A method for PCB two-dimensional code reading and PCB hole inspection data binding

Abstract

The application discloses a method for PCB two-dimensional code reading and PCB hole inspection data binding, which comprises the following steps: obtaining composite image information corresponding to a PCB board through a same set of integrated optical image acquisition units and performing light efficiency separation processing to obtain reflected light image data and penetrating light image data; determining position reference correlation data corresponding to the reflected light image and the penetrating light image based on the reflected light image data and the penetrating light image data, performing blob extraction and analysis on the penetrating light image data, and generating a standard file in combination with a UI interactive interface; completing PCB board positioning, blob quality determination and two-dimensional code identification reading based on the standard file and the position reference correlation data, and generating PCB hole inspection data binding data in association, and simultaneously performing PCB data binding verification to complete binding and archiving of the PCB hole inspection data and the two-dimensional code. The application can significantly improve the judgment accuracy of one-to-one detection of PCB hole inspection data binding.

Description

Technical Field

[0001] This invention relates to the field of data binding technology, specifically, to a method for reading PCB QR codes and binding PCB inspection data. Background Technology

[0002] As the core carrier of PCB board identification, the QR code, when linked to the inspection data, is a crucial link in achieving full lifecycle traceability and quality control of PCBs. Traditional inspection methods require the deployment of two independent optical systems for QR code reading and inspection, using software time difference matching to achieve data binding. This not only results in low equipment integration and large space occupation, but also makes it prone to data correlation due to matching errors, which can no longer meet the high-efficiency requirements of large-scale PCB production.

[0003] Currently, several methods have been proposed for binding PCB QR code reading with via inspection data. These methods typically employ separate QR code readers (optical imaging system A) and via inspection optical systems (optical imaging system B) operating independently. Optical imaging system A is only responsible for decoding and outputting the PCB QR codes, while optical imaging system B is responsible for collecting via information and performing quality assessments. After both systems have completed their respective tasks, a time difference parameter is set via software to determine whether the two sets of data belong to the same PCB board. However, existing methods are affected by the different imaging positions and inconsistent operation times of the two systems, making it difficult to set the time difference parameter, which can easily lead to misjudgments or failures in data binding. On the other hand, when there are multiple QR codes on the PCB board, the traditional single-reader decoding mode cannot achieve a one-to-one match between the QR code and the via inspection data, further increasing the risk of binding failure and thus reducing the reliability of data traceability in PCB production inspection. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for reading PCB QR codes and binding PCB inspection data, comprising the following steps:

[0005] Step S1: Acquire composite image information corresponding to the PCB board through the same set of integrated optical imaging units, and perform optical effect separation processing on the composite image information to obtain reflected light image data and transmitted light image data; Step S2: Determine the location reference association data corresponding to the reflected light image and the transmitted light image based on the reflected light image data and the transmitted light image data; extract and parse the transmitted light image data from blobs, and generate a standard file by selecting the positioning blobs and reading the location using the QR code in the UI interactive interface. Step S3: Based on the standard file and position reference data, complete the PCB board positioning, blob quality judgment and QR code recognition and reading, and generate QR code information and hole inspection results; Step S4: Associate the QR code information with the hole inspection results to generate PCB hole inspection data binding data, and perform PCB data binding verification based on the PCB hole inspection data binding data to complete the binding and archiving of PCB hole inspection data and QR codes.

[0006] The beneficial effects of this application are as follows: By using the same integrated optical imaging unit to acquire composite images of the PCB board, the traditional mode of two independent optical systems operating separately is completely eliminated. This ensures, from a hardware perspective, that the imaging positions of the reflected light image and the transmitted light image are completely consistent, avoiding positional deviations caused by installation errors of different systems. The composite image undergoes optical effect separation processing, simultaneously obtaining reflected light image data (for QR code reading) and transmitted light image data (for via detection), achieving synchronous acquisition of QR code information and via data. This eliminates the impact of the time difference between the two systems, simplifies the hardware architecture, reduces equipment debugging difficulty, and avoids the risk of data binding misjudgment or failure due to asynchrony between the two systems, thus improving the consistency and reliability of PCB inspection data acquisition. Secondly, by determining the positional reference correlation data based on the reflected light and transmitted light images, a correspondence between the two sets of images is established, providing a unified spatial reference for the subsequent one-to-one correlation of QR code and via data, avoiding data misalignment caused by missing positional references. The system extracts and analyzes blobs from transmitted light images. Combined with a user-friendly interface, it selects blobs for positioning and QR code reading locations to generate a standard profile. This allows for flexible specification of the reading location and associated blob area for each QR code on a PCB board with multiple QR codes. This overcomes the limitation of traditional single-head decoding, which cannot distinguish between multiple QR codes. The standard profile clearly defines the binding rules between QR codes and corresponding verification areas, avoiding the problem of confusion between multiple QR codes and verification data in traditional methods. Then, relying on the positioning blobs and location reference data in the standard profile, PCB board positioning can be quickly completed, ensuring consistent positioning for each inspection. Simultaneously, based on the QR code reading locations set in the standard profile, the system selectively reads corresponding QR code information, achieving recognition and correspondence in multi-QR code scenarios. While completing positioning, blob quality assessment and QR code recognition are performed simultaneously, generating corresponding QR code information and verification results. This ensures that the two sets of data are generated synchronously for the same PCB board and the same area, completely eliminating data attribution errors caused by the time difference in traditional dual-system operations. Finally, the QR code information and the hole inspection results are linked to generate PCB hole inspection data binding data, achieving a tight association between the QR code and the corresponding hole inspection data, providing a data foundation for PCB lifecycle traceability. Based on the bound data, data binding verification can check the rationality and completeness of the association, promptly identify and correct potential binding errors, and prevent erroneous data from flowing into subsequent traceability stages. Compared to traditional models lacking verification mechanisms and where binding errors are difficult to detect, this significantly improves the accuracy of data binding. After verification, the PCB hole inspection data and QR codes are bound and archived, forming a standardized traceability data archive. This ensures that each PCB's QR code corresponds to its hole inspection data, achieving one-to-one traceability even in multi-QR code scenarios, thus significantly improving the reliability of PCB production inspection data traceability. Attached Figure Description

[0007] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the method for reading PCB QR codes and binding PCB hole inspection data in this embodiment; Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S1. Detailed Implementation

[0008] The following drawings will disclose several embodiments of the present invention. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential. Furthermore, for the sake of simplicity, some conventional structures and components will be shown in the drawings in a simple schematic manner.

[0009] To further understand the invention's content, features, and effects, the following embodiments are provided, and detailed descriptions are given below in conjunction with the accompanying drawings: Reference Figure 1 , Figure 1 This is a flowchart illustrating the method for reading PCB QR codes and binding PCB hole verification data in this embodiment. The method for reading PCB QR codes and binding PCB hole verification data in this embodiment includes the following steps: Step S1: Acquire composite image information corresponding to the PCB board through the same set of integrated optical imaging units, and perform optical effect separation processing on the composite image information to obtain reflected light image data and transmitted light image data; In this embodiment of the invention, by activating the same integrated optical imaging unit, which includes a line scanning module and a flash control module, the line scanning module, equipped with a line camera, moves at a uniform speed of 30 mm per second along the length of the PCB board, scanning and acquiring data line by line across the entire PCB board. The flash control module alternately switches between reflected and transmitted light emission states at a fixed frequency of 50 Hz, with a switching interval of 10 milliseconds. When reflected light is on, transmitted light is off, and vice versa, eliminating light superposition interference. Simultaneously, reflected light images of the QR code area and transmitted light images of the via area on the PCB board surface are captured, forming composite image information. The composite image is transmitted to the memory unit for real-time storage at a 2MB per frame specification, generating a time-series image dataset. By splitting the dataset according to the parity of the time sequence number, odd-numbered frames are the reflected light image frame sequence, and even-numbered frames are the transmitted light image frame sequence. After frame fusion noise reduction, contour enhancement, and light effect frame filtering and purification processing, clear-textured reflected light image data and complete-contour transmitted light image data are finally obtained. The grayscale range of the reflected light image is 180-255, and the grayscale range of the transmitted light image is 80-220.

[0010] Step S2: Determine the location reference association data corresponding to the reflected light image and the transmitted light image based on the reflected light image data and the transmitted light image data; extract and parse the transmitted light image data from blobs, and generate a standard file by selecting the positioning blobs and reading the location using the QR code in the UI interactive interface. In this embodiment of the invention, based on reflected and transmitted light image data, a PCB board image coordinate system is established with the upper left corner of the board as the origin, the X-axis pointing horizontally to the right, and the Y-axis pointing vertically downwards, with 1 pixel corresponding to an actual size of 0.01 mm. The coordinates of the four vertices of the QR code and the center of the corner vias are extracted, and the positional offset is calculated and corrected (controlled within ±2 pixels) to determine the positional reference data. Grayscale thresholding (threshold 80) and morphological operations on 3×3 rectangular structural elements are performed on the transmitted light image data. After connected component analysis and morphological filtering (bounding rectangle aspect ratio 1.0-1.2, circularity 0.85-1.0), blob extraction and invalid stray blob removal are completed. The valid blob set is transmitted to the UI interactive interface, and positioning blobs are selected according to the four quadrants, with an adjacent spacing of not less than 50% of the board's side length. The reference positioning parameters are recorded. Simultaneously, a 30×30 pixel rectangle is designated as the QR code reading position, and the vertex coordinates are recorded. By integrating parameters and transmitted light image features, a standardized comparison template is built, a standard file is generated and stored in memory, and bound to a time-series image dataset.

[0011] Step S3: Based on the standard file and position reference data, complete the PCB board positioning, blob quality judgment and QR code recognition and reading, and generate QR code information and hole inspection results; In this embodiment of the invention, fixed positional relationship data of source images are extracted from the location reference association data, and a positioning comparison framework is built by integrating standard file parameters, limiting the allowable range of positional deviation to ±2 pixels. The effective blobs of the PCB board transmitted light image are compared with the features of the standard file positioning blobs, and core matching blobs with a fit ≥0.95 are selected to generate preliminary positioning data. After deviation correction and dual-image position collaborative verification (deviation ≤2 pixels), positioning calibration data is obtained. Based on the calibration data, blob parameters are verified one by one, and four results are judged: qualified, abnormal shape, abnormal light transmission, and positional offset, generating blob quality judgment results. The coordinates of the QR code reading are mapped to the reflected light image through a coordinate mapping reference factor. After texture verification and deviation correction (deviation ±1 pixel after calibration), the QR code recognition area is locked. Gray-scale stretching enhancement and distortion correction are performed on the area, and texture features are extracted line by line to complete decoding. The QR code information and blob quality judgment results are associated to generate corresponding data.

[0012] Step S4: Associate the QR code information with the hole inspection results to generate PCB hole inspection data binding data, and perform PCB data binding verification based on the PCB hole inspection data binding data to complete the binding and archiving of PCB hole inspection data and QR codes.

[0013] In this embodiment of the invention, by associating QR code information with via inspection results, and sorting according to the PCB board image coordinate system, the quality status and coordinate information of each via are bound to the QR code information, generating PCB via inspection data binding data. The binding data includes the QR code encoding, full blob quality judgment details, positioning calibration parameters, and image feature summary, stored in a fixed byte stream format. The binding data undergoes dual verification: the first verification checks the positional correlation between the QR code information and the corresponding via area to ensure no binding misalignment; the second verification checks the consistency between the blob quality judgment result and the original transmitted light image features, eliminating data with judgment deviations. Qualified binding data is synchronized to the memory unit for archiving, establishing a unique correspondence between the QR code encoding and the via inspection data. Archived data is stored according to PCB board batches, retaining a complete binding link record, thus realizing the binding and archiving of PCB via inspection data and QR codes.

[0014] Furthermore, the integrated optical imaging unit includes a line scanning module and a flash control module, wherein, referring to Figure 2 , Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S1 in this embodiment. Step S1 includes the following steps: Step S11: Start the line scanning module and flash control module corresponding to the integrated optical imaging unit. The line scanning module performs full-area scanning acquisition on the PCB board. The flash control module alternately switches the light emission state corresponding to the reflected light and the transmitted light according to the preset frequency, and simultaneously acquires the composite image information of the PCB board containing the reflected light and the transmitted light. The acquired composite image information is then transmitted to the memory for real-time storage to generate a time-series image dataset. In this embodiment of the invention, by activating the line scanning module and flash control module in the integrated optical imaging unit, the line scanning module, equipped with a line camera, moves linearly at a uniform speed of 30 mm per second along the length of the PCB board. During the movement, the line camera scans the QR code area and the through-hole area on the PCB board surface line by line, achieving full-area acquisition without blind spots. The flash control module is directly connected to the reflected light and transmitted light emitting components via circuitry, and is synchronized with the scanning rhythm of the line camera. It alternately switches the emission states of the two types of light at a fixed frequency of 50 Hz. When the power supply circuit of the reflected light is connected, the power supply circuit of the transmitted light is disconnected, and vice versa. The interval between each switch is strictly controlled to 10 milliseconds, completely avoiding overlapping interference and light effect crosstalk during the acquisition of the two types of light. During the scanning process, reflected light images of the QR code area on the PCB board surface and transmitted light images of the through-hole area inside the board are captured simultaneously. The two images are superimposed according to the scanning sequence to form composite image information. The composite image is transmitted sequentially to the memory unit along the data transmission line at a specification of 2MB per frame. It is written to the storage area in real time in the form of a continuous byte stream. Each frame of the image is marked with a unique scanning sequence number and the corresponding PCB board scanning position coordinates. Finally, a time-series image dataset is generated, which is arranged in chronological order and includes position information. The dataset completely covers all image information of the QR code area texture and the through-hole area morphology.

[0015] Step S12: Perform frame separation processing on the time-series image dataset to obtain the image frame sequence of reflected light and transmitted light corresponding to the PCB board; In this embodiment of the invention, the light type corresponding to each frame is deduced by relying on the time sequence number of each frame in the time-series image dataset and a preset 50Hz light switching frequency, and frame separation processing is performed on the dataset. Image frames are divided by the parity of the time sequence number. Image frames with odd time sequence numbers are extracted to form a reflected light image frame sequence, and image frames with even time sequence numbers are extracted to form a transmitted light image frame sequence. Each image frame maintains a resolution of 1024×768 pixels and a color depth of 24 bits to ensure that image details are not lost. During the frame separation process, based on the PCB board scanning position coordinates attached to each image frame, a one-to-one correspondence is established between the reflected light image frame and the transmitted light image frame corresponding to the same scanning position. This strictly locks the positional association of the two image frame sequences for the same PCB board, preventing positional offset phenomena and providing a basic guarantee for the positional binding of subsequent QR code reading and via detection.

[0016] Step S13: Perform light effect frame filtering and purification on the image frame sequences of reflected light and transmitted light corresponding to the PCB board to obtain the set of reflected light image frames and the set of transmitted light image frames. In this embodiment of the invention, the reflected light image frame sequence is purified using a grayscale threshold filtering method. A fixed grayscale threshold range of 180-255 is set via hardware circuitry. The pixel grayscale distribution of each reflected light image frame is detected frame by frame. Image frames with an overall grayscale value within this range and a grayscale compliance rate exceeding 95% in the QR code area are retained. Invalid frames with grayscale values ​​below 180, shadow interference, or blurred QR code textures are removed, ultimately forming a set of reflected light image frames. The transmitted light image frame sequence is purified using a contrast filtering method. The difference between the maximum and minimum grayscale values ​​of each frame is calculated, i.e., the contrast ratio. Image frames with a contrast ratio higher than 80 and clear distinction between the via area and the board background are retained. Invalid frames with insufficient contrast or blurred via outlines are removed, resulting in a set of transmitted light image frames. Both sets are sorted according to the PCB board scanning position coordinates, retaining image frames at corresponding positions on the same PCB board to ensure complete preservation and consistent positional correspondence between the QR code area and the via area images.

[0017] Step S14: Perform frame fusion and noise reduction processing on the set of reflected light image frames to obtain reflected light image data; In this embodiment of the invention, frame fusion and noise reduction processing is performed on the set of reflected light image frames. The image frames within the set are grouped according to the PCB board scanning position coordinates, with each group corresponding to multiple reflected light image frames of the same QR code area. Pixel-by-pixel grayscale values ​​are superimposed on the multiple image frames within each group. The sum of the grayscale values ​​of corresponding pixels within each group is calculated and then divided by the number of image frames in that group to obtain the average grayscale value. This average grayscale value is used as the grayscale value of the corresponding pixel in the final reflected light image. All pixels are processed one by one to complete frame fusion. This process effectively cancels random noise interference in single-frame images and weakens image deviations caused by slight light fluctuations during scanning. The resulting reflected light image data has clear texture and uniform grayscale, presenting detailed features such as the edges and textures of the QR code on the PCB board surface, providing a clear and stable image foundation for feature recognition during subsequent QR code interpretation.

[0018] Step S15: Perform contour enhancement processing based on the set of transmitted light image frames to obtain transmitted light image data.

[0019] In this embodiment of the invention, contour enhancement processing is performed based on a set of transmitted light image frames. First, edge detection is performed frame by frame on all transmitted light image frames in the set. The algorithm identifies the grayscale threshold between the pixels at the via edge and the background pixels, thus locking the via contour range. Then, by adjusting and enhancing the grayscale difference between the via edge and the background area, the grayscale value of the pixels at the via edge is uniformly increased to 220, while the grayscale value of the pixels in the background area is reduced to 50. This widens the grayscale difference between the two to highlight the via contour shape and avoid confusion between the via edge and the background. During the processing, the continuity of the via contour is simultaneously calibrated, minor contour breaks are filled, and contour offsets caused by scanning are corrected. The final transmitted light image data can clearly present the complete contour, aperture size, and wall condition of the PCB via, providing identifiable image support for subsequent via quality inspection and blob analysis.

[0020] Furthermore, step S13 includes the following steps: Optical effect frame spectrum analysis is performed based on the image frame sequence of reflected and transmitted light corresponding to the PCB board to generate optical effect feature maps of each frame in the image frame sequence under different optical effects. In this embodiment of the invention, relying on the previously separated reflected and transmitted light image frame sequences, a light effect frame spectrum analysis is performed through a calculation and comparison unit, with the analysis process proceeding synchronously with the image frame time sequence. For the reflected light image frame sequence, the QR code area and the surrounding 5 mm area are locked frame by frame, and two core features are extracted: the distribution data of pixel grayscale values ​​within this area and the boundary coordinates of the light effect coverage range. For the transmitted light image frame sequence, the through-hole area and the plate background transition area are located frame by frame, and two core features are extracted: the average pixel grayscale value of the through-hole area and the grayscale attenuation amplitude corresponding to the light effect penetration depth. Each feature data is recorded one by one according to the time sequence number of each image frame, generating independent light effect feature maps for reflected and transmitted light. Each map is labeled with the time sequence number of the corresponding image frame, the light effect type, and specific feature parameters, realizing the visualization and data retention of the light effect features of each image frame under different light effects, providing complete data support for subsequent gradient evaluation.

[0021] Furthermore, pixel grayscale distribution gradients are evaluated based on the light effect feature maps corresponding to each frame in the image frame sequence under different light effects to obtain the light effect feature values ​​corresponding to the image frame sequence under different light effects; the light effect purity factor corresponding to each frame under different light effects is calculated based on the light effect feature values. In this embodiment of the invention, pixel grayscale distribution gradient evaluation is performed based on core data from two types of optical effect feature maps. The evaluation process strictly corresponds to the region division standard of each frame image. The rate of change of grayscale value from the core region to the edge region of each frame image is calculated. The reflected light image frame is divided into three gradient regions with the center of the QR code as the origin and a radius of 3 mm, and the width of each region is 1 mm. The difference in grayscale value and the rate of change between adjacent two layers are calculated in turn. The transmitted light image frame is divided into two gradient regions with the center of the aperture as the origin and a radius of 2 mm, and the width of each region is 1 mm. The rate of change of grayscale in the gradient direction is calculated, and the optical effect feature value is determined according to the stability of the rate and the value range. The optical effect purity factor is calculated according to the ratio of the optical effect feature value to the preset standard optical effect feature benchmark value. The value range of the reflected light purity factor is strictly controlled within 0.8-1.0, and the value range of the transmitted light purity factor is controlled within 0.75-1.0. The factor value is positively correlated with the optical effect purity. The closer the value is to the upper limit, the more stable the optical effect and the less interference.

[0022] Furthermore, the flash switching timing data corresponding to the integrated optical imaging unit is obtained, and the flash switching timing data is used to perform flash switching timing calibration on the light effect purity factor corresponding to each frame under different light effects, so as to eliminate the light effect interference factor corresponding to the moment of flash switching and generate the light effect purity parameter corresponding to each frame under different light effects after timing calibration. In this embodiment of the invention, preset flash switching timing data is retrieved from the flash control module. This data completely records the start time, end time, and stable emission duration of each light switch, matching the previously set 50Hz switching frequency and 10ms switching cycle. The acquisition time of each frame of image is compared with the timing data one by one to determine whether each frame of image is at the moment of light switching. The moment of switching is defined as the period from 1ms before the start of switching to 1ms after the end of switching. For image frames within this period, the average value of the light effect purity factor of 3 adjacent stable time periods under the same light effect type is retrieved. The purity factor of the current frame is corrected according to the average value to completely eliminate the light effect interference factor caused by the superposition of the two light rays at the moment of switching. The corrected light effect purity parameter is retained to three decimal places and bound to the timing number, light effect type, and position coordinates of the corresponding image frame to ensure that the parameter corresponds to the image without deviation.

[0023] Furthermore, based on the light effect purity parameters corresponding to each frame under different light effects after time-series calibration, the statistical distribution of light effect purity is obtained, and the screening benchmark value of light effect frames under different light effects is constructed based on the statistical distribution of light effect purity. Based on the screening benchmark value of light effect frames under different light effects, the image frame sequences of reflected light and transmitted light corresponding to the PCB board are subjected to light effect frame pre-screening to obtain the initial screening image frame set corresponding to reflected light and transmitted light. In this embodiment of the invention, the distribution range of all luminous efficacy purity parameters after statistical time-series calibration is calculated and compared by the comparison unit. The extreme parameters at both ends (7.5%) are removed, and the screening benchmark values ​​are set according to the quantiles of the remaining 85% of parameters. The screening benchmark value for reflected light parameters is set to 0.88, and the screening benchmark value for transmitted light parameters is set to 0.83. The purity parameters of each image frame are compared with the corresponding luminous efficacy screening benchmark values ​​frame by frame. Image frames with parameters higher than the benchmark values ​​are retained, while image frames with parameters lower than the benchmark values, insufficient luminous efficacy, or significant interference are resolutely removed, forming initial screening image frame sets for reflected light and transmitted light respectively. The image frames within the sets are strictly sorted according to their time sequence numbers, fully preserving the positional coordinate relationship between each image frame and the PCB board, ensuring a one-to-one correspondence between the QR code area and the via area image, laying a solid foundation for subsequent secondary purification and processing.

[0024] Furthermore, the light effect attenuation data corresponding to each frame in the initial screening image frame set is obtained, and the inter-frame light effect coordination factor corresponding to different light effects is calculated based on the light effect attenuation data corresponding to each frame in the initial screening image frame set. Based on the inter-frame light effect coordination factor corresponding to different light effects, the initial screening image frame set corresponding to reflected light and transmitted light is subjected to secondary purification processing to remove abnormal frames corresponding to the inter-frame light effect coordination factor deviating from the normal range, so as to obtain the reflected light image frame set and the transmitted light image frame set.

[0025] In this embodiment of the invention, the luminous efficacy attenuation data of each frame in the initial screening set is acquired through an optical imaging unit, and the luminous efficacy intensity attenuation amount and attenuation trend at the time of acquisition of each frame are recorded. Taking three adjacent frames as a group, the fit and variation law of the attenuation amounts of the three frames are calculated to obtain the luminous efficacy coordination factor corresponding to each frame. The normal range of the reflected light coordination factor is limited to 0.9-1.1, and the normal range of the transmitted light coordination factor is limited to 0.85-1.15. The coordination factor is determined frame by frame to determine whether it is within the normal range, and abnormal frames with factors deviating from the range are removed. For regions with three or more consecutive abnormal frames, the intermediate frame is retained and fine-tuned by combining the attenuation data of the preceding and following normal frames. Finally, a set of reflected and transmitted light image frames with stable luminous efficacy, no abnormal interference, and accurate positional correspondence is obtained, effectively ensuring the processing accuracy of subsequent frame fusion, contour enhancement, QR code reading, and hole verification.

[0026] Furthermore, step S2 includes the following steps: Step S21: Construct a PCB board image coordinate system based on reflected light image data and transmitted light image data, and determine the position reference association data corresponding to the reflected light image and transmitted light image based on the PCB board image coordinate system; In this embodiment of the invention, based on the reflected light image data and transmitted light image data obtained above, a PCB board image coordinate system is constructed through a calculation and comparison unit. The origin is the top left corner of the PCB board, with the positive X-axis pointing horizontally to the right and the positive Y-axis pointing vertically downwards. The coordinate system unit is set to pixels, with each pixel corresponding to an actual PCB board size of 0.01 mm. The coordinates of the four vertices of the QR code in the reflected light image and the center coordinates of the four corner vias in the transmitted light image are extracted. These two types of coordinates are then incorporated into the same coordinate system for calibration. By calculating the positional offsets of the two types of images, the X-axis and Y-axis offsets are controlled within ±2 pixels. Based on the offset-corrected coordinate data, positional reference association data between the reflected light image and the transmitted light image is determined. This data includes the coordinate mapping relationship between the two types of images, the deviation correction value, and the correspondence of reference points, ensuring that the QR code area and the via area are aligned in the coordinate system. This provides coordinate support for subsequent blob analysis and QR code reading position binding.

[0027] Step S22: Perform connected component analysis on the transmitted light image data to complete blob extraction and obtain the full blob dataset; In this embodiment of the invention, connected component analysis is performed on the transmitted light image data after contour enhancement processing. The analysis is executed by a calculation and comparison unit, setting a pixel grayscale value difference ≤10 as the connectivity criterion. The transmitted light image data is traversed pixel by pixel. Scanning begins from the upper left corner of the image, marking the first pixel whose grayscale value matches the via feature (50-220 range) as a seed point. This seed point is then expanded to its eight neighboring regions, grouping all pixels that meet the connectivity condition into the same connected component. This process is repeated across the entire image to complete the labeling of all connected components. Each connected component corresponds to a blob. The core parameters of each blob are extracted, including center coordinates, boundary coordinates, total number of pixels, and average grayscale value. These parameters are recorded sequentially according to the connected component labeling order, forming a full blob dataset. The dataset is sorted according to the PCB board image coordinate system, and each blob is bound to the coordinate information of the corresponding via region, ensuring a one-to-one correspondence between the dataset and the actual via locations on the PCB board, with no omissions or duplicate labels.

[0028] Step S23: Perform morphological representation and classification filtering on the full blob dataset to remove invalid and stray blobs and obtain a valid blob set; In this embodiment of the invention, morphological characterization and classification filtering are performed on the full blob dataset. Morphological parameters of each blob are extracted, including the aspect ratio of the circumscribed rectangle, roundness, and area. The aspect ratio of the circumscribed rectangle is calculated as the ratio of the longer side to the shorter side. Roundness is calculated as the ratio of the blob area to the area of ​​the circumscribed circle. The area is converted to the actual area based on the total number of bound pixels (each pixel corresponds to 0.0001 square millimeters). Filtering thresholds are set: circumscribed rectangle aspect ratio 1.0-1.2, roundness 0.85-1.0, and actual area corresponding to ±5% of the standard via area of ​​the PCB board. Blobs with a grayscale mean deviating from the standard grayscale range of vias (80-200) are removed. The morphological parameters of each blob are compared with the thresholds one by one. Blobs whose parameters all meet the threshold requirements are retained, while invalid stray blobs with parameters exceeding the standard, belonging to board impurities, or background interference are removed. Finally, a valid blob set is obtained. The blobs in the set are arranged in the order of the PCB board image coordinate system, completely preserving the via shape and position features.

[0029] Step S24: By combining the UI interactive interface to display the effective blob set, and selecting the positioning blob and QR code reading location, the baseline positioning parameters and QR code reading coordinates are obtained; In this embodiment of the invention, the effective blob set is transmitted to the UI interactive interface. The interface restores and displays the full-frame transmitted light image and effective blobs according to the PCB board image coordinate system ratio. Each blob is marked with a red outline, and the corresponding center coordinates and shape parameters are displayed simultaneously. The interface has a built-in fixed selection mechanism, limiting the selection of four corner effective blobs as positioning blobs. These four blobs must be located in the four quadrants of the PCB board image coordinate system, and the distance between adjacent positioning blobs must not be less than 50% of the side length of the PCB board. The positioning blobs are locked through the interface touch operation, and the system automatically records their center coordinates, roundness, and area as reference positioning parameters. At the same time, the QR code area is locked in the reflected light image area of ​​the interface. With the center of the QR code as the reference, a 30×30 pixel rectangular area is defined as the QR code reading position. The coordinates of the four vertices of the rectangular area are recorded as the QR code reading coordinates. The coordinate data is retained to integer pixels to ensure the accuracy of positioning and the accuracy of the QR code reading range. All parameters are synchronized to the calculation and comparison unit in real time for storage.

[0030] Step S25: Obtain the features of the transmitted light image based on the transmitted light image data, and construct a standardized comparison template based on the benchmark positioning parameters, QR code reading coordinates and transmitted light image features to generate a standard file.

[0031] In this embodiment of the invention, transmitted light image features are obtained based on transmitted light image data, including three core features: the grayscale distribution pattern of effective blobs, the grayscale gradient value of via edges, and the grayscale reference value of the board background. The grayscale distribution pattern is represented by the pixel grayscale statistical curve of each effective blob, the edge grayscale gradient value is set to 20-30, and the background grayscale reference value is fixed at 50. The reference positioning parameters, QR code reading coordinates, and transmitted light image features are integrated to build a standardized comparison template. The template is based on the PCB board image coordinate system and includes built-in parameter thresholds for positioning blobs, the QR code reading area range, and standard values ​​for transmitted light image features. During template generation, the correlation of various parameters is simultaneously calibrated to ensure that the positioning parameters correspond to the QR code reading coordinates and that the image features match the shape of the effective blobs. Finally, a standard file is generated and stored in memory unit in byte stream form, bound to the corresponding PCB board's time-series image dataset, providing a unified standard basis for subsequent board positioning, via verification, and QR code reading.

[0032] Furthermore, step S22 includes the following steps: Gray-scale thresholding is performed on the transmitted light image data to distinguish between the foreground blob region and the background region, resulting in a binarized image. In this embodiment of the invention, grayscale threshold segmentation is performed on the transmitted light image data after contour enhancement processing. The operation is completed by a calculation and comparison unit. Based on the grayscale range of the transmitted light image set above, a fixed segmentation threshold of 80 is set. Pixels below this threshold are determined to be background areas and uniformly assigned a grayscale value of 0; pixels above or equal to this threshold are determined to be foreground blob areas and uniformly assigned a grayscale value of 255. During the segmentation process, the entire image is traversed pixel by pixel, and the regions are strictly divided according to the threshold standard. The foreground areas corresponding to the vias are locked in the key, ensuring that the via blobs required for QR code binding are distinguished from the board background, avoiding the omission of foreground areas or the misjudgment of background areas due to threshold ambiguity. After segmentation, a binary image is obtained, which contains only black and white grayscale values. The foreground blob areas are presented in white, and the background areas are presented in black, completely preserving the contour shape and position information of the via areas, laying the foundation for subsequent morphological operations.

[0033] Furthermore, morphological operations are performed on the binarized image to eliminate tiny noise points and fill in the holes inside the blob, resulting in an optimized binarized image. In this embodiment of the invention, morphological operations are performed on the binarized image. Dilation and erosion operations are executed sequentially to eliminate minute noise points and fill holes within the blob. The dilation operation uses a 3×3 pixel rectangular structuring element. It iterates through each white pixel in the binarized image, converting any black pixels within its eight-neighborhood to white, repeating this operation once to thoroughly remove minute black noise points smaller than 3×3 pixels. The erosion operation uses the same rectangular structuring element, iterating through the dilated image and converting any target pixels containing black pixels within their eight-neighborhood to black, repeating this operation once to restore the original contour of the foreground blob region and fill tiny holes within the blob with a diameter less than 2 pixels. The resulting optimized binarized image has a continuous and complete foreground blob region contour, free from noise interference and internal defects, ensuring the accuracy of subsequent connected component analysis.

[0034] Furthermore, connected component labeling analysis is performed on the corresponding foreground blob regions in the optimized binarized image to extract all independent blobs; In this embodiment of the invention, connected component labeling analysis is performed on the foreground blob region in the optimized binarized image. By performing a scanning labeling operation, an eight-neighbor connectivity determination rule is set, where adjacent white pixels (grayscale value 255) are grouped into the same connected component. Starting from the top-left pixel of the image, the image is scanned line by line. When the first unlabeled white pixel is encountered, a unique label number is assigned, and the process expands to its eight neighbors, assigning the same label number to all connected white pixels. This process is repeated across the entire image to complete the labeling of all independent blobs. The labeling process strictly follows the PCB board image coordinate system established earlier, with each connected component label number bound to its corresponding position coordinates. This ensures that independent blobs corresponding to different vias are not labeled in confusion. Finally, all independent blobs are extracted, each distinguished by a unique label number, with complete outlines and clear boundaries.

[0035] Furthermore, the morphological parameters, coordinate information, and grayscale features corresponding to each independent blob are recorded to generate a blob feature dataset; In this embodiment of the invention, a blob feature dataset is generated by recording the morphological parameters, coordinate information, and grayscale features corresponding to each independent blob. Morphological parameters include the aspect ratio, circularity, and area of ​​the circumscribed rectangle. The aspect ratio is calculated as the ratio of the longer side to the shorter side. Circularity is calculated as the ratio of the blob area to the area of ​​the circumscribed circle. The area is converted to the actual area based on the total number of pixels (0.0001 square millimeters per pixel). The coordinate information is extracted from the blob center coordinates, calculated based on the PCB board image coordinate system, and retained to the nearest integer pixel. The grayscale features record the average grayscale value of the blob region; since all values ​​are 255 after binarization, they are uniformly recorded as a fixed value. Each independent blob corresponds to a complete set of feature data, recorded sequentially according to the label number to ensure no parameters are omitted and no data deviations, providing a detailed basis for subsequent classification and filtering.

[0036] Furthermore, the feature datasets corresponding to all independent blobs are integrated to form a full blob dataset.

[0037] In this embodiment of the invention, a full blob dataset is formed by integrating the blob feature datasets corresponding to all independent blobs. During the integration process, the data are sorted according to the positive X-axis direction of the PCB board image coordinate system, and the feature data of each blob are concatenated sequentially. The label number, morphological parameters, coordinate information, and grayscale features of each blob correspond one-to-one, constructing a structured data set. At the same time, the data correlation is calibrated to ensure that the feature data of each blob completely matches the actual location and label number in the image, with no data misalignment or omission. The integrated full blob dataset contains blob information corresponding to all vias on the PCB board, stored in a unified format in the memory unit, and linked with the time-series image dataset and location reference correlation data mentioned above. This provides complete data support for subsequent invalid stray blob removal and valid blob screening, ensuring the continuity of PCB QR code reading and via data binding.

[0038] Furthermore, the connected component labeling analysis of the corresponding foreground blob region in the optimized binarized image includes the following steps: Extract the pixel distribution data of the foreground region corresponding to the foreground blob region in the optimized binarized image to obtain the pixel coordinate dataset; In this embodiment of the invention, foreground pixel distribution data of the foreground blob region in the optimized binarized image is extracted and linked with the PCB board image coordinate system established earlier. Only white foreground pixels with a grayscale value of 255 are collected. The entire optimized binarized image is traversed row by row and column by column, recording the X-axis and Y-axis coordinates corresponding to each white pixel. The coordinate values ​​are calculated based on the origin of the coordinate system and the actual size corresponding to the pixel (1 pixel corresponds to 0.01 mm), and are retained to the nearest integer. During the acquisition process, background pixels with a grayscale value of 0 are simultaneously removed to ensure that only the foreground pixel data corresponding to the vias are retained. Finally, a pixel coordinate dataset is formed. The dataset is arranged in ascending order of X-axis coordinates, and each coordinate point is bound to unique pixel position information, providing data support for subsequent connectivity analysis.

[0039] Furthermore, based on the pixel coordinate dataset, the connectivity features corresponding to adjacent pixels are calculated to generate a pixel connectivity matrix; In this embodiment of the invention, a pixel connectivity matrix is ​​generated by calculating the connectivity features of adjacent pixels based on a pixel coordinate dataset. An eight-neighbor connectivity criterion is established: eight pixels above, below, to the left, right, and diagonally adjacent to a given pixel are considered adjacent pixels. If all adjacent pixels are foreground pixels (their coordinates exist within the pixel coordinate dataset), they are considered connected and marked as 1; if adjacent pixels are background pixels (their coordinates are not within the dataset), they are considered unconnected and marked as 0. Using the coordinate order of the pixel coordinate dataset as the matrix row and column indices, the connectivity between each pixel and its eight neighboring pixels is calculated one by one, constructing a two-dimensional pixel connectivity matrix. The matrix dimension is consistent with the optimized binarized image pixel resolution (1024×768), clearly presenting the connectivity status of all foreground pixels.

[0040] Furthermore, the discrete foreground blob regions are clustered and labeled based on the pixel connectivity matrix to obtain a preliminary set of connected components; domain differentiation identifier data is generated based on this preliminary set of connected components. In this embodiment of the invention, discrete foreground blob regions are clustered and labeled based on a pixel connectivity matrix. Starting with the first connected pixel labeled 1 in the top left corner of the matrix, it is grouped into the first cluster group. Simultaneously, the clustering is expanded along the eight-neighbor direction, incorporating all pixels labeled 1 into the same cluster group and assigning them unique domain identifiers. After labeling the first cluster group, the matrix is ​​traversed to find unlabeled connected pixels, and the clustering operation is repeated until all foreground pixels are grouped into their corresponding cluster groups, resulting in a preliminary set of connected domains. Domain differentiation identifiers are generated based on this set. Each identifier is bound to the pixel coordinate range and total number of pixels in the corresponding cluster group, ensuring that different connected domains can be distinguished by their identifiers, preventing confusion and overlap.

[0041] Furthermore, by combining the grayscale features of the transmitted light image, each connected component is verified, and broken pseudo-connected components are removed to obtain effective connected component data. In this embodiment of the invention, each connected component is verified by combining the grayscale features of the transmitted light image. The grayscale standard of the transmitted light image after contour enhancement (grayscale range of 80-220 for the via region, and 50 for the background) is used. The original transmitted light image region corresponding to each cluster in the preliminary connected component set is extracted one by one, and the average grayscale value of the region is calculated. If the average grayscale value is in the range of 80-220 and the difference between the grayscale value and the background region is ≥30, it is determined to be a valid connected component. If the average grayscale value deviates from this range, or if there is a grayscale discontinuity causing false connectivity, it is determined to be a broken pseudo-connected component and directly removed. During the verification process, the average grayscale value of the valid connected components is recorded simultaneously to ensure that it conforms to the via image features, ultimately obtaining the valid connected component data.

[0042] Furthermore, contour boundaries are extracted from the effective connected component data to generate all independent blobs.

[0043] In this embodiment of the invention, contour boundaries are extracted from the effective connected component data. Based on the pixel connectivity matrix, the edge pixels of each effective connected component are identified, i.e., foreground pixels with background pixels (connectivity marked as 0) within their eight neighboring regions. The edge pixels are traversed clockwise, their coordinates are recorded, and they are concatenated to form a closed contour. During the extraction process, the PCB board image coordinate system is strictly followed, the smoothness of the contour boundaries is calibrated, and individual pixels protruding on the contour are removed to ensure the continuity and integrity of the contour. Each effective connected component corresponds to an independent blob, arranged in order of domain distinction identifier number. The contour boundary coordinates of each independent blob are recorded synchronously, and finally all independent blobs are generated, laying the foundation for subsequent feature extraction and classification filtering.

[0044] Furthermore, step S3 includes the following steps: Step S31: Obtain fixed positional relationship data of source images based on location benchmark association data, and construct a positioning comparison framework based on the standard file and the fixed positional relationship data of source images; In this embodiment of the invention, based on the location reference association data obtained above, fixed positional relationship data of the same source image is extracted. This data includes the coordinate mapping coefficients of reflected and transmitted light images, the reference point deviation correction value, and the positional correspondence ratio between the QR code area and the via area. The PCB board image coordinate system parameters (1 pixel corresponds to 0.01 mm) are strictly followed to ensure that the positional relationship data is consistent with the calibration standard mentioned above. The standard file stored in the memory unit is called, and the positioning blob parameters, QR code reading coordinates, and fixed positional relationship data of the same source image built into the standard file are integrated to build a positioning comparison framework. The framework includes built-in coordinate mapping rules, blob matching thresholds, and QR code area positioning standards, clearly defining the correspondence between the PCB board image and the standard file, limiting the allowable positional deviation range to ±2 pixels, providing a unified comparison benchmark for subsequent positioning matching, blob verification, and QR code reading, and ensuring the consistency of operations at each stage.

[0045] Step S32: Based on the positioning comparison framework, compare the transmitted light image corresponding to the PCB board with the corresponding positioning in the standard file using blobs to perform positioning matching and generate positioning calibration data; In this embodiment of the invention, based on a positioning comparison framework, positioning matching is performed between the transmitted light image corresponding to the PCB board and the positioning blobs in the standard file. After the transmitted light image of the PCB board is processed using the same procedure described above, four blobs at its corners are extracted as candidate positioning points. The center coordinate deviation, roundness deviation, and area deviation of each candidate positioning point and the positioning blobs in the standard file are calculated one by one. The coordinate deviation is calculated separately for the X-axis and Y-axis, with the roundness deviation controlled within ±0.03 and the area deviation controlled within ±3%. Candidate positioning points with acceptable deviations are matched to generate positioning calibration data. This data includes the deviation correction value, coordinate calibration coefficient, and overall positioning accuracy parameters for each matched point. If the deviation of a candidate positioning point exceeds the standard, it is corrected based on the parameters of adjacent matched points to ensure the accuracy of the positioning calibration data and provide positional reference support for subsequent blob verification.

[0046] Step S33: Based on the comparison of positioning calibration data, verify each blob in the corresponding transmitted light image of the PCB board to obtain the blob quality judgment result; In this embodiment of the invention, blobs in the transmitted light image of the PCB board are verified one by one based on positioning calibration data. The verification process strictly follows the standards set by the positioning comparison framework. The morphological parameters (boundary rectangle aspect ratio 1.0-1.2, circularity 0.85-1.0, area corresponding to ±5% of the standard via area) and grayscale features (average grayscale value 80-220) of each blob are extracted one by one. Combined with the coordinate position corrected by the positioning calibration data, the parameters of the blob at the corresponding position in the standard file are compared. If all parameters meet the standards, it is judged as qualified; if the morphological parameters exceed the standard, it is judged as morphological abnormality; if the grayscale value deviates from the range, it is judged as light transmittance abnormality; if the position deviation exceeds ±2 pixels, it is judged as positional offset. The verification results of each blob are recorded one by one, generating a blob quality judgment result. The result is bound to the coordinate information and identification number of the corresponding blob, ensuring that the quality status of each via is traceable.

[0047] Step S34: Based on the fixed positional relationship data of the same source image, map the corresponding QR code reading coordinates in the standard file to the reflected light image to lock the QR code recognition area; In this embodiment of the invention, coordinate mapping rules based on the fixed positional relationship data of the same source image are used to map the QR code reading coordinates (coordinates of the four vertices of a 30×30 pixel rectangular area) within the standard file to the reflected light image of the comparison PCB board. During the mapping process, positioning calibration data is used to correct coordinate deviations, ensuring that the mapped QR code reading coordinates are aligned with the QR code area in the reflected light image, with the coordinate deviation controlled within ±1 pixel. After mapping, the QR code recognition area in the reflected light image is locked. This area is bounded by the mapped rectangular coordinates, preserving the complete texture range of the QR code while eliminating interference from surrounding irrelevant areas. This defines the scope for subsequent QR code texture enhancement, distortion correction, and reading operations, ensuring the targeted and accurate reading of the QR code.

[0048] Step S35: Perform texture enhancement and distortion correction on the QR code recognition area, and generate QR code information and verification results by combining the QR code recognition area reading with the blob quality judgment results.

[0049] In this embodiment of the invention, enhancement processing is performed on the QR code recognition area. A grayscale stretching algorithm is used to expand the grayscale difference between the QR code texture and the background, adjusting the grayscale value of the texture area to 180-255 and the grayscale value of the background area to 0-50, thereby enhancing the recognition of texture details. By correcting image distortion, the geometry of the QR code area is calibrated based on the PCB board image coordinate system to eliminate slight stretching distortion generated during scanning and ensure the regularity of the QR code texture. After calibration, the recognition area is scanned line by line to extract the QR code texture features and complete the decoding to obtain the QR code information. The QR code information is integrated with the previously generated blob quality judgment result, and the coordinate information and quality status of the corresponding vias are associated according to a fixed format to generate comprehensive data containing QR code information and via verification results, which is stored in the memory unit to realize the binding of PCB QR code reading and via verification data.

[0050] Furthermore, step S32 includes the following steps: Extract the set of valid blobs from the transmitted light image corresponding to the PCB board to obtain the comparison blob dataset; In this embodiment of the invention, a comparison blob dataset is obtained by extracting and comparing the effective blob set from the transmitted light image corresponding to the PCB board. After the compared PCB board transmitted light image is processed using the same process as described above (grayscale threshold segmentation, morphological operations, connected component analysis, and morphological filtering), invalid stray blobs are removed by strictly adhering to the filtering thresholds (bounding rectangle aspect ratio 1.0-1.2, circularity 0.85-1.0, area corresponding to standard via area ±5%). The extracted effective blobs are sorted in the positive direction of the X-axis of the PCB board image coordinate system. Each blob records complete feature data, including morphological parameters, center coordinates (calculated based on 1 pixel corresponding to 0.01 mm, retaining integer pixels), and grayscale mean (range 80-220). Finally, a comparison blob dataset is formed. The dataset is consistent with the standard file positioning blob data format stored in the memory unit, providing a unified data foundation for subsequent feature comparison.

[0051] Furthermore, based on the localization comparison framework, feature comparison is performed between the blob dataset and the corresponding localization blobs in the standard file to extract matching feature points and calculate feature fit. In this embodiment of the invention, feature comparison is performed between the blob dataset and the location blobs in the standard file based on a location comparison framework. The location blobs are the four corner blobs preset in the standard file, and they are compared one by one in quadrant order. The core feature parameters of both blobs are extracted, including center coordinates, roundness, area, and edge gray-level gradient value. The edge gray-level gradient value is limited to 20-30. The fit of each feature parameter is calculated. The coordinate fit is calculated inversely proportional to the absolute value of the deviation. The allowable deviation for roundness fit is ±0.03, the allowable deviation for area fit is ±3%, and the allowable deviation for edge gray-level gradient value fit is ±2. Matching feature points that meet the fit requirements for all feature parameters are extracted. The overall feature fit is calculated by combining all parameters. The fit value ranges from 0 to 1. Values ​​higher than 0.95 are judged as high-fit matching points, providing a basis for subsequent core matching blob selection.

[0052] Furthermore, based on feature matching, core matching blobs are selected, and the positional correspondence between the PCB board and the standard document is established to generate preliminary positioning data. In this embodiment of the invention, core matching blobs are selected based on feature fit. A fit threshold of 0.95 is set, and blobs corresponding to matching feature points with a feature fit higher than this threshold are retained as core matching blobs, ensuring that at least one core matching blob is retained in each quadrant. If no blob meets the threshold in a certain quadrant, the blob with the highest fit in that quadrant and not lower than 0.90 is selected, and its feature value is corrected by combining the parameters of core matching blobs in adjacent quadrants, and then included in the core matching blob set. Based on the coordinate correspondence of the core matching blobs, a positional correspondence between the PCB board and the standard document is established, the mapping ratio of the coordinate systems of both sides and the deviation of the reference point are calculated, and preliminary positioning data is generated. This data includes the correspondence of the core matching blobs, coordinate mapping coefficients, and overall fit statistics, providing a basic positioning basis for subsequent deviation correction.

[0053] Furthermore, the initial positioning data is corrected for deviation based on the fixed positional relationship data of the same source images to eliminate the influence of small offsets during the image acquisition process. The reflected light and transmitted light image coordinate mapping parameters in the fixed positional relationship data of the same source images are retrieved to construct the deviation correction benchmark. In this embodiment of the invention, deviation correction is performed on the preliminary positioning data based on the fixed positional relationship data of the same source image. The fixed positional relationship data of the same source image includes the coordinate mapping coefficients of the reflected light and transmitted light images and the reference point deviation correction value, strictly adhering to the allowable positional deviation range of ±2 pixels. The correction process focuses on eliminating the influence of minor offsets during image acquisition. According to the coordinate mapping rules in the fixed positional relationship data, the blob coordinates in the preliminary positioning data are adjusted to ensure that the positional reference of the PCB board is consistent with that of the standard file. Simultaneously, the coordinate mapping parameters of the reflected light and transmitted light images in the fixed positional relationship data of the same source image are retrieved. These parameters are fixed values ​​after calibration mentioned above. Based on this, a deviation correction reference is constructed. The reference incorporates a coordinate correction formula, an upper limit for offset, and parameter calibration standards, providing a unified correction basis for subsequent adaptive calibration.

[0054] Furthermore, texture feature points in the reflected light image corresponding to the PCB board are extracted and compared with the feature points corresponding to the reflected light image in the standard file to calculate the horizontal and vertical offsets; a deviation correction model is constructed based on the offsets, and the blob coordinates in the preliminary positioning data are adaptively calibrated based on the deviation correction model to generate calibrated coordinates; at the same time, the positional coordination of the two images of the PCB board is verified to generate positional coordination verification data. In this embodiment of the invention, texture feature points of the QR code area in the PCB board reflected light image are extracted and compared. Twenty feature points at the edges and corners of the QR code are selected and arranged in coordinate order. Each QR code texture feature point is compared with the one corresponding to the reflected light image in the standard file, and the horizontal (X-axis) and vertical (Y-axis) offsets are calculated. The offsets are calculated in pixels, rounded to the nearest integer. A deviation correction model is constructed based on the offsets. The model is set according to linear correction rules, and the horizontal and vertical offsets are substituted into the correction formulas to adaptively calibrate the blob coordinates in the initial positioning data, generating calibrated coordinates. The deviation of the calibrated coordinates is controlled within ±1 pixel. Simultaneously, the positional coordination of the PCB board reflected light and transmitted light dual images is verified. The coordinate deviation between the QR code area and the corresponding via area is calculated. A deviation ≤2 pixels is considered acceptable for coordination, generating positional coordination verification data, which includes the offset, calibration coefficient, and coordination judgment result.

[0055] Furthermore, positioning calibration data is generated based on location collaborative verification data and calibrated coordinates.

[0056] In this embodiment of the invention, positioning calibration data is generated based on location coordination verification data and calibrated coordinates. First, the coordination judgment results in the location coordination verification data are verified, retaining only the calibrated coordinates that pass coordination. If coordination failures exist, recalibration is performed based on the deviation correction benchmark until the coordination meets the requirements. The calibrated coordinates, core matching blob correspondence, deviation correction values, coordinate mapping coefficients, and location coordination verification results are integrated to construct structured positioning calibration data. The data is sorted according to the PCB board image coordinate system, and each core matching blob is bound to corresponding calibration parameters and coordination verification information. The finally generated positioning calibration data is stored in a memory unit and linked with the comparison blob dataset and standard file data to provide positioning support for subsequent blob verification and QR code reading coordinate mapping.

[0057] Furthermore, step S34 includes the following steps: Step S341: Based on the fixed positional relationship data of the same source images, analyze the spatial correspondence pattern of the two images to obtain the coordinate mapping reference factor; In this embodiment of the invention, based on the fixed positional relationship data of the same source images obtained above, the spatial correspondence between the reflected light and transmitted light dual images is analyzed. This data includes the coordinate mapping coefficients of the dual images, the reference point deviation correction value, and the positional correspondence ratio, strictly following the PCB board image coordinate system parameters (1 pixel corresponds to 0.01 mm). During the analysis process, the focus is on extracting the coordinate correspondence of the reference points of the dual images (the vertex of the QR code and the center of the corner through-hole), calculating the mapping ratio coefficients in the X and Y axes, locking the linear law of coordinate mapping, and eliminating nonlinear offset interference. Based on the analysis results, a coordinate mapping reference factor is generated. This factor includes the X-axis mapping coefficient, the Y-axis mapping coefficient, and the reference point deviation compensation value. The mapping coefficient is fixed as a constant value after calibration above, and the deviation compensation value is controlled within ±1 pixel, providing a reference basis for subsequent coordinate mapping calculations.

[0058] Step S342: Retrieve the coordinates of the QR code corresponding to the standard file, and perform coordinate mapping calculation on the reflected light image in combination with the coordinate mapping reference factor to generate the initial mapped coordinates of the reflected light image; In this embodiment of the invention, coordinates are read by retrieving a QR code stored in the standard file. These coordinates are the coordinates of the four vertices of a 30×30 pixel rectangular area, recorded based on the PCB board image coordinate system and retained to the nearest integer pixel. The coordinate mapping reference factor is then substituted into a preset mapping formula, specifically... ; ,in , The coordinates for reading the QR code are stored in the standard file. X-axis mapping coefficient, Y-axis mapping coefficient, This is the X-axis reference point deviation compensation value. To compensate for the Y-axis reference point deviation, the initial mapped coordinates of the reflected light image are calculated one by one by performing mapping coordinate calculations on the reflected light image and substituting them into a preset mapping formula. During the calculation, X-axis and Y-axis mapping coefficients and deviation compensation values ​​are simultaneously substituted to correct the spatial positional deviation between the two images, ensuring that the initial mapped coordinates are approximately aligned with the QR code area in the reflected light image, with the initial mapped coordinate deviation controlled within ±3 pixels. The generated initial mapped coordinates of the reflected light image are arranged in the order of the rectangle vertices and bound to the corresponding coordinate calculation parameters, laying the foundation for subsequent texture verification and deviation correction.

[0059] Step S343: Extract texture features of the corresponding region of the reflected light image based on the initial mapping coordinates of the reflected light image, and construct a texture verification dataset; In this embodiment of the invention, texture features of the corresponding region of the reflected light image are extracted based on the initial mapping coordinates of the reflected light image. The texture extraction region is defined by expanding the initial mapping coordinates by 5 pixels to ensure complete coverage of the potential QR code texture. Core texture features within this region are extracted, including the arrangement pattern of the QR code's black and white modules, edge texture grayscale gradient values ​​(range 180-255), corner feature point coordinates, and module spacing. Module spacing is calculated in pixels, down to 1 pixel. The extracted texture features are organized by category, and the specific parameters of each feature are recorded sequentially to construct a texture verification dataset. The dataset maintains the same format as the standard QR code texture feature data to ensure consistency in subsequent comparison operations.

[0060] Step S344: Based on the texture verification dataset, compare the QR code texture features corresponding to the standard file and calculate the coordinate deviation correction amount; In this embodiment of the invention, the texture verification dataset is compared one by one with the QR code texture features corresponding to the standard file, with a focus on comparing the arrangement pattern of black and white modules, the position of corner feature points, and the gray-scale gradient values ​​at the edges. The coordinate deviations of the corner feature points in the X and Y axes are calculated, and the average deviation of all feature points is statistically analyzed to generate a coordinate deviation correction amount. If there is a misalignment in the arrangement of black and white modules, the deviation correction amount is adjusted according to the number of misaligned pixels; if the edge gray-scale gradient value deviation exceeds ±5, the correction parameters are finely adjusted based on the gradient distribution pattern. The final determined coordinate deviation correction amount is calculated separately for the X and Y axes, down to 0.1 pixels, ensuring the accuracy of subsequent coordinate calibration and completely eliminating the positional deviation caused by the initial mapping.

[0061] Step S345: Based on the coordinate deviation correction, calibrate the initial mapping coordinates of the reflected light image to generate an identification mapping coordinate set, and delineate and lock the QR code identification area based on the identification mapping coordinate set.

[0062] In this embodiment of the invention, by substituting the coordinate deviation correction amount into the initial mapping coordinates of the reflected light image, the coordinates of the four vertices are calibrated one by one to generate an identification mapping coordinate set. After calibration, the coordinate deviation is strictly controlled within ±1 pixel to ensure alignment with the QR code area in the reflected light image. Based on the identification mapping coordinate set, a closed rectangle is formed by concatenating the vertices in sequence to delineate the QR code identification area. The area boundary strictly fits the edge of the QR code, eliminating irrelevant background areas and noise interference. After delineation, the identification mapping coordinate set and area range parameters are recorded simultaneously and stored together with texture verification data and coordinate deviation correction amount. This provides area positioning support for subsequent QR code texture enhancement, distortion correction, and reading operations, ensuring the accuracy of QR code reading.

[0063] Furthermore, step S341 includes the following steps: Extract the fixed positional relationship data of the same source image to separate the feature anchor point information corresponding to the reflected light and transmitted light images, and obtain the dual light effect anchor point dataset; In this embodiment of the invention, a dual-light-effect anchor point dataset is obtained by separating the feature anchor point information corresponding to reflected light and transmitted light images from the fixed positional relationship data of the same source images. The fixed positional relationship data of the same source images uses the calibration parameters mentioned above and strictly follows the PCB board image coordinate system specifications (1 pixel corresponds to 0.01 mm). The feature anchor points are selected from the regions with fixed positions and stable features in the two types of images. The anchor points of the reflected light image are selected from the four vertices of the QR code and the midpoints of the four sides, for a total of 8 anchor points; the anchor points of the transmitted light image are selected from the centers of the four corner positioning blobs and the centers of the two standard through holes in the middle area of ​​the board, for a total of 6 anchor points. The X-axis and Y-axis coordinates, gray-level feature values ​​and edge gradient parameters of each anchor point are recorded one by one. The gray-level values ​​of the reflected light anchor points are limited to 180-255, the gray-level values ​​of the transmitted light anchor points are limited to 80-220, and the edge gradient parameters are controlled within 20-30. The dataset is classified and organized according to the light effect type to form a dual-light-effect anchor point dataset, which provides feature basis for subsequent anchor point matching.

[0064] Furthermore, a matching analysis was performed on the dual-light-effect anchor point dataset to establish a one-to-one correspondence between anchor points and generate an anchor point association map. In this embodiment of the invention, a one-to-one correspondence between anchor points is established by performing matching analysis on the dual-light-effect anchor point dataset, generating an anchor point association map. The matching process uses the vertex anchor points of the reflected light image QR code as a benchmark, and combines the coordinate mapping ratio in the fixed positional relationship data of the same source image to compare the coordinate positions, grayscale features, and edge gradient parameters of the transmitted light image anchor points one by one. Matching thresholds are set: coordinate deviation ≤ 2 pixels, grayscale feature deviation ≤ 5, and edge gradient parameter deviation ≤ 2. Anchor points that meet all threshold conditions are established with corresponding relationships. Specifically, the four vertex anchor points of the reflected light QR code correspond to the center anchor points of the four corner positioning blobs of the transmitted light, and the midpoint anchor point of the reflected light QR code corresponds to the center anchor point of the through-hole in the middle area of ​​the transmitted light, ensuring the anchor point correspondence and covering the key areas of the entire board. An anchor point association map is generated based on the correspondence, and the map labels the coordinates, feature parameters, and correspondence of each anchor point, clearly presenting the matching logic of the dual-light-effect anchor points.

[0065] Furthermore, based on the anchor point association map, the spatial position deviation corresponding to the anchor points of the two images is calculated to obtain the anchor point deviation dataset; In this embodiment of the invention, an anchor point deviation dataset is obtained by calculating the spatial position deviation between anchor points of two images based on the anchor point association map. The X-axis and Y-axis coordinate differences between the reflected light anchor point and the corresponding transmitted light anchor point are calculated one by one according to the anchor point correspondence, with the coordinate difference calculated in pixels, down to 0.1 pixels. The spatial position deviation of each anchor point is calculated multiple times and the average is taken to eliminate single calculation errors. Anchor points with deviations exceeding ±1 pixel are marked, and the reasonableness of the deviation is verified by combining the parameters of adjacent anchor points. If the deviation is abnormal, the anchor point correspondence is recalibrated and the calculation is repeated. The spatial position deviations, average deviations, deviation ranges, and calibration records of all anchor points are compiled and summarized to form an anchor point deviation dataset. The dataset is sorted according to the anchor point correspondence, providing deviation data support for subsequent spatial correspondence pattern analysis and coordinate mapping model construction.

[0066] Furthermore, by combining the anchor point deviation dataset to analyze the spatial correspondence pattern, a coordinate mapping association model is constructed, and a coordinate mapping reference factor is generated through the calculation of the coordinate mapping association model.

[0067] In this embodiment of the invention, the spatial correspondence between reflected and transmitted light images is analyzed using an anchor point deviation dataset. This establishes a linear relationship in the coordinate mapping of the two images, eliminates nonlinear offsets and random interference, and clarifies the mapping ratios and overall deviation trends in the X and Y axes. Based on the analyzed spatial correspondence, a coordinate mapping association model is constructed. This model incorporates an anchor point deviation correction formula, a coordinate mapping algorithm, and parameter calibration standards. The anchor point deviation dataset is substituted into the model for calculation and fitting, optimizing the mapping coefficients and deviation compensation values. A coordinate mapping benchmark factor is generated through model calculation. This factor includes X-axis mapping coefficients, Y-axis mapping coefficients, and benchmark deviation compensation values. The mapping coefficients are fixed as constant values ​​after fitting, and the deviation compensation values ​​are controlled within ±1 pixel. This ensures that the benchmark factor adapts to the spatial correspondence of the two images, providing a core benchmark for subsequent QR code reading coordinate mapping calculations.

[0068] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for reading PCB QR codes and binding PCB inspection data, characterized in that, Includes the following steps: Step S1: Acquire composite image information corresponding to the PCB board through the same set of integrated optical imaging units, and perform optical effect separation processing on the composite image information to obtain reflected light image data and transmitted light image data; Step S2: Determine the location reference association data corresponding to the reflected light image and the transmitted light image based on the reflected light image data and the transmitted light image data; extract and parse the transmitted light image data from blobs, and generate a standard file by selecting the positioning blobs and reading the location using the QR code in the UI interactive interface. Step S3: Based on the standard file and position reference data, complete the PCB board positioning, blob quality judgment and QR code recognition and reading, and generate QR code information and hole inspection results; Step S4: Associate the QR code information with the hole inspection results to generate PCB hole inspection data binding data, and perform PCB data binding verification based on the PCB hole inspection data binding data to complete the binding and archiving of PCB hole inspection data and QR codes.

2. The method for reading PCB QR codes and binding PCB inspection data according to claim 1, characterized in that, The integrated optical imaging unit includes a line scanning module and a flash control module, wherein step S1 includes the following steps: Step S11: Start the line scanning module and flash control module corresponding to the integrated optical imaging unit. The line scanning module performs full-area scanning acquisition on the PCB board. The flash control module alternately switches the light emission state corresponding to the reflected light and the transmitted light according to the preset frequency, and simultaneously acquires the composite image information of the PCB board containing the reflected light and the transmitted light. The acquired composite image information is then transmitted to the memory for real-time storage to generate a time-series image dataset. Step S12: Perform frame separation processing on the time-series image dataset to obtain the image frame sequence of reflected light and transmitted light corresponding to the PCB board; Step S13: Perform light effect frame filtering and purification on the image frame sequences of reflected light and transmitted light corresponding to the PCB board to obtain the set of reflected light image frames and the set of transmitted light image frames. Step S14: Perform frame fusion and noise reduction processing on the set of reflected light image frames to obtain reflected light image data; Step S15: Perform contour enhancement processing based on the set of transmitted light image frames to obtain transmitted light image data.

3. The method for reading PCB QR codes and binding PCB inspection data according to claim 2, characterized in that, Step S13 includes the following steps: Optical effect frame spectrum analysis is performed based on the image frame sequence of reflected and transmitted light corresponding to the PCB board to generate optical effect feature maps of each frame in the image frame sequence under different optical effects. The pixel grayscale distribution gradient is evaluated based on the light effect feature map of each frame in the image frame sequence under different light effects, and the light effect feature value of the image frame sequence under different light effects is obtained; the light effect purity factor of each frame under different light effects is calculated based on the light effect feature value. The flash switching timing data corresponding to the integrated optical imaging unit is obtained, and the flash switching timing data is used to perform flash switching timing calibration on the light effect purity factor corresponding to each frame under different light effects, so as to eliminate the light effect interference factor corresponding to the flash switching moment and generate the light effect purity parameter corresponding to each frame under different light effects after timing calibration. Based on the light effect purity parameters of each frame under different light effects after time-series calibration, the statistical distribution of light effect purity is obtained, and the screening benchmark value of light effect frames under different light effects is constructed based on the statistical distribution of light effect purity. Based on the screening benchmark value of light effect frames under different light effects, the image frame sequence of reflected light and transmitted light corresponding to the PCB board is pre-screened to obtain the initial screening image frame set corresponding to reflected light and transmitted light. Obtain the light effect attenuation data corresponding to each frame in the initial screening image frame set, and calculate the inter-frame light effect coordination factor under different light effects based on the light effect attenuation data corresponding to each frame in the initial screening image frame set; perform secondary purification processing on the initial screening image frame set corresponding to reflected light and transmitted light based on the inter-frame light effect coordination factor under different light effects to remove abnormal frames corresponding to the inter-frame light effect coordination factor deviating from the normal range, and obtain the reflected light image frame set and the transmitted light image frame set.

4. The method for reading PCB QR codes and binding PCB inspection data according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Construct a PCB board image coordinate system based on reflected light image data and transmitted light image data, and determine the position reference association data corresponding to the reflected light image and transmitted light image based on the PCB board image coordinate system; Step S22: Perform connected component analysis on the transmitted light image data to complete blob extraction and obtain the full blob dataset; Step S23: Perform morphological representation and classification filtering on the full blob dataset to remove invalid and stray blobs and obtain a valid blob set; Step S24: By combining the UI interactive interface to display the effective blob set, and selecting the positioning blob and QR code reading location, the baseline positioning parameters and QR code reading coordinates are obtained; Step S25: Obtain the features of the transmitted light image based on the transmitted light image data, and construct a standardized comparison template based on the benchmark positioning parameters, QR code reading coordinates and transmitted light image features to generate a standard file.

5. The method for reading PCB QR codes and binding PCB inspection data according to claim 4, characterized in that, Step S22 includes the following steps: Gray-scale thresholding is performed on the transmitted light image data to distinguish between the foreground blob region and the background region, resulting in a binarized image. Morphological operations are performed on the binarized image to eliminate tiny noise points and fill in holes inside the blob, resulting in an optimized binarized image. Connectivity labeling analysis is performed on the corresponding foreground blob regions in the optimized binarized image to extract all independent blobs; Record the morphological parameters, coordinate information, and grayscale features of each independent blob to generate a blob feature dataset; Integrate the blob feature datasets corresponding to all independent blobs to form a full blob dataset.

6. The method for reading PCB QR codes and binding PCB inspection data according to claim 5, characterized in that, The connected component labeling analysis of the corresponding foreground blob region in the optimized binarized image includes the following steps: Extract the pixel distribution data of the foreground region corresponding to the foreground blob region in the optimized binarized image to obtain the pixel coordinate dataset; Based on the pixel coordinate dataset, the connectivity features corresponding to adjacent pixels are calculated to generate a pixel connectivity matrix. Based on the pixel connectivity matrix, discrete foreground blob regions are clustered and labeled to obtain a preliminary set of connected components; domain differentiation identifier data is generated based on this preliminary set of connected components. By combining the grayscale features of the transmitted light image, each connected component is verified, and broken pseudo-connected components are removed to obtain valid connected component data. Extract the contour boundaries of the valid connected component data to generate all independent blobs.

7. The method for reading PCB QR codes and binding PCB inspection data according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Obtain fixed positional relationship data of source images based on location benchmark association data, and construct a positioning comparison framework based on the standard file and the fixed positional relationship data of source images; Step S32: Based on the positioning comparison framework, compare the transmitted light image corresponding to the PCB board with the corresponding positioning in the standard file using blobs to perform positioning matching and generate positioning calibration data; Step S33: Based on the comparison of positioning calibration data, verify each blob in the corresponding transmitted light image of the PCB board to obtain the blob quality judgment result; Step S34: Based on the fixed positional relationship data of the same source image, map the corresponding QR code reading coordinates in the standard file to the reflected light image to lock the QR code recognition area; Step S35: Perform texture enhancement and distortion correction on the QR code recognition area, and generate QR code information and verification results by combining the QR code recognition area reading with the blob quality judgment results.

8. The method for reading PCB QR codes and binding PCB inspection data according to claim 7, characterized in that, Step S32 includes the following steps: Extract the set of valid blobs from the transmitted light image corresponding to the PCB board to obtain the comparison blob dataset; Based on the localization comparison framework, the blob dataset is compared with the corresponding localization blob in the standard file to perform feature comparison, extract matching feature points and calculate feature fit. Based on feature matching, core matching blobs are selected, and the positional correspondence between the PCB board and the standard document is established to generate preliminary positioning data. Based on the fixed positional relationship data of the same source image, the initial positioning data is corrected for deviation to eliminate the influence of small offsets during the image acquisition process. The reflected light and transmitted light image coordinate mapping parameters in the fixed positional relationship data of the same source image are retrieved to construct the deviation correction benchmark. Texture feature points in the reflected light image of the PCB board are extracted and compared with the feature points in the reflected light image in the standard file to calculate the horizontal and vertical offsets. Based on the offsets, a deviation correction model is constructed, and the blob coordinates in the preliminary positioning data are adaptively calibrated based on the deviation correction model to generate calibrated coordinates. At the same time, the positional coordination of the two images of the PCB board is verified to generate positional coordination verification data. Positioning calibration data is generated based on location collaborative verification data and calibrated coordinates.

9. The method for reading PCB QR codes and binding PCB inspection data according to claim 7, characterized in that, Step S34 includes the following steps: Step S341: Based on the fixed positional relationship data of the same source images, analyze the spatial correspondence pattern of the two images to obtain the coordinate mapping reference factor; Step S342: Retrieve the coordinates of the QR code corresponding to the standard file, and perform coordinate mapping calculation on the reflected light image in combination with the coordinate mapping reference factor to generate the initial mapped coordinates of the reflected light image; Step S343: Extract texture features of the corresponding region of the reflected light image based on the initial mapping coordinates of the reflected light image, and construct a texture verification dataset; Step S344: Based on the texture verification dataset, compare the QR code texture features corresponding to the standard file and calculate the coordinate deviation correction amount; Step S345: Based on the coordinate deviation correction, calibrate the initial mapping coordinates of the reflected light image to generate an identification mapping coordinate set, and delineate and lock the QR code identification area based on the identification mapping coordinate set.

10. The method for reading PCB QR codes and binding PCB inspection data according to claim 9, characterized in that, Step S341 includes the following steps: Extract the fixed positional relationship data of the same source image to separate the feature anchor point information corresponding to the reflected light and transmitted light images, and obtain the dual light effect anchor point dataset; Matching analysis was performed on the dual-light-effect anchor point dataset to establish a one-to-one correspondence between anchor points and generate an anchor point association map. Based on the anchor point association map, the spatial position deviation corresponding to the anchor points of the two images is calculated to obtain the anchor point deviation dataset; By combining the anchor point deviation dataset to analyze the spatial correspondence, a coordinate mapping association model is constructed, and a coordinate mapping reference factor is generated through the calculation of the coordinate mapping association model.

Images