is 100 pixels, then each sequence has a length of 5000 pixels of grayscale values. The grayscale values ​​are integers between 0 and 255. Each pixel grayscale value in the sequence corresponds to a specific location in the image. A record set consisting of four fields is established: frame number, region number, pixel position, and grayscale value. Multiple regions correspond to multiple sequences. For example, the boundary range of the sluggish region Z03 in frame 20 is (400, 300) to (500, 400), and the grayscale extraction band range is (350, 250) to (550, 450), corresponding to a grayscale sequence length of 10000. Finally, all region grayscale sequences are aggregated according to the frame number to obtain the boundary grayscale sequence set.

[0031] The grayscale fluctuation detection submodule calculates the difference between the maximum and minimum grayscale values ​​of each pixel sequence in the boundary grayscale sequence set, filters out sequence segments with grayscale fluctuation less than or equal to the grayscale fluctuation judgment threshold, records the starting position and length of the pixels, and generates low dynamic grayscale interval segments. For each pixel sequence in the boundary grayscale sequence set, the difference between the maximum and minimum grayscale values ​​in the sequence is calculated. This requires reading each sequence one by one and extracting the maximum value within each sequence. and minimum value The grayscale fluctuation calculation formula is adopted. This difference is an important indicator for judging the range of texture structure fluctuations. A threshold of 20 is set, derived from the international image coding standard ITU-T 81, which defines texture-deficient regions. All differences are then considered. The sequence is identified as a low dynamic range sequence, and the starting pixel position and sequence length of the sequence in the original image are recorded for subsequent matching. For example, if a sequence is 5000 pixels long, with a maximum value of 198 and a minimum value of 182, the difference is 16, which meets the set threshold condition. The starting pixel position is recorded as (350, 250), and the sequence is marked as a low dynamic range sequence. This detection operation is repeated, and the difference is calculated and determined for each sequence. A dataset containing the starting position, length, and grayscale fluctuation value of all sequences that meet the condition is established to obtain the low dynamic range grayscale segments.

[0032] The fuzzy structure recognition submodule is based on the starting coordinates and continuous pixel lengths of all grayscale sequences in the low dynamic grayscale interval. Sequences with continuous pixel lengths greater than the grayscale fluctuation judgment threshold are marked as candidate segments. The spatial position index of each corresponding region in the image frame is calculated and matched with the coordinates of the initial stagnant region to obtain fuzzy structure candidate segments. The process retrieves the starting coordinates and continuous pixel lengths of all grayscale sequences within the low dynamic grayscale range. First, it extracts the starting coordinates (x, y) and sequence length L for each record. Then, it performs a filtering operation on L, setting a threshold of 20 for valid candidate segments. This means that only sequences with a continuous length greater than 20 pixels can be considered candidate segments. The filtering method uses an interval comparison approach, comparing the L value of each record with the constant 20. If the condition is met... If the sequence is found to be true, it enters the candidate set. Then, the spatial position of each candidate segment in the image frame is marked and matched with the coordinates of the stagnant region in the corresponding frame. The overlap area ratio is used for calculation, and the spatial overlap threshold is set to 90%. The calculation formula is as follows: ,in, This represents the area of ​​overlap between the candidate fragment and the sluggish region. This represents the area of ​​a candidate segment. If the R value is greater than or equal to 90%, the segment is retained as a fuzzy structure region. For example, if a candidate segment has an area of ​​1200 pixels in frame 28 and overlaps with a sluggish region by 1120 pixels, then R = 93.3%, which meets the condition and is marked as a fuzzy segment. Finally, segments that meet the spatial overlap condition are extracted from all candidates, and a region number, frame number, and coordinate index table are established to obtain fuzzy structure candidate segments.

[0033] Specifically, such as Figure 2 , Figure 6 As shown, the gradient closure verification module includes: The grayscale difference extraction submodule obtains the grayscale value sequence of all pixels in the image frame in the candidate segment of the fuzzy structure, extracts the grayscale difference between two adjacent pixels according to the pixel arrangement order in each segment, records each difference in each gradient chain and the corresponding pixel position, and generates a continuous grayscale gradient chain. To obtain the grayscale value sequence of all pixels in the candidate fragment of the fuzzy structure in the image frame, it is necessary to read the pixel coordinate index of each fragment in the image one by one, and extract the corresponding grayscale value sequence according to the horizontal or vertical pixel arrangement. Assuming that there is a fragment numbered P07 in frame number 28, whose grayscale pixel distribution coordinates are from (150, 200) to (170, 200), a total of 21 consecutive pixels, whose corresponding grayscale values ​​are 190, 192, 193, 194, 192, 191, 190... etc., extract the difference between the grayscale values ​​of two adjacent pixels according to the original pixel arrangement order, and let the first... The grayscale value of each pixel is Then the gradient difference If the grayscale values ​​of the first and second pixels are 190 and 192 respectively, then And so on, forming a complete gradient chain sequence {\DeltaG_1,\DeltaG_2,\DeltaG_3, ...,\DeltaG_{n-1}}, where The number of pixels in a segment is used to record the sequence number of each gradient difference in the gradient chain and the coordinate positions of the pixels before and after it, forming a one-dimensional grayscale gradient chain data structure. At the same time, the corresponding frame number and segment number are labeled for archiving, and a continuous grayscale gradient chain is obtained.

[0034] The gradient error calculation submodule extracts the gray level difference between the first and last points of each continuous gray level gradient chain, calculates the absolute value and records it as the closure difference, filters all segments with closure differences greater than the gradient closure tolerance, records the segment number, the first and last pixel coordinates and the error value, and obtains the first and last gradient error values. Based on the continuous grayscale gradient chain, the grayscale value difference between the first and last points of each chain is extracted. First, the grayscale value of the first pixel is extracted from each gradient chain. grayscale value of the last pixel Calculate the difference in gray levels between the first and last gray levels. The allowable error threshold is set to 5, which is the maximum grayscale error value specified in the gradient closure tolerance standard. If the actual calculated... If the gradient closure condition is not met, an error filtering operation is performed. For example, if the grayscale value of the first pixel of segment P12 is 188 and the grayscale value of the last pixel is 196, then the closure difference is... If the grayscale value is greater than the set threshold of 5, it is determined to be an abnormal closed chain. If the grayscale values ​​of the first and last segments of another segment P13 are 210 and 206 respectively, the closure difference is 4, which belongs to a normal closed chain. Record the segment number, frame number, first and last coordinate position and error value of all abnormal gradient chains with a closure difference greater than 5, establish a temporary label list of abnormal gradient chains, and obtain the first and last gradient error values.

[0035] The abnormal region determination submodule performs region index extraction operation on all segments whose error values ​​are greater than the gradient closure tolerance based on the error data corresponding to each segment in the first and last gradient error values. It counts the coordinate range, number of pixels and gradient chain length of the segment in the image frame, marks the abnormal region number and archives it according to the corresponding frame number, and obtains the structural chain closure abnormal region. The error data corresponding to each segment in the first and last gradient error values ​​is retrieved. For all segments with error values ​​greater than 5, a region index extraction operation is performed. The abnormal gradient chain segment number and its corresponding image frame number are read, and the start and end pixel coordinates of the segment are located. For example, segment P07 is located in frame 36, with start coordinates (400, 210) and end coordinates (430, 210), and a total of 31 pixels. The grayscale difference chain length of each abnormal segment is then calculated. ,in The number of pixels in the segment corresponds to the number of differences in the gradient chain. If multiple abnormal segments appear in the same region or adjacent frames, they are merged and classified into an abnormal structural region. An abnormal region index table is constructed. The fields in the table include region number, frame number, set of segment numbers, first and last coordinates, and error value range. The region numbers are sorted from largest to smallest according to the number of segments. Finally, all segment regions that meet the gradient error threshold condition are merged and registered to obtain the structural chain closure abnormal region.

[0036] Specifically, such as Figure 2 , Figure 7 As shown, the behavior trajectory evolution module includes: The inter-frame marking and recording submodule obtains the frame number and coordinate index information of each abnormal segment in the structural chain closure abnormal region, traverses the continuous image frame sequence, marks whether there are abnormal regions with the same number and the same position in each frame image, constructs a binary sequence according to the frame order, and generates abnormal region frame sequence record values. To obtain the frame number and coordinate index information of each abnormal segment within a structural chain closure anomaly region, the region number, image frame number, and start and end coordinate information of each segment need to be extracted from the anomaly region label table obtained from the preceding module. An anomaly region tracking matrix is ​​established for each segment. Each frame of image data is traversed in the frame sequence, checking if a region with the same region number or a coordinate range overlap rate greater than 90% exists in each frame. If it exists, it is marked as 1; otherwise, it is marked as 0. The generated results are stored in a binary label vector, arranged in frame number order to form a binary sequence. For example, if anomaly region number Z07 actually appears in frames 31, 32, 33, 35, and 38 in frames 31 to 40, the constructed binary sequence is {1, 1, 1, 0, 0, 1, 0, 0, 1, 0}, with a sequence length equal to the number of image frames (10 frames). A single label indicates whether the region appears within the corresponding frame. The region matching method uses a pixel coordinate interval similarity judgment mechanism, based on the region coordinate set in the frame image. With the original coordinate set of the anomaly region The ratio of the intersection area to the original area is used as the criterion for judgment. A valid match is considered when the frame is checked frame by frame. After each frame is checked, a binary existence sequence with a length equal to the number of image frames is generated for each abnormal segment, and the frame sequence record value of the abnormal region is obtained.

[0037] The continuous response tracking submodule calculates the length of the continuous existence segment of each abnormal region in the frame sequence based on the inter-frame changes of each record in the abnormal region frame sequence record value. If the length of the continuous existence segment is greater than or equal to the benchmark length of the stable segment, it is determined to be a continuous response segment. The module extracts the abnormal region numbers and start and end frame indices that meet the conditions to obtain the continuous response frame segment interval. Based on the inter-frame changes of each record in the abnormal region frame sequence, a sliding window operation needs to be performed on the binary sequence to determine whether there is a segment with consecutive 1 values. Assuming the sliding window size is 3 frames, if a region's sequence contains three or more consecutive subsequences of 1 values, then that segment is defined as an abnormal response continuous frame segment. For example, if a region's sequence is {0, 1, 1, 1, 0, 1, 0, 1, 1, 0}, then frames 2 to 4 constitute a valid continuous frame segment with start and end indices [2, 4]. The sliding window method involves continuously scanning 3 frames of data starting from frame 1. If the condition is met, the frame index and region number are recorded. This operation is repeated until the end of the sequence. If multiple segments meet the condition, all start and end frame indices are recorded. The final result is a list of abnormal regions and their corresponding one or more response segments, with the length of each segment recorded. ,in To establish the start and end frame indices, this tracing process is performed across all regions to obtain the corresponding continuous frame segment intervals.

[0038] The abnormal state determination submodule filters the region numbers that meet the stable abnormal judgment threshold based on the abnormal region numbers and the length of the continuous frame segment in the response, integrates and statistically analyzes the corresponding coordinate range and response frequency in the image frame, and archives them by frame index to obtain the circuit board processing abnormal detection results. The call retrieves the index of the abnormal region and the duration of the continuous frame segment within the response frame interval. First, a stable anomaly detection threshold is set to a duration greater than or equal to 5 frames. That is, if an abnormal region has a response length of [missing information] within a continuous image frame interval... If it is a stable abnormal region, then during the filtering process, each record will be selected from the stable abnormal regions. The value is compared with the constant 5. If the condition is met, the region number is included in the result region. The corresponding frame index range and original coordinate interval are extracted. This type of region is grouped and integrated according to frame number, and the response frequency in the corresponding image is counted. ,in, This indicates the number of frames where the abnormal region occurred. Given the total number of frames, if a region numbered Z12 appears 9 times in 30 frames, the response frequency is 30%, and it is classified as a stable abnormal region. All regions that meet the conditions are uniformly numbered and a data table is established. The fields in the table include region number, start and end frames, duration, response frequency in the frame, and spatial coordinate index, to obtain the circuit board processing abnormality detection results.

[0039] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A circuit board processing anomaly detection system, characterized in that, The system includes: The image anchoring module acquires continuous process image frames during the circuit board etching stage, sets positioning anchor points in each frame, collects the standard grayscale values ​​of each anchor point as the first frame reference data, and statistically generates grayscale anchor point reference data. The grayscale comparison module obtains grayscale data at the same position in subsequent image frames based on the coordinates of each anchor point in the grayscale anchor point reference data, performs inter-frame grayscale value comparison operation, identifies anchor point sequences in consecutive frames with differences lower than the etching process monitoring threshold, and obtains the etching response sluggish region. The candidate region filtering module extracts the pixel grayscale sequence of the corresponding image boundary region based on the position of the etch response sluggish region in the image frame, identifies continuous pixel segments whose grayscale value changes are stable in the low dynamic range range, determines them as structurally blurred fragment regions, and obtains blurred structural candidate fragments. The gradient closure verification module extracts the grayscale difference between adjacent pixels based on the pixel grayscale sequence in the fuzzy structure candidate segment, performs absolute value extraction on the grayscale difference between the beginning and end of the gradient chain and performs error judgment with the gradient closure tolerance, records the segment region whose error exceeds the gradient closure tolerance as an abnormal structure region, and generates an abnormal region of structure chain closure.

2. The circuit board processing anomaly detection system according to claim 1, characterized in that: The grayscale anchor reference data includes anchor position index, standard grayscale level, and pattern boundary mapping relationship. The etch response lag region includes inter-frame grayscale change sequence, low-variation anchor point set, and threshold trigger marker. The blurred structure candidate fragment includes low dynamic range grayscale sequence, continuous pixel segment index, and image boundary region identifier. The structural chain closure anomalous region includes gradient chain start and end grayscale difference anomalous segments, grayscale closure error record, and anomalous structural fragment label.

3. The circuit board processing anomaly detection system according to claim 1, characterized in that: The standard grayscale value is specifically the grayscale level defined by the international standard for electronic circuit board acceptance; the etching process monitoring threshold is specifically the allowable range of grayscale changes defined by the rigid printed circuit board qualification and performance specifications; and the low dynamic range is specifically the grayscale fluctuation range defined by the international standard for image coding.

4. The circuit board processing anomaly detection system according to claim 1, characterized in that: The image anchoring module includes: The grayscale acquisition submodule acquires image frames of continuous processes during the circuit board etching stage. In each image frame, multiple pixels within the pattern boundary are set as positioning anchor points. The standard grayscale value of each anchor point is acquired, and interference pixels at the image edge are removed to generate a set of image anchor point grayscale values. The reference grayscale setting submodule selects the first frame data as the reference image frame according to the time series based on the grayscale values ​​of each frame anchor point in the image anchor grayscale value set, extracts the standard grayscale values ​​of all anchor points in the first frame, calls the standard color card grayscale level reference value for comparison and calibration, filters the effective reference anchor points, and obtains the effective anchor point grayscale reference value. The grayscale reference generation submodule calculates the grayscale change range in subsequent consecutive image frames based on the grayscale values ​​of each anchor point in the effective anchor point grayscale reference values, and statistically analyzes the grayscale mean, maximum deviation, and trend value of each anchor point in all image frames. It then classifies and groups the overall anchor points to generate grayscale anchor point reference data.

5. The circuit board processing anomaly detection system according to claim 1, characterized in that: The grayscale contrast module includes: The image grayscale extraction submodule obtains the coordinate information of all anchor points in the grayscale anchor point reference data, collects the grayscale values ​​of the corresponding coordinate positions in subsequent image frames, selects continuous frame images with the same frame sequence length as the reference data, locates and extracts the grayscale values ​​of all anchor point coordinates in each frame image, and generates anchor point grayscale sequence values. The anchor point frame difference determination submodule calculates the gray level difference between each anchor point in the gray level sequence value of the anchor point in the adjacent frames. If the absolute value does not exceed the etching process monitoring threshold, the anchor point is marked as a continuous and consistent frame point. The set of anchor points that meet the continuous and consistent condition is extracted to obtain the gray level consistent anchor point sequence. The response anomaly identification submodule performs spatial clustering based on the coordinates of anchor point regions in the grayscale consistent anchor point sequence where the difference between consecutive frames is lower than the etching process monitoring threshold. It identifies continuous consistent anchor point groups that appear repeatedly in multiple frames, determines whether the proportion of the anchor point group in the total anchor points is higher than the etching uniformity threshold, and marks it as a response sluggish region if it is higher. It then establishes a sluggish region index map under the corresponding frame sequence to obtain the etching response sluggish region.

6. The circuit board processing anomaly detection system according to claim 1, characterized in that: The candidate region filtering module includes: The image region extraction submodule obtains the region index and coordinate information of all corresponding image frames in the etch response sluggish region. In each frame image, it reads the image boundary range corresponding to the region marked by the index, extends outward according to the coordinates to construct the boundary pixel extraction band, traverses all pixels in the boundary band and records the gray level value, arranges the extracted gray level value in row and column order to form a one-dimensional gray level sequence, establishes the gray level boundary record set of all sluggish regions in the corresponding image frame, and generates the boundary gray level sequence set. The grayscale fluctuation detection submodule calculates the difference between the maximum and minimum grayscale values ​​of each pixel sequence in the boundary grayscale sequence set, filters out sequence segments with grayscale fluctuation less than or equal to the grayscale fluctuation judgment threshold, and records the pixel start position and length to generate low dynamic grayscale interval segments. The fuzzy structure recognition submodule, based on the starting coordinates and continuous pixel lengths of all grayscale sequences in the low dynamic grayscale interval, marks sequences with continuous pixel lengths greater than the grayscale fluctuation judgment threshold as candidate segments, calculates the spatial position index of each corresponding region in the image frame, and matches it with the coordinates of the initial stagnant region to obtain fuzzy structure candidate segments.

7. The circuit board processing anomaly detection system according to claim 1, characterized in that: The gradient closure verification module includes: The grayscale difference extraction submodule obtains the grayscale value sequence of all pixels in the image frame of the candidate segment of the fuzzy structure, extracts the grayscale difference between two adjacent pixels according to the pixel arrangement order in each segment, records each difference in each gradient chain and the corresponding pixel position, and generates a continuous grayscale gradient chain. The gradient error calculation submodule extracts the gray level difference between the first and last points of each chain based on the continuous gray level gradient chain, calculates the absolute value and records it as the closure difference, filters all segments with closure differences greater than the gradient closure tolerance, records the segment number, the first and last pixel coordinates and the error value, and obtains the first and last gradient error values. The abnormal region determination submodule performs a region index extraction operation on all segments with error values ​​greater than the gradient closure tolerance based on the error data corresponding to each segment in the first and last gradient error values. It counts the coordinate range, number of pixels and gradient chain length of the segment in the image frame, marks the abnormal region number and archives it according to the corresponding frame number, and obtains the structural chain closure abnormal region.

8. The circuit board processing anomaly detection system according to claim 1, characterized in that, The system also includes: The behavior trajectory evolution module marks whether the structural chain closure abnormal region appears in different frames according to the corresponding position of the abnormal region in consecutive image frames, tracks the response changes in the image frame index order, determines whether the corresponding region continues to show stable abnormal signs, and generates circuit board processing abnormality detection results.

9. The circuit board processing anomaly detection system according to claim 8, characterized in that: The circuit board processing anomaly detection results include anomaly area trajectory index results, anomaly response continuous marking results, and stable anomaly judgment identification records.

10. The circuit board processing anomaly detection system according to claim 8, characterized in that: The behavior trajectory evolution module includes: The inter-frame marking and recording submodule obtains the frame number and coordinate index information of each abnormal segment in the structural chain closure abnormal region, traverses the continuous image frame sequence, marks whether there are abnormal regions with the same number and the same position in each frame image, constructs a binary sequence according to the frame order, and generates abnormal region frame sequence recording values. The continuous response tracking submodule calculates the length of the continuous existence segment of each abnormal region in the frame sequence based on the inter-frame changes of each record in the abnormal region frame sequence record value. If the length of the continuous existence segment is greater than or equal to the stable segment judgment benchmark length, it is determined to be a response continuous segment. All abnormal region numbers and start and end frame indices that meet the conditions are extracted to obtain the response continuous frame segment interval. The abnormal state determination submodule filters out the region numbers that meet the stable abnormal judgment threshold based on the abnormal region numbers and the length of the continuous frame segment in the response, integrates and statistically analyzes the corresponding coordinate range and response frequency in the image frame, and archives them by frame index to obtain the circuit board processing abnormality detection results.

Images