A multi-stage multi-defect visual inspection method and system for an automatic web inspection machine
By employing a multi-stage visual inspection method, combined with median filtering, morphological processing, and Gaussian blurring filter, the efficiency and accuracy issues of various defect detection in capacitive screen printing were resolved, achieving efficient defect detection and report generation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGGUAN UNIV OF TECH
- Filing Date
- 2026-01-26
- Publication Date
- 2026-06-12
Smart Images
Figure CN122199379A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of machine vision technology, and in particular to a multi-stage, multi-defect visual inspection method and system for an automatic mesh inspection machine. Background Technology
[0002] In the field of screen printing for electronic components, there are very high requirements for screen printing quality inspection, including inspection accuracy, inspection speed, and inspection types. Currently, in the existing capacitor screen printing production process, manual inspection of defects in capacitor screen printing suffers from low efficiency and poor accuracy, making it difficult to meet the needs of large-scale production. Traditional machine vision inspection methods often can only detect single types of defects and cannot effectively detect and identify multiple different types of defects. Moreover, due to the complex and variable inspection environment, the images acquired during actual inspection have high noise levels, resulting in inconsistent image quality. Especially when the screen printing is black and white interwoven and different defects appear in different areas, regional inspection is required during actual screen printing inspection, leading to low inspection efficiency. The multi-stage, multi-defect visual inspection method and system for automatic screen printing proposed in this invention, through multi-stage processing steps, from image preprocessing to defect detection and report generation, can quickly and accurately detect multiple defects in capacitor screen printing, such as screen blockage, latex peeling, and dirt. Summary of the Invention
[0003] To overcome the shortcomings of existing technologies, this invention provides a multi-stage, multi-defect visual inspection method and system for automatic screen printing machines, enabling rapid and accurate detection of various defects in capacitor screen printing, such as screen clogging defects, latex peeling defects, and dirt defects, thereby improving the overall quality and production efficiency of capacitor screen printing products.
[0004] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:
[0005] The first aspect of this application provides a multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine, comprising the following steps: S101. Perform median filtering and morphological image preprocessing on the capacitive screen printing image signal to enhance image quality; S102. Based on a multi-level filtering mechanism, perform coarse localization processing of the contour extraction region on the image preprocessing results; S103. Perform feature enhancement and fine localization processing on the coarse localization results of the region based on Gaussian fuzzy filter; S104. Perform defect detection processing on the region fine localization processing results based on morphological opening and closing operations; S105. Perform defect marking and information annotation processing on the defect detection results, and generate a defect detection report.
[0006] Furthermore, median filtering and morphological image preprocessing are performed on the capacitive screen printing image signal to enhance image quality, including the following steps: The capacitive screen printing image is converted to grayscale and median filtering is used to suppress salt-and-pepper noise and preserve edge structure information; Image quality is enhanced by combining Otsu thresholding with a sequence of morphological operations, which includes closing and opening operations. Closing operations fill small holes inside rectangles, while opening operations eliminate isolated noise points.
[0007] Furthermore, the coarse localization of contour extraction regions based on the image preprocessing results using a multi-level filtering mechanism includes the following steps: Only the outermost contour is extracted from the image preprocessing results; The extracted outermost contour is represented by the minimum bounding rectangle, whose attributes include center coordinates, length, width, and rotation angle. By setting area thresholds and aspect ratio ranges, rectangular candidate areas that meet the characteristics of screen printing are selected. Edge buffer checks are used to distinguish between complete and incomplete rectangles among the screened candidate rectangular regions that meet the screen mesh characteristics.
[0008] Furthermore, the feature enhancement and fine localization processing of the region coarse localization results based on the Gaussian blur filter includes the following steps: Image smoothing processing is performed on the coarse localization results of the region based on Gaussian blur filter to enhance the contour features of the target region; Image segmentation is performed on the image smoothing result to separate the target region from the background, thereby achieving precise localization of the capacitive screen printing feature region; Morphological operations are used to optimize the image segmentation results.
[0009] Furthermore, the defect detection processing based on the region fine localization results using morphological opening and closing operations includes the following steps: Based on morphological opening and closing operations, the results of precise regional positioning are used to detect and process network blockage defects. Latex shedding defect detection is performed based on the results of precise region localization processing using morphological opening and closing operations. Dirt and defect detection processing is performed based on the results of precise region localization processing using morphological opening and closing operations.
[0010] Furthermore, the process of detecting network blockage defects based on the region's precise localization results using morphological opening and closing operations includes the following steps: Potential white mesh regions are extracted from each candidate rectangular region in the image by threshold segmentation. The black gaps in the white area are filled using a closing operation to enhance the connectivity of the area; Opening operations are used to remove small-area noise in binary images to optimize defect contours; The white area is then further segmented to identify dark-colored blockage defects.
[0011] Furthermore, the latex peeling defect detection processing based on the region fine-positioning processing results using morphological opening and closing operations includes the following steps: The detection area is limited to the offset area outside the large white rectangle, and a detection mask is constructed by extending the rectangle outward by a specific pixel distance. Set the interior of the offset rectangle as the background and the outer area as the foreground; The opening operation is used to remove noise from the image after threshold segmentation. The discrete detachment regions are connected by a closing operation to form a coherent defect profile, and latex detachment defects are identified.
[0012] Furthermore, the process of detecting contamination defects based on the region's precise localization results using morphological opening and closing operations includes the following steps: The detection area is isolated by constructing a mask, where the inside of the rectangle is set as the background and the outside area is used as the foreground. Grayscale extraction is performed on the foreground region, and threshold segmentation is applied to highlight the features of dirt and defects. Opening and closing operations are used to remove noise in binarization; Closure operations are used to fill the gaps in the white area to enhance the integrity of the dirty area and to identify dirt defects.
[0013] Furthermore, the defect detection results are processed by defect marking and information annotation, and a defect detection report is generated, including the following steps: By drawing graphic frames of different colors, the defect detection results are marked with defects and information annotations, including detection time, image size, defect coordinates, and defect type. Generate a defect detection report from images that have been marked with defects and annotated with information; The defect detection report is encapsulated into a JSON structure and saved as a UTF-8 encoded JSON file.
[0014] The second aspect of this application provides a multi-stage, multi-defect visual inspection system for an automatic mesh inspection machine, comprising: The image acquisition unit is used to acquire capacitive screen printing image signals; The image preprocessing unit is used to perform median filtering and morphological image preprocessing on the capacitive screen printing image signal to enhance image quality; The first data processing unit is used to perform coarse localization processing of the contour extraction region based on the image preprocessing results using a multi-level filtering mechanism. The second data processing unit is used to perform feature enhancement and fine localization processing on the coarse localization results of the region based on a Gaussian blur filter. The defect identification unit is used to perform defect detection processing on the results of precise region localization based on morphological opening and closing operations. The report generation unit is used to perform defect marking and information annotation processing on the defect detection results, and generate a defect detection report.
[0015] The beneficial effects of this application are as follows: Image quality is enhanced through median filtering and morphological image preprocessing; a multi-level filtering mechanism accurately coarsely locates the contour extraction region; and by screening and distinguishing effective regions, interference from invalid regions is avoided. Gaussian blurring filters further enhance the features of the target region, achieving precise region localization and making defect detection more accurate. In the defect detection stage, morphological opening and closing operations are used to target different types of defects, such as mesh blockage defects, latex peeling defects, and dirt defects, improving the accuracy and reliability of defect detection. Through defect marking and information annotation processing, and the generation of detailed defect detection reports, defects in capacitor wire mesh can be detected quickly and accurately, allowing for timely identification of problems in the production process, reducing the defect rate, and improving production efficiency. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a schematic diagram illustrating the steps of a multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine according to the present invention. Detailed Implementation
[0018] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0019] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0020] Example 1 like Figure 1 As shown, a multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine includes the following steps: S101. Perform median filtering and morphological image preprocessing on the capacitive screen printing image signal to enhance image quality; Capacitive screen printing image signals can be acquired using a 16K line scan camera, a high-brightness line scan light source, and a motion control board. After acquiring the image signal, median filtering is used to remove noise points, preventing noise from interfering with subsequent detection. Median filtering sorts the grayscale values of pixels within a window and takes the median as the new grayscale value of the center pixel, effectively preserving edge information. Morphological processing improves the image's shape and structure, including closing and opening operations. Closing uses dilation followed by erosion, while opening uses erosion followed by dilation. Dilation enlarges the target object, connecting previously separated parts and better identifying the overall shape; erosion shrinks the target object, removing small interfering parts. The capacitive screen printing image signal, after median filtering and morphological preprocessing, has clearer contours and more prominent features, more accurately reflecting the actual state of the screen printing, thus improving the accuracy and reliability of subsequent detection.
[0021] Median filtering and morphological image preprocessing are performed on capacitive screen printing image signals to enhance image quality, including the following steps: The capacitive screen printing image is converted to grayscale and median filtering is used to suppress salt-and-pepper noise and preserve edge structure information; Image quality is enhanced by combining Otsu thresholding with a sequence of morphological operations, which includes closing and opening operations. Closing operations fill small holes inside rectangles, while opening operations eliminate isolated noise points.
[0022] For example, for a capacitor screen-printed image with significant salt-and-pepper noise and small internal holes, the image is first converted to grayscale. Then, median filtering is applied to each pixel, sorting its neighboring pixel values and using the median value as the new value. After median filtering, the salt-and-pepper noise is effectively suppressed while preserving edge structure information. Next, the Otsu thresholding method is used to automatically calculate a suitable threshold, binarizing the image. Then, a sequence of morphological operations is used to enhance image quality. First, a closing operation (kernel size 11×11) is performed to dilate the binarized image, filling the small holes within the rectangles. Then, an opening operation (kernel size 3×3) is performed to erode the dilated image, eliminating isolated noise points. After processing with this sequence of morphological operations, the image quality is significantly enhanced, providing clearer image data for subsequent visual inspection.
[0023] S102. Based on a multi-level filtering mechanism, perform coarse localization processing of the contour extraction region on the image preprocessing results; Median filtering and morphological image preprocessing are performed on the capacitive screen printing image signal to obtain the image preprocessing result. By retrieving only the outermost contour from the image preprocessing result, redundant contour processing is reduced, avoiding interference from nested contours and improving image processing speed. Each extracted contour is represented by a minimum bounding rectangle, through which key geometric attributes such as center coordinates, length, width, and rotation angle can be obtained. To further filter out candidate rectangular regions that conform to screen printing characteristics, area thresholds and aspect ratio ranges are set to exclude interference contours that obviously do not conform to template characteristics. Finally, edge buffer checks are used to distinguish between complete rectangles and incomplete rectangles caused by image truncation, ensuring that subsequent defect detection is only performed on valid areas. This coarse-area localization method significantly improves localization accuracy while maintaining recall through a multi-level filtering mechanism.
[0024] The coarse localization of contour extraction regions based on image preprocessing results using a multi-level filtering mechanism includes the following steps: Only the outermost contour is extracted from the image preprocessing results; The extracted outermost contour is represented by the minimum bounding rectangle, whose attributes include center coordinates, length, width, and rotation angle. By setting area thresholds and aspect ratio ranges, rectangular candidate areas that meet the characteristics of screen printing are selected. Edge buffer checks are used to distinguish between complete and incomplete rectangles among the screened candidate rectangular regions that meet the screen mesh characteristics.
[0025] For example, in the contour extraction stage, the RETR_EXTERNAL mode can be used to preprocess the image results, initially retrieving only the outermost contour to eliminate interference from nested contours. The minimum bounding rectangle is then used to represent these contours, obtaining important attributes such as their center coordinates, size, and rotation angle, providing foundational data for further analysis. A reasonable area threshold (e.g., a minimum area of 660 pixels) can be set. 2 The image is filtered to select candidate rectangular areas that meet the requirements, based on the aspect ratio range (e.g., a maximum aspect ratio of 100), while excluding areas that do not meet the characteristics. Edge buffer checks (e.g., 10-pixel boundaries) carefully distinguish complete rectangles from incomplete rectangles caused by image truncation. After contour extraction and coarse positioning, the contours and features of the capacitive screen printing in the image are clearer and more defined, providing a reliable basis for subsequent accurate defect detection and analysis. This helps to promptly identify potential problems in the screen printing process, ensuring product quality and smooth production. In subsequent defect detection, only valid areas are targeted, improving detection accuracy and efficiency.
[0026] S103. Perform feature enhancement and fine localization processing on the coarse localization results of the region based on Gaussian fuzzy filter; A multi-level filtering mechanism is used to perform coarse localization of the contour extraction region in the image preprocessing results, yielding a coarse localization result. A Gaussian blur filter is then used to enhance the feature and refine the localization of the coarse localization result, further improving the clarity and recognizability of target features in the image. The Gaussian blur filter smooths the image, reducing noise interference and making the contour features of the target region more prominent, thus enhancing the edge feature information of the capacitor screen printing contour and making the contour features sharper. A threshold segmentation algorithm is then used to segment the smoothed image, separating the target region from the background and avoiding interference from background information in subsequent defect detection, achieving precise localization of the feature regions of the capacitor screen printing. Morphological operations are then used to optimize the segmented image. Dilation fills small holes in the target region, making it more complete, while erosion removes small protrusions at the edges of the target region, making its shape more regular. The image optimized by morphological operations has clearer and more reliable target region features, contributing to improved accuracy in defect detection.
[0027] The process of feature enhancement and fine localization of the region coarse localization results based on Gaussian blur filter includes the following steps: Image smoothing processing is performed on the coarse localization results of the region based on Gaussian blur filter to enhance the contour features of the target region; Image segmentation is performed on the image smoothing result to separate the target region from the background, thereby achieving precise localization of the capacitive screen printing feature region; Morphological operations are used to optimize the image segmentation results.
[0028] S104. Perform defect detection processing on the region fine localization processing results based on morphological opening and closing operations; Morphological opening and closing operations are used to perform defect detection processing on the results of precise region localization. Defect detection includes detecting mesh blockage, latex detachment, and contamination. Morphological opening and closing operations consist of morphological opening and closing operations. The morphological opening operation (erosion followed by dilation) is mainly used to eliminate minor noise and separate adjacent objects, while the morphological closing operation (dilation followed by erosion) is used to fill internal voids and connect adjacent components, thereby improving the connectivity of the defective region.
[0029] Defect detection processing based on morphological opening and closing operations for region fine localization includes the following steps: Based on morphological opening and closing operations, the results of precise regional positioning are used to detect and process network blockage defects. Latex shedding defect detection is performed based on the results of precise region localization processing using morphological opening and closing operations. Dirt and defect detection processing is performed based on the results of precise region localization processing using morphological opening and closing operations.
[0030] The process of detecting network blockage defects based on the results of precise region localization using morphological opening and closing operations includes the following steps: Potential white mesh regions are extracted from each candidate rectangular region in the image by threshold segmentation. The black gaps in the white area are filled using a closing operation to enhance the connectivity of the area; Opening operations are used to remove small-area noise in binary images to optimize defect contours; The white area is then further segmented to identify dark-colored blockage defects.
[0031] The latex peeling defect detection processing based on the region fine positioning processing results using morphological opening and closing operations includes the following steps: The detection area is limited to the offset area outside the large white rectangle, and a detection mask is constructed by extending the rectangle outward by a specific pixel distance. Set the interior of the offset rectangle as the background and the outer area as the foreground; The opening operation is used to remove noise from the image after threshold segmentation. The discrete detachment regions are connected by a closing operation to form a coherent defect profile, and latex detachment defects are identified.
[0032] The process of detecting contamination defects based on the results of precise region localization processing using morphological opening and closing operations includes the following steps: The detection area is isolated by constructing a mask, where the inside of the rectangle is set as the background and the outside area is used as the foreground. Grayscale extraction is performed on the foreground region, and threshold segmentation is applied to highlight the features of dirt and defects. Opening and closing operations are used to remove noise in binarization; Closure operations are used to fill the gaps in the white area to enhance the integrity of the dirty area and to identify dirt defects.
[0033] For example, in detecting screen blockage defects, filling the black gaps in the white area enhances the connectivity of the region, allowing blockage areas that might otherwise be misjudged due to gaps to be fully identified. Removing small-area noise makes the defect outline clearer, facilitating subsequent analysis and processing. Secondary segmentation further accurately identifies dark-colored blockage defects, precisely locating potential blockages. In detecting latex detachment defects, the detection area is limited to an offset region outside the large white rectangle. Using a constructed detection mask, interference from the main filter body is effectively eliminated, focusing on edge areas prone to latex detachment. After noise removal through opening operations and connection of discrete detachment areas through closing operations, a coherent defect outline is clearly presented, improving the accuracy of latex detachment defect identification. In detecting dirt defects, a mask is constructed to isolate the detection area, distinguishing the foreground from the background. Grayscale extraction and threshold segmentation highlight dirt defect features. After noise removal and gap filling through opening and closing operations, the integrity of the dirty area is enhanced, thereby improving the accuracy of dirt defect identification.
[0034] S105. Perform defect marking and information annotation on the defect detection results, and generate a defect detection report; Defect detection is performed on the region precision localization results using morphological opening and closing operations to obtain the defect detection results. Different colored graphical frames are used to mark and annotate the defect detection results; for example, red circles mark blocked mesh defects, blue circles mark latex peeling defects, and yellow circles mark dirt defects. Information annotations include detection time, image size, defect coordinates, and defect type. A defect detection report is generated from the defect-marked and annotated image, and the report is encapsulated in a JSON structure and saved as a UTF-8 encoded JSON file to ensure data parsing by the host computer.
[0035] The process of marking and annotating defects in the detection results and generating a defect detection report includes the following steps: By drawing graphic frames of different colors, the defect detection results are marked with defects and information annotations, including detection time, image size, defect coordinates, and defect type. Generate a defect detection report from images that have been marked with defects and annotated with information; The defect detection report is encapsulated into a JSON structure and saved as a UTF-8 encoded JSON file.
[0036] Example 2 The above describes a multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine provided in the embodiments of this application. The following describes a multi-stage, multi-defect visual inspection system for an automatic mesh inspection machine provided in the embodiments of this application.
[0037] A multi-stage, multi-defect visual inspection system for an automatic mesh inspection machine includes: The image acquisition unit is used to acquire capacitive screen printing image signals; The image preprocessing unit is used to perform median filtering and morphological image preprocessing on the capacitive screen printing image signal to enhance image quality; The first data processing unit is used to perform coarse localization processing of the contour extraction region based on the image preprocessing results using a multi-level filtering mechanism. The second data processing unit is used to perform feature enhancement and fine localization processing on the coarse localization results of the region based on a Gaussian blur filter. The defect identification unit is used to perform defect detection processing on the results of precise region localization based on morphological opening and closing operations. The report generation unit is used to perform defect marking and information annotation processing on the defect detection results, and generate a defect detection report.
[0038] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system and unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0039] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine, characterized in that, Includes the following steps: S101. Perform median filtering and morphological image preprocessing on the capacitive screen printing image signal to enhance image quality; S102. Based on a multi-level filtering mechanism, perform coarse localization processing of the contour extraction region on the image preprocessing results; S103. Perform feature enhancement and fine localization processing on the coarse localization results of the region based on Gaussian fuzzy filter; S104. Perform defect detection processing on the region fine localization processing results based on morphological opening and closing operations; S105. Perform defect marking and information annotation processing on the defect detection results, and generate a defect detection report.
2. The multi-stage, multi-defect visual inspection method for automatic mesh inspection machines according to claim 1, characterized in that, Step S101 includes the following steps: The capacitive screen printing image is converted to grayscale and median filtering is used to suppress salt-and-pepper noise and preserve edge structure information; Image quality is enhanced by combining Otsu thresholding with a sequence of morphological operations, which includes closing and opening operations. Closing operations fill small holes inside rectangles, while opening operations eliminate isolated noise points.
3. The multi-stage, multi-defect visual inspection method for automatic mesh inspection machines according to claim 1, characterized in that, Step S102 includes the following steps: Only the outermost contour is extracted from the image preprocessing results; The extracted outermost contour is represented by the minimum bounding rectangle, whose attributes include center coordinates, length, width, and rotation angle. By setting area thresholds and aspect ratio ranges, rectangular candidate areas that meet the characteristics of screen printing are selected. Edge buffer checks are used to distinguish between complete and incomplete rectangles among the screened candidate rectangular regions that meet the screen mesh characteristics.
4. The multi-stage, multi-defect visual inspection method for automatic mesh inspection machines according to claim 1, characterized in that, Step S103 includes the following steps: Image smoothing processing is performed on the coarse localization results of the region based on Gaussian blur filter to enhance the contour features of the target region; Image segmentation is performed on the image smoothing result to separate the target region from the background, thereby achieving precise localization of the capacitive screen printing feature region; Morphological operations are used to optimize the image segmentation results.
5. The multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine according to claim 1, characterized in that, Step S104 includes the following steps: Based on morphological opening and closing operations, the results of precise regional positioning are used to detect and process network blockage defects. Latex shedding defect detection is performed based on the results of precise region localization processing using morphological opening and closing operations. Dirt and defect detection processing is performed based on the results of precise region localization processing using morphological opening and closing operations.
6. The multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine according to claim 5, characterized in that, The process of detecting network blockage defects based on the results of precise regional localization processing using morphological opening and closing operations includes the following steps: Potential white mesh regions are extracted from each candidate rectangular region in the image by threshold segmentation. The black gaps in the white area are filled using a closing operation to enhance the connectivity of the area; Opening operations are used to remove small-area noise in binary images to optimize defect contours; The white area is then further segmented to identify dark-colored blockage defects.
7. The multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine according to claim 5, characterized in that, The latex peeling defect detection processing based on the region fine positioning processing results of morphological opening and closing operations includes the following steps: The detection area is limited to the offset area outside the large white rectangle, and a detection mask is constructed by extending the rectangle outward by a specific pixel distance. Set the interior of the offset rectangle as the background and the outer area as the foreground; The opening operation is used to remove noise from the image after threshold segmentation. The discrete detachment regions are connected by a closing operation to form a coherent defect profile, and latex detachment defects are identified.
8. The multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine according to claim 5, characterized in that, The process of detecting contamination defects based on the results of precise region localization processing using morphological opening and closing operations includes the following steps: The detection area is isolated by constructing a mask, where the inside of the rectangle is set as the background and the outside area is used as the foreground. Grayscale extraction is performed on the foreground region, and threshold segmentation is applied to highlight the features of dirt and defects. Opening and closing operations are used to remove noise in binarization; Closure operations are used to fill the gaps in the white area to enhance the integrity of the dirty area and to identify dirt defects.
9. The multi-stage, multi-defect visual inspection method for an automatic mesh inspection machine according to claim 1, characterized in that, Step S105 includes the following steps: The detection area is isolated by constructing a mask, where the inside of the rectangle is set as the background and the outside area is used as the foreground. Grayscale extraction is performed on the foreground region, and threshold segmentation is applied to highlight the features of dirt and defects. Opening and closing operations are used to remove noise in binarization; Closure operations are used to fill the gaps in the white area to enhance the integrity of the dirty area and to identify dirt defects.
10. A multi-stage, multi-defect visual inspection system for an automatic mesh inspection machine, used to implement the multi-stage, multi-defect visual inspection method for the automatic mesh inspection machine according to any one of claims 1-9, characterized in that, include: The image acquisition unit is used to acquire capacitive screen printing image signals; The image preprocessing unit is used to perform median filtering and morphological image preprocessing on the capacitive screen printing image signal to enhance image quality; The first data processing unit is used to perform coarse localization processing of the contour extraction region based on the image preprocessing results using a multi-level filtering mechanism. The second data processing unit is used to perform feature enhancement and fine localization processing on the coarse localization results of the region based on a Gaussian blur filter. The defect identification unit is used to perform defect detection processing on the results of precise region localization based on morphological opening and closing operations. The report generation unit is used to perform defect marking and information annotation processing on the defect detection results, and generate a defect detection report.