A visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates

Abstract

This invention relates to the field of online visual inspection technology for printed materials, and discloses a visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates. The system includes: acquiring planar polarized images, micro-bending polarized images, and edge reference strip images of the printed material to be inspected; performing illumination correction on the two types of images using the edge reference strip images, and registering the micro-bending image to planar coordinates; generating a water-based ink printing area map from the cross-polarized images of the planar images; generating a micro-bending adhesion defect response map and a micro-bending edge change map based on the two types of images; generating an adhesion defect candidate area map within the water-based ink printing area map, and outputting an adhesion defect marker map after connectivity confirmation. This invention is used for online adhesion defect detection of water-based ink printed materials on environmentally friendly substrates.

Description

Technical Field

[0001] This invention relates to the field of online visual inspection technology for printed materials, and more specifically, to a visual inspection system for defects in the adhesion of water-based inks to environmentally friendly substrates. Background Technology

[0002] Water-based inks on environmentally friendly substrates are widely used in packaging, labeling, film lamination, and paper-based coating printing. These substrates can include biodegradable films, paper-based composite films, recycled fiber coated films, and low surface energy substrates treated with corona or plasma. Compared to conventional absorbent paper, the surface energy, coating water absorption capacity, micro-roughness, and surface treatment uniformity of these substrates have a greater impact on the film formation and adhesion stability of water-based inks. In actual production, even if pre-press wetting meets process requirements, batch variations in roll stock, local coating fluctuations, surface contaminants, ink drying shrinkage, and changes in conveyor tension can still lead to insufficient ink adhesion in localized areas.

[0003] Current online printing inspection typically uses a color orthogonal camera to capture images, and then identifies appearance problems such as dirt, missing prints, misregistration, color difference, and scratches through template comparison, grayscale difference, color thresholding, edge detection, or depth detection models. This approach is suitable for defects with significant grayscale, color, or contour variations, but its adaptability to early stages of insufficient adhesion of water-based inks is limited. Early insufficient adhesion usually manifests as localized micro-voids at the interface, slight ink layer lifting, narrow peeling edges, or micro-cracks. The brightness and color differences produced in ordinary color orthogonal images are relatively small and easily confused with substrate fiber texture, coating particles, filler scattering, printed pattern details, and uneven lighting.

[0004] In continuous conveying inspection scenarios, there are also on-site factors such as slight lateral movement of the roll, tension fluctuations, light source angle drift, camera exposure fluctuations, and differences in substrate reflection. If only static whiteboard correction or single-frame image thresholding is used, changes in lighting, pattern edges, or substrate textures are easily judged as defects. If only template difference is used, registration errors and micro-slippage during conveying can create small-scale pseudo-defects. If relying on training samples, the number, morphological distribution, and batch coverage of early samples with insufficient adhesion are often inadequate, which can easily affect on-site transfer and use.

[0005] Therefore, the detection of adhesion defects of water-based inks on environmentally friendly substrates needs to incorporate imaging acquisition conditions, conveyor position calibration, illumination correction, and defect image generation into the detection process while the roll material is running continuously. More specifically, it is necessary to generate comparable polarized images of the same printed micro-area at the planar station and the micro-bending station, and to use edge reference bars to reduce the influence of illumination and inter-line fluctuations. Then, the micro-bending station image is registered to the planar station coordinates, so that the polarization reflection changes and edge changes caused by local insufficient ink adhesion can be separated from the pattern color, substrate texture, and uneven illumination, thereby generating image results that indicate the location and boundary of the adhesion defects. Summary of the Invention

[0006] This invention provides a visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates, which solves the technical problems mentioned in the background art.

[0007] This invention provides the following technical solution: A visual inspection system for defects in water-based ink adhesion on environmentally friendly substrates is applied to a visual inspection system comprising a planar support imaging station and a micro-bent guide roller imaging station. The two imaging stations are arranged sequentially along a continuous conveying direction. Each imaging station is equipped with a light source, a camera, a parallel polarization channel, and a cross polarization channel, and is configured to perform: Collect planar polarization images, micro-bending polarization images, and edge reference strip images of the water-based ink printed on the environmentally friendly substrate to be tested; The planar workstation polarization image and the micro-bending workstation polarization image are respectively illuminated and corrected using the edge reference strip image. The illuminated and corrected micro-bending workstation polarization image is then registered to the planar workstation coordinates to obtain the planar workstation corrected image and the micro-bending workstation registered image. A water-based ink printing area map is generated from the cross-polarization image in the planar workstation correction image; Based on the planar station correction image, the micro-bending station registration image, and the water-based ink printing area map, a micro-bending adhesion defect response map and a micro-bending edge change map are generated. Within the water-based ink printing area map, an adhesion defect candidate area map is generated based on the micro-bending adhesion defect response map; By combining the microbend attachment defect response map and the microbend edge change map, the connectivity of the attachment defect candidate region map is confirmed to obtain the confirmed attachment defect region map. Output an attachment defect marking map based on the confirmed attachment defect area map.

[0008] Preferably, the process involves acquiring planar polarization images, micro-bending polarization images, and edge reference strip images of the water-based ink printed on the environmentally friendly substrate to be tested, including: The water-based ink printed material on the environmentally friendly substrate to be tested is passed sequentially through the planar support imaging station and the micro-bent guide roller imaging station. In the planar support imaging station, the area to be photographed is brought into contact with the planar support surface to form a planar workstation imaging state; In the micro-bending guide roller imaging station, a micro-bending station imaging state is formed by a cylindrical guide roller that does not contact the printing surface. The encoder triggers a linear or area color camera, enabling the same printed micro-area to form paired images at two imaging stations. Parallel polarized images, cross-polarized images, and edge reference strip images located at the edge of the substrate and not overlapping with the pattern to be detected are acquired from the same row of images or the same image.

[0009] Preferably, the planar station polarization image and the micro-bending station polarization image are respectively illuminated and corrected using the edge reference strip image, and the illuminated and corrected micro-bending station polarization image is registered to the planar station coordinates to obtain a planar station corrected image and a micro-bending station registered image, including: Perform dark field subtraction for each imaging station, each polarization channel, and each color channel; The brightness of the edge reference strip image is read along the continuous conveying direction, and the planar station reference lighting map and the micro-bend station reference lighting map are obtained by interpolation along the horizontal direction of the image. The image after dark field subtraction is illuminated using the planar workstation reference lighting diagram and the micro-bend workstation reference lighting diagram to obtain an illuminated corrected image. Calculate the coordinate transformation relationship of the workstation based on the printing boundary, registration mark, and pattern texture area; The illumination correction image of the micro-bending guide roller imaging station is resampled according to the workstation coordinate transformation relationship to obtain the micro-bending workstation registration image, and the illumination correction image of the planar support imaging station is used as the planar workstation correction image.

[0010] Preferably, generating a water-based ink printing area map from the cross-polarization image in the planar station correction image includes: Select a planar workstation cross-polarization image from the planar workstation correction image; Based on the brightness of the planar workstation reference lighting map and the pixel brightness of each color channel in the planar workstation cross-polarization image, an aqueous ink absorption map for each color channel is formed. In the water-based ink absorption maps of the red, green, and blue channels, the channel value with the largest absorption value is selected pixel by pixel to form an ink absorption brightness map; Perform inter-class variance threshold segmentation on the ink absorption brightness map to obtain the water-based ink printing area map; The printing area pixels in the water-based ink printing area map are used to generate the micro-bending adhesion defect response map and the adhesion defect candidate area map.

[0011] Preferably, based on the planar station correction image, the micro-bending station registration image, and the water-based ink printing area map, a micro-bending adhesion defect response map and a micro-bending edge change map are generated, including: In the planar workstation correction image and the micro-bending workstation registration image, channel difference processing and reflection brightness normalization processing are performed on the parallel polarization image and the cross polarization image respectively to obtain the planar workstation mirror reflection image and the micro-bending workstation mirror reflection image. The difference between the micro-bending station mirror reflection image and the planar station mirror reflection image is used to obtain the micro-bending mirror change image. The cross-polarization images in the planar station correction image and the micro-bending station registration image are respectively averaged in three color channels and edge extracted to obtain diffuse reflection edge maps of the two imaging stations; The difference between the two diffuse edge maps is corrected by using the background value of the neighboring edge within the water-based ink printing area map to obtain the micro-curved edge variation map; By combining the micro-bend mirror surface change map, the micro-bend edge change map, and the neighborhood mirror surface fluctuation value within the water-based ink printing area map, the micro-bend adhesion defect response map is obtained.

[0012] Preferably, within the water-based ink printing area map, generating an adhesion defect candidate area map based on the micro-bending adhesion defect response map includes: The optical resolution unit is determined based on the camera pixel size, lens magnification, and the system modulation transfer function cutoff spatial frequency. According to the optical resolution unit, the printing design boundary pixels are removed from the water-based ink printing area map to obtain the printing main body area map; Within the printed main area map, the quantile judgment threshold of the micro-bending adhesion defect response map is obtained to obtain the adhesion defect response threshold. Pixels located within the water-based ink printing area map and whose micro-bending adhesion defect response map is greater than the adhesion defect response threshold are identified as candidate defect pixels. The candidate region map of the attachment defect is generated from the candidate defect pixels.

[0013] Preferably, by combining the microbend adhesion defect response map and the microbend edge change map, the connectivity of the candidate adhesion defect region map is confirmed to obtain a confirmed adhesion defect region map, including: Connected region marking is performed on the candidate region map of adhesion defects to obtain candidate adhesion defect regions; Remove candidate attachment defect regions whose area is smaller than the projected area of ​​one of the optical resolution units; Extract the region skeleton and region boundary from the deleted candidate attachment defect regions; Within a neighborhood measured by the optical resolution unit, the peak response pixel of the microbend attachment defect response map and the peak edge pixel of the microbend edge variation map are found respectively. The candidate attachment defect region whose region skeleton includes the response peak pixel and whose region boundary neighborhood intersects with the edge peak pixel is determined as the confirmed attachment defect region; The confirmed attachment defect region map is generated from the confirmed attachment defect region, wherein the width of the region boundary neighborhood is equal to the projection width of one of the optical resolution units in the image plane.

[0014] Preferably, outputting an adhesion defect marking map based on the confirmed adhesion defect area map includes: Morphological closure and void filling are performed on the confirmed attachment defect area map to obtain the attachment defect area map; The projection shape of one of the optical resolution units in the planar workstation coordinates is used as a morphological structural element; Extract the contour of the adhesion defect area map; superimpose the contour onto the cross-polarized image in the planar workstation correction image; Maintaining the coordinates of the plane workstation, output the attachment defect marking diagram that indicates the location and boundary of the attachment defect.

[0015] Beneficial effects include: This invention sequentially feeds the water-based ink printed material on environmentally friendly substrates to be inspected into a planar support imaging station and a micro-bending guide roller imaging station, and acquires parallel polarization images, cross-polarization images, and edge reference strip images at the two imaging stations respectively, so that the same printed micro-area forms paired images at the planar station and the micro-bending station; illumination correction is performed on the two types of station images through the edge reference strip, and the micro-bending station image is registered to the planar station coordinates, reducing the impact of illumination fluctuations and positional deviations on defect segmentation during continuous conveying; the processing range is limited by the water-based ink printing area map, and candidate generation and connectivity confirmation are performed by combining the micro-bending adhesion defect response map and the micro-bending edge change map, so that substrate texture, pattern color transition, coating particles, and uneven lighting are less likely to be incorporated into the adhesion defect area; finally, an adhesion defect marking map is output, which can provide image results containing defect location and defect boundary for online quality inspection of water-based ink printed materials on environmentally friendly substrates. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the continuous conveying and detection system for water-based ink printed materials with environmentally friendly substrates according to the present invention. Detailed Implementation

[0017] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0018] The following combination Figure 1 This paper provides a detailed description of a visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates. Figure 1 In the process, the water-based ink printed product 200 on the environmentally friendly substrate to be tested is conveyed along the continuous conveying direction 240. The water-based ink printed product 200 on the environmentally friendly substrate to be tested includes an environmentally friendly substrate 210, a water-based ink printing layer 220, and an edge reference strip 230. The environmentally friendly substrate 210 is one of a biodegradable film, a paper-based composite film, a recycled fiber coated film, and a low surface energy modified substrate treated with corona or plasma. Defects in water-based ink adhesion include interfacial micropores, ink layer micro-lifting, narrow peeling edges, and micro-cracks.

[0019] like Figure 1 As shown, the visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates includes a processor 100, an image acquisition interface 110, an adhesion defect marker output interface 120, a planar support imaging station 300, a micro-bent guide roller imaging station 400, an encoder 500, and a conveyor roller 510. The planar support imaging station 300 and the micro-bent guide roller imaging station 400 are arranged sequentially along the continuous conveying direction 240. The encoder 500 is located at the conveyor roller 510 and provides a position synchronization signal to the processor 100.

[0020] The planar support imaging station 300 includes a planar support 310, a planar station polarized light source 320, a planar station camera 330, and a planar station imaging area 340. The planar support 310 is located on the back of the environmentally friendly substrate 210, so that the area to be photographed within the planar station imaging area 340 is in contact with the planar support surface to form a planar station imaging state. The planar station polarized light source 320 illuminates from one side of the water-based ink printing layer 220, and the planar station camera 330 acquires planar station polarized images from one side of the water-based ink printing layer 220.

[0021] The micro-bending guide roller imaging station 400 includes a micro-bending guide roller 410, a micro-bending station polarization light source 420, a micro-bending station camera 430, and a micro-bending station imaging area 440. The micro-bending guide roller 410 is a cylindrical guide roller that does not contact the printed surface. The micro-bending guide roller 410 contacts the back of the environmentally friendly substrate 210, causing the area to be photographed within the micro-bending station imaging area 440 to form a micro-bending station imaging state. The micro-bending station polarization light source 420 illuminates from one side of the water-based ink printing layer 220, and the micro-bending station camera 430 acquires micro-bending station polarization images from one side of the water-based ink printing layer 220.

[0022] A polarization beam splitter is installed in the imaging optical path between the planar polarization light source 320 and the planar camera 330 to form a parallel polarization channel and a cross-polarization channel. A polarization beam splitter is also installed in the imaging optical path between the micro-bending polarization light source 420 and the micro-bending camera 430 to form a parallel polarization channel and a cross-polarization channel. The polarization direction of the parallel polarization channel is parallel to the polarization direction of the incident light from the light source, while the polarization direction of the cross-polarization channel is perpendicular to the polarization direction of the incident light from the light source. The parallel polarization channel and the cross-polarization channel are formed by simultaneous acquisition through beam splitting by the dual cameras.

[0023] The image acquisition interface 110 receives data output from the planar workstation camera 330, the micro-bending workstation camera 430, and the encoder 500. The processor 100 performs paired processing on the planar station polarization image, the micro-bending station polarization image, and the edge reference strip image of the same printing micro-area based on the position synchronization signal, and sequentially acquires the planar station polarization image, the micro-bending station polarization image, and the edge reference strip image of the water-based ink printed material 200 to be inspected on the environmentally friendly substrate. It uses the edge reference strip image to perform illumination correction on the planar station polarization image and the micro-bending station polarization image respectively, and registers the illumination-corrected micro-bending station polarization image to the planar station coordinates. It generates a water-based ink printing area map from the cross-polarization image in the planar station correction image. Based on the planar station correction image, the micro-bending station registration image, and the water-based ink printing area map, it generates a micro-bending adhesion defect response map and a micro-bending edge change map. Within the water-based ink printing area map, it generates an adhesion defect candidate area map based on the micro-bending adhesion defect response map. Combining the micro-bending adhesion defect response map and the micro-bending edge change map, it confirms the connectivity of the adhesion defect candidate area map, obtaining a confirmed adhesion defect area map. Finally, it outputs an adhesion defect marker map based on the confirmed adhesion defect area map. The attachment defect marking output interface 120 outputs an attachment defect marking diagram that indicates the location and boundary of the attachment defect.

[0024] Example 1: The polarization beam splitter uses a polarization beam splitter prism as its core element. Polarized light emitted from the planar polarization light source 320 illuminates the surface of the water-based ink printed material 200 on the environmentally friendly substrate to be tested. The reflected light enters the polarization beam splitter prism and is split into a transmitted beam and a reflected beam by the beam-splitting film inside the prism. The transmitted beam corresponds to a parallel polarization channel, and the reflected beam corresponds to a cross-polarization channel. The polarization direction of the parallel polarization channel is parallel to the polarization direction of the incident light from the light source, while the polarization direction of the cross-polarization channel is perpendicular to the polarization direction of the incident light from the light source.

[0025] Parallel polarization channel beams and cross-polarization channel beams are respectively incident on the photosensitive surfaces of two identical planar positioning cameras 330. The two identical planar positioning cameras 330 are mounted using a common reference mount, ensuring complete overlap of the two object-side fields of view, and allowing the transmitted and reflected beams formed by the polarizing beam splitter to enter their respective photosensitive surfaces. The intensity splitting ratio of the polarizing beam splitter is... It is used to balance the signal-to-noise ratio of parallel polarization channels and cross polarization channels.

[0026] During polarization direction calibration, a standard polarizer is placed on the object plane within the planar imaging area 340. The standard polarizer is rotated until the average brightness of the parallel polarization channel image reaches its maximum value. At this point, the relative angle between the polarizing beam splitter and the planar polarizing light source 320 is locked. After calibration, the orthogonality error of the polarization directions of the parallel polarization channel and the cross polarization channel does not exceed [a certain value]. The structure, optical path, and calibration method of the micro-bending station polarization beam splitter are consistent with those of the planar support imaging station 300.

[0027] Both the planar polarizing light source 320 and the micro-bending polarizing light source 420 employ linear LED polarizing light sources. Their spectra cover the red, green, and blue characteristic wavelengths of the visible light band. The polarization of both sources is achieved by mounting linear polarizers on the light-emitting surface of the light source, with an extinction ratio of at least [value missing]. .

[0028] The illumination angle of the planar polarized light source 320 is set to be perpendicular to the normal of the environmentally friendly substrate 210. Angle, lighting distance set as The uniformity of illumination within the 340° imaging area of ​​the planar workstation is not less than The illumination angle, illumination distance, and uniformity of the micro-bending station polarized light source 420 are consistent with those of the planar station polarized light source 320. Brightness calibration of the planar station polarized light source 320 and the micro-bending station polarized light source 420 involves placing a standard diffuse reflection whiteboard in the corresponding planar station imaging area 340 or micro-bending station imaging area 440, acquiring parallel polarization channel whiteboard images from the planar support imaging station 300 and the micro-bending guide roller imaging station 400 respectively, and adjusting the light source drive current to ensure that the average brightness difference of the three channels of the whiteboard images from the planar support imaging station 300 and the micro-bending guide roller imaging station 400 is less than [a certain value]. The brightness consistency calibration is then completed. After calibration, the processor 100 records the drive current parameters of each channel.

[0029] The planar support component 310 is made of black matte aluminum oxide material, with a surface treated by anodizing. Its reflectivity in the visible light band is lower than that of aluminum oxide. This is to avoid interference from back-reflected light on polarization imaging results. The flatness of the support surface of the planar support 310 is no higher than... It is used to ensure that the 200 environmentally friendly substrate water-based ink printed on the substrate to be tested is completely flat within the 340 imaging area of ​​the planar workstation.

[0030] Edge reference strip 230 is disposed on one side edge of the environmentally friendly substrate 210 along the continuous conveying direction 240, and its width is set to to The edge reference strip 230 is a blank substrate area without water-based ink printing, does not overlap with the printed pattern area, and is continuously distributed along the continuous conveying direction 240. The reflectivity of the edge reference strip 230 is consistent with the reflectivity of the blank environmentally friendly substrate 210. Before illumination correction, the processor 100 determines the linear correspondence between the pixel brightness of the area where the edge reference strip 230 is located and the actual illumination intensity of the object surface through initialization calibration; the initialization calibration uses a standard illuminance meter to measure the illumination intensity of the object surface point by point and performs linear fitting with the corresponding pixel brightness, with a goodness of fit of not less than .

[0031] The micro-bending guide roller 410 adopts a cylindrical high-precision polished roller body, and the surface roughness of the roller body is no higher than that of the guide roller 410. The range of values ​​for the roller diameter is: to For different types of environmentally friendly substrates, the 210 roller diameter is matched accordingly, with biodegradable films and recycled fiber coated films being matched... Roller diameter, paper-based composite film matching Roller diameter, matched with low surface energy modified substrates treated with corona or plasma. Roller diameter.

[0032] The axis of the slightly curved guide roller 410 is perpendicular to the continuous conveying direction 240, and the installation radial runout is no greater than [value missing]. This is used to ensure that the environmentally friendly substrate 210 within the imaging area 440 of the micro-bending station produces a uniform bending curvature. The defect excitation curvature is determined through pre-experiments, ensuring that the local surface micro-angle change at the four types of water-based ink adhesion defects—interface micro-voids, ink layer micro-lifting, narrow peeling edges, and micro-cracks—is not less than [a certain value]. This ensures that the polarization mirror reflection component produces a stable change that can be captured by the micro-bending station camera 430.

[0033] Example 2: The water-based ink printed material 200 on the environmentally friendly substrate to be tested is transported in roll form along the continuous conveying direction 240 under the traction of the conveying roller 510, and passes sequentially through the planar support imaging station 300 and the micro-bending guide roller imaging station 400. The distance between the two imaging stations, the conveying speed of the conveying roller 510, and the pulse signal of the encoder 500 jointly determine the trigger delay, so that when the same printed micro-area enters the planar station imaging area 340 and the micro-bending station imaging area 440 respectively, paired images are formed.

[0034] In the planar support imaging station 300, the water-based ink printed material 200 on the environmentally friendly substrate to be inspected passes against the planar support surface of the planar support member 310, and the area to be photographed within the planar workstation imaging area 340 remains in a planar workstation imaging state. The planar support member 310 is located on the back of the environmentally friendly substrate 210 and does not contact the water-based ink printing layer 220.

[0035] In the micro-bending guide roller imaging station 400, the water-based ink printed material 200 of the environmentally friendly substrate to be inspected bypasses the micro-bending guide roller 410, with the water-based ink printing layer 220 facing the micro-bending station camera 430. The micro-bending guide roller 410 does not contact the water-based ink printing layer 220, but only contacts the back of the environmentally friendly substrate 210. The axis of the micro-bending guide roller 410 is perpendicular to the continuous conveying direction 240, causing the environmentally friendly substrate 210 within the micro-bending station imaging area 440 to produce a uniform bending curvature.

[0036] The encoder 500 is mounted on the coaxial end of the conveyor roller 510 and rotates synchronously with the conveyor roller 510 to output pulse signals. The pulse equivalent of the encoder 500 is calculated using the diameter of the conveyor roller 510 and the line count of the encoder 500, ensuring that the longitudinal sampling accuracy of the object side is consistent with the lateral pixel accuracy of the camera. The pulse equivalent of the encoder 500 is calculated as follows: ; in, The pulse equivalent of encoder 500. The diameter of the conveyor roller 510 is... This represents the number of pulses per revolution of encoder 500.

[0037] Both the planar station camera 330 and the micro-bending station camera 430 use an external encoder trigger mode for data acquisition. The trigger delay is determined by the distance between the imaging center lines of the planar support imaging station 300 and the micro-bending guide roller imaging station 400, the conveying speed of the water-based ink printed material 200 on the environmentally friendly substrate to be inspected, and the pulse frequency of the encoder 500. The trigger delay time is calculated as follows: ; in, To trigger the delay time, The distance between the imaging center lines of the planar support imaging station 300 and the micro-bent guide roller imaging station 400 is defined as follows: The conveying speed of the 200 samples of water-based ink printed on the environmentally friendly substrate to be tested is given. The trigger delay pulse count is calculated as follows: ; in, To trigger the delay pulse number, This refers to the pulse frequency of encoder 500. Processor 100 calculates and dynamically adjusts the number of trigger delay pulses in real time to compensate for positional deviations caused by fluctuations in conveyor speed. Image matching accuracy is verified by the longitudinal positional deviation of the printed pattern edge, and the matching error is controlled within [specific parameters]. Within a certain number of pixels; if the threshold is exceeded, the processor 100 automatically corrects the number of trigger delay pulses.

[0038] The planar station camera 330 and the micro-bending station camera 430 are triggered by pulse signals from the encoder 500 to acquire images of the same printing micro-area in the planar station imaging area 340 and the micro-bending station imaging area 440. The planar station camera 330 and the micro-bending station camera 430 are either linear array color cameras or area array color cameras. When using a linear array color camera, the same row of images includes a parallel polarized image, a cross-polarized image, and an edge reference strip image. When using an area array color camera, the same image includes a parallel polarized image, a cross-polarized image, and an edge reference strip image.

[0039] The planar workstation polarization images include parallel polarization images and cross-polarization images, while the micro-bending workstation polarization images include parallel polarization images and cross-polarization images. Each type of image contains red, green, and blue channels. The edge reference bar image is derived from the region where edge reference bar 230 is located and is used to characterize the illumination intensity at the current acquisition time.

[0040] Example 3: Processor 100 performs dark-field subtraction for each imaging station, each polarization channel, and each color channel. Dark-field image acquisition is performed during the initialization phase and after daily power-on warm-up. During dark-field image acquisition, the polarization light source for the corresponding imaging station is turned off, and the camera's exposure time and gain parameters are kept consistent with normal production acquisition settings. Each imaging station, each polarization channel, and each color channel acquires images continuously. Frame dark field image, for The final dark field image is obtained by averaging the corresponding pixel positions of the frame dark field image. The final dark field image is calculated as follows: ; in, The pixel brightness values ​​of the final dark-field image. For the first Pixel brightness values ​​of the dark-field image captured in frames. This indicates the planar workstation corresponding to the planar support imaging station 300 or the micro-bending workstation corresponding to the micro-bending guide roller imaging station 400. Indicates the polarization channel type. Indicates color channels, Represents the pixel coordinates of the image.

[0041] Dark field image update cycle set to At the same time, when the ambient temperature changes by more than Automatic re-acquisition is triggered to compensate for deviations caused by dark current drift. Before dark field subtraction, the processor 100 marks and interpolates to repair sensor defects. The defect threshold is the brightness deviation from the neighborhood mean in the dark field image. For pixels exceeding one standard deviation, dead pixel repair is performed using... Neighborhood bilinear interpolation algorithm.

[0042] Processor 100 normalizes the pixel brightness values ​​entering the calculation formula to and take for Processor 100 performs dark field subtraction on the original acquired image to obtain a dark field-subtracted image. The dark field subtraction calculation is as follows: ; in, To subtract pixel brightness values ​​from the image for dark scenes, The pixel brightness values ​​of the original acquired image. This refers to the pixel brightness value of the dark field image corresponding to the imaging station, polarization channel, and color channel.

[0043] Lateral relative illumination distribution is acquired during the initialization phase. During acquisition, a blank environmentally friendly substrate 210 without printed patterns is laid flat in the corresponding imaging area to ensure full lateral coverage of the imaging field of view. The processor 100 selects images at equal intervals along the lateral direction of the image. Each sampling point is evenly distributed across the entire imaging width, and the pixel brightness value at each sampling point is read. A piecewise linear interpolation algorithm is used to interpolate the brightness values ​​of the sampling points to obtain the brightness distribution curve of the full horizontal resolution. Then, the curve is normalized according to the average pixel brightness of the area where the edge reference bar 230 is located to obtain the horizontal relative illumination distribution.

[0044] ; in, For lateral relative lighting distribution, The interpolated lateral brightness distribution. This refers to the horizontal pixel range corresponding to edge reference strip 230. If there is no blank substrate area during production, the horizontal relative illumination distribution is updated using a periodic shutdown calibration method, with an update cycle not exceeding [number missing]. .

[0045] Processor 100 reads the brightness of the edge reference strip image along the continuous conveying direction 240. For each row of the image, processor 100 extracts the pixel brightness value of the area where the edge reference strip 230 is located, calculates the reference illumination intensity of that row, and forms a one-dimensional illumination curve along the continuous conveying direction 240; then, it performs linear interpolation expansion along the horizontal direction of the image according to the horizontal relative illumination distribution obtained in the initialization stage, to obtain a two-dimensional reference illumination map of the same size as the printing area image. For the planar support imaging station 300, a planar station reference illumination map is generated; for the micro-bending guide roller imaging station 400, a micro-bending station reference illumination map is generated. The reference illumination maps are generated separately for the polarization channel and the color channel.

[0046] The area where edge reference bar 230 is located is in the The set of pixels corresponding to a row is denoted as The reference lighting intensity for this line is calculated as follows: ; in, For the first Reference lighting intensity of the line, For the region where edge reference bar 230 is located in the first... The set of pixels in a row. The reference lighting map values ​​are calculated as follows: ; in, These are the reference illumination map values ​​obtained from the edge reference strip image. It is a lateral relative lighting distribution.

[0047] Processor 100 uses the planar station reference lighting map and the micro-bend station reference lighting map to perform illumination correction on the image after dark subtraction, obtaining an illumination-corrected image. The mean value of the reference lighting map after dark subtraction in the region where the edge reference bar 230 is located is calculated as follows: ; in, The set of pixels in the region where edge reference bar 230 is located. This is the mean value in the area where the edge reference bar 230 is located after subtracting the dark field from the reference lighting diagram. The lighting correction calculation is as follows: ; in, To correct the pixel brightness values ​​of an image for illumination, To subtract pixel brightness values ​​from the image for dark scenes, For reference lighting diagram values, This represents the pixel brightness value of the dark-field image. To prevent small positive numbers with a denominator of zero.

[0048] Example 4: The registration marks are circular markers with a diameter of [missing information]. The registration marks are evenly spaced along the horizontal direction. Each is arranged at equal intervals of 240 along the continuous conveying direction. The registration mark is printed in the blank area inside the edge reference strip 230 and does not overlap with the printed pattern.

[0049] The processor 100 calculates the station coordinate transformation relationship based on the printing boundary, registration marks, and pattern texture area. The processor 100 extracts the center coordinates of the registration marks from the planar station correction image and the micro-bending station correction image, and supplements the texture corner points of the printed pattern as auxiliary feature points; the total number of feature points is not less than... The processor 100 uses a normalized cross-correlation algorithm to calculate the matching similarity of feature points, with a similarity threshold set to [value missing]. Mismatched points below the similarity threshold are discarded. If the water-based ink printed material 200 on the environmentally friendly substrate to be tested does not have a dedicated registration mark, the processor 100 selects the fixed corners and line intersections of the printed pattern as alternative feature points to complete the registration.

[0050] Before registration, the processor 100 first performs camera distortion correction on the planar station correction image and the micro-bending station correction image, using... order radial distortion model and A step-tangential distortion model is used to correct lens distortion. The micro-bending station further compensates for perspective distortion caused by bending of the environmentally friendly 210 substrate through thin-plate spline transformation.

[0051] The coordinate mapping employs a thin-plate spline transformation model to establish the mapping relationship from the coordinates of the micro-bending workstation to the coordinates of the planar workstation. The calculation of the thin-plate spline transformation model is as follows: ; ; ; in, For the coordinates of the micro-bending workstation, For planar workstation coordinates, The current pixel coordinates, For the first Coordinates of matching feature points, for The corresponding weighting coefficients, for The corresponding weighting coefficients, For thin plate spline basis functions, The distance between the current pixel coordinates and the coordinates of the matched feature points is the Euclidean distance. Time to take The transformation parameters are solved by least squares fitting, and the fitting reprojection error is controlled within... Within one pixel.

[0052] The processor 100 performs bilinear resampling on the illumination correction image of the micro-bending guide roller imaging station 400 according to the station coordinate transformation relationship to obtain the micro-bending station registration image. The illumination correction image of the planar support imaging station 300 maintains the planar station coordinates and serves as the planar station correction image. The micro-bending station registration image and the planar station correction image have the same size and coordinate system, and pixels with the same coordinates correspond to the same physical position on the environmentally friendly substrate 210.

[0053] Example 5: The processor 100 selects a planar cross-polarized image from the planar station correction image. The cross-polarized channel filters out surface specular reflections while retaining diffuse reflection and absorption information from the ink layer and substrate, which is used to generate the water-based ink printing area map.

[0054] The absorption value of water-based inks is calculated using the natural logarithm base. The reference brightness is the average brightness of the area containing the edge reference strip 230 in the planar cross-polarization channel, representing the incident light intensity benchmark under blank substrate conditions. The processor 100 generates water-based ink absorption maps for each color channel based on the brightness of the planar reference illumination map and the pixel brightness of each color channel in the planar cross-polarization image. The water-based ink absorption value is calculated as follows: ; in, For the first Each color channel is located at coordinates The water-based ink absorption value at that location, For the planar workstation cross-polarization channel Average brightness of the reference bars for each color channel. The first cross-polarized image of the planar workstation Each color channel is located at coordinates The corrected brightness value at the location, This indicates the planar workstation corresponding to the planar support imaging station 300. This indicates a cross-polarization channel.

[0055] For colored substrate scenarios, the processor 100 first collects the cross-polarized three-channel brightness of the blank colored substrate as a background reference. When calculating the absorption value of water-based ink, the background brightness of the colored substrate is used instead of the average brightness of the reference strip to eliminate the absorption deviation caused by the color of the substrate itself.

[0056] Processor 100 selects the channel value with the highest absorption value pixel by pixel from the water-based ink absorption maps of the red, green, and blue channels to form an ink absorption brightness map. The calculation is as follows: ; in, These are the pixel values ​​of the ink absorption brightness map. , , These represent the red, green, and blue channels, respectively. The ink absorption brightness map integrates the absorption information from the three color channels and is used for extracting multi-color printing areas.

[0057] For light-screen ink and gradient printing scenarios, processor 100 first performs an analysis on the ink absorption brightness map. Gaussian smoothing is applied to the window, followed by inter-class variance thresholding to suppress segmentation errors caused by dot noise. Processor 100 performs inter-class variance thresholding on the ink absorption brightness map to obtain the water-based ink printing area map. This water-based ink printing area map is denoted as a binary mask. ,satisfy: ; in, For indicator functions, The segmentation threshold is obtained by maximizing the inter-class variance. Let be the set of pixels in the printed region. .

[0058] After obtaining the initial binary mask through threshold segmentation, processor 100 first executes the process with an area of... Fill tiny holes smaller than one pixel, then remove areas smaller than [number] pixels. Isolated noise regions of 1 pixel were finally selected. Circular structural elements perform boundary smoothing operations. The printed area pixels in the water-based ink printing area map are used to generate a micro-bend adhesion defect response map and an adhesion defect candidate area map.

[0059] Example 6: The processor 100 performs channel difference processing and reflection brightness normalization processing on the parallel polarization image and the cross polarization image in the planar station correction image and the micro-bending station registration image, respectively, to obtain the planar station mirror reflection image and the micro-bending station mirror reflection image. Correcting the polarization channel in a planar workstation image and color channels pixel brightness value, Registering polarization channels in images for micro-bending stations and color channels pixel brightness value, This indicates the micro-bending station corresponding to the micro-bending guide roller imaging station 400.

[0060] The polarization mirror residual is normalized by summing the brightness of the red, green, and blue channels. This is based on the fact that the specular reflection component has consistent polarization characteristics in the three visible light channels, and summing them improves the signal-to-noise ratio and detection stability. The red, green, and blue channels are calculated with equal weights to adapt to the detection scenarios of colored water-based inks. The polarization mirror residual is calculated as follows: ; in, For workstation In coordinates The residual value of the polarizing mirror at that location, For workstation Parallel polarization channel Corrected image brightness values ​​for each color channel. For workstations Cross-polarization channel number Corrected image brightness values ​​for each color channel. Indicates a parallel polarization channel. This indicates a cross-polarization channel.

[0061] Processor 100 calculates the difference between the reflection image of the micro-bending station mirror and the reflection image of the planar station mirror to obtain the micro-bending mirror change image. The calculation is as follows: ; in, This represents the change value of the slightly curved mirror surface. This is the reflection value of the mirror surface at the micro-bending workstation. This is the value of the mirror reflection diagram of the planar workstation. The sign corresponds to the deflection direction of the surface micro-angle at the water-based ink adhesion defect. Interface micro-voids and micro-cracks usually exhibit a mixed distribution of positive and negative values, while ink layer micro-lifting and narrow peeling edges show unipolar changes. Taking the absolute value can uniformly characterize the specular response intensity of the defect and avoid missed detections caused by sign differences.

[0062] Processor 100 performs three-channel averaging on the cross-polarization images in the planar workstation correction image and the micro-bending workstation registration image, respectively, to obtain a single-channel diffuse reflectance brightness map. The calculation is as follows: ; in, For workstations The cross-polarized average brightness. Processor 100 performs processing on the diffuse reflectance brightness map after averaging the three cross-polarized channels. Gaussian smoothing, with a smoothing standard deviation of . This is used to suppress spurious gradients caused by image noise. Processor 100 applies a method to the smoothed diffuse reflectance map. Edge extraction is performed using the Sobel operator for the size, and the lateral gradient components are obtained respectively. and longitudinal gradient components This yields diffuse reflection edge maps for planar workstations and micro-bending workstations, characterized by both the gradient vector and gradient magnitude. The gradient vector is denoted as... The gradient magnitude is calculated as follows: ; in, The gradient vector of the brightness map. This represents the L2 norm. The gradient magnitude is not binarized; continuous values ​​are retained for subsequent calculations of edge change intensity.

[0063] Processor 100 uses the background values ​​of neighboring edges within the water-based ink printing area map to correct the difference between two diffuse edge maps, resulting in a micro-curved edge variation map. Local Neighborhood Window Determined by the optical resolution unit, This represents the set of pixels representing the printed area in the water-based ink printing area map. The micro-curved edge variation map is calculated as follows: ; in, The numerator is the L2 norm of the gradient vector difference between the micro-bending station and the planar station, and the denominator is the median and minimum threshold of the gradient magnitude of the planar station pixels in the local neighborhood of the printing area. The larger value in Pick The denominator uses the median of the gradient magnitude within the local neighborhood of the planar workstation to avoid the normalization reference being too high due to strong edges of the printed pattern itself; when the local neighborhood is entirely a uniform region and the median gradient approaches 0. When the minimum gradient magnitude threshold is used as the denominator, it is used to avoid the value from approaching zero. This leads to computational instability and noise amplification.

[0064] Processor 100 combines the micro-bend mirror surface change map, the micro-bend edge change map, and the neighborhood mirror surface fluctuation value within the water-based ink printing area map to obtain the micro-bend adhesion defect response map. The calculation is as follows: ; in, This represents the response value for micro-bending adhesion defects. This represents the absolute value of the change in the slightly curved mirror surface. Indicates the absolute deviation of the median. This represents the intersection of the local neighborhood and the printed area. The median absolute deviation is calculated as follows: ; in, This represents the set of numerical values ​​used in the statistics. The response map of micro-bending adhesion defects is normalized using the median absolute deviation to robustly characterize the response fluctuation level of the local background. All four types of water-based ink adhesion defects satisfy the response characteristics of simultaneous specular abrupt changes and edge enhancement. Among them, interface micro-voids and ink layer micro-curving are dominated by polarization specular changes and supplemented by edge enhancement, while narrow peeling edges and micro-cracks are dominated by edge enhancement and supplemented by specular changes. The product form allows regions where both types of features coexist to obtain a high response, while the response of single-feature regions is suppressed.

[0065] Example 7: The processor 100 determines the optical resolution unit based on the camera pixel size, lens magnification, and the system modulation transfer function cutoff spatial frequency. Measurements were performed using a standard resolution test chart. During measurement, the standard resolution test chart was placed on the object plane of the corresponding workstation's imaging area. After image acquisition, the system modulation transfer function curve was calculated using the knife-edge method, and the data was read. The corresponding spatial frequency is used as The measurement process is repeated. Take the average value each time.

[0066] Camera pixel size is denoted as Lens magnification is denoted as The size of the optical resolution unit of the object is denoted as The projection width of the optical resolution unit in the image plane is denoted as The optical resolution unit size of the object is calculated as follows: ; in, The size of the optical resolution unit of the object. for The corresponding system modulation transfer function cutoff spatial frequency. The projection width of the optical resolution unit in the image plane is calculated as follows: ; in, The projection width of the optical resolution unit in the image plane. Indicates rounding up. This indicates taking the maximum value. The projection of the optical resolution unit onto the image plane is denoted as... , Centered on the current pixel and with a side length of The projected area of ​​the optical resolution unit for a square pixel set is calculated as follows: ; in, This represents the projected area of ​​the optical resolution unit. Local neighborhood window. In pixels Centered on the optical resolution unit, the calculation is as follows: ; in, In pixels A local neighborhood window centered on the optical resolution unit. All scale parameters based on the optical resolution unit are passed through... The lens magnification is calculated by converting the camera pixel size step by step.

[0067] Processor 100 removes printing design boundary pixels from the water-based ink printing area map using optical resolution units to obtain the main printing area map. Processor 100 performs an erosion operation on the water-based ink printing area map, with the structural elements projected onto the image plane using optical resolution units. This causes the boundaries of the printing area to shrink inward. Remove pixels near the design edges. The set of pixels corresponding to the etched printing area is denoted as... For printing scenarios involving fine lines and small characters, the size of the structural element in the erosion operation should be half the projected size of the original optical resolution unit to avoid excessive erosion of the main printing area, which could cause the pattern to disappear. If the area of ​​the main printing area after erosion is smaller than the original printing area... If so, the processor 100 will skip the etching step and directly use the original water-based ink printed area map for subsequent testing.

[0068] Within the printed main area map, processor 100 calculates the quantile threshold for the micro-bending adhesion defect response map to obtain the adhesion defect response threshold. (Tail quantile parameters) Pick The value is based on the fact that the defect response of normal printed materials follows an approximately normal distribution. Quantiles, used as thresholds, can control the background false detection rate. Within. For production scenarios with varying precision requirements, the tail portion parameter... exist to Within the range, a smaller value is used for high-precision inspection scenarios, and a larger value is used for ordinary quality inspection scenarios. The attachment defect response threshold is calculated as follows: ; in, The threshold for the attachment defect response. Indicates taking Quantile operator for quantiles The quantile parameter is the tail portion parameter. The quantile determination threshold is adaptively determined based on the distribution of micro-bend attachment defect response values ​​in the current image.

[0069] The processor 100 identifies pixels located within the water-based ink printing area map and whose micro-bending adhesion defect response map is greater than the adhesion defect response threshold as candidate defect pixels. The determination is as follows: ; in, These are the candidate defect label values. All candidate defect pixels constitute a binary form of the attachment defect candidate region map.

[0070] Example 8: Processor 100 performs connected component labeling on the candidate attachment defect region map to obtain candidate attachment defect regions. The connected component labeling uses an 8-connected component labeling algorithm, grouping spatially connected candidate defect pixels into the same candidate attachment defect region. These candidate attachment defect regions are denoted as... ,in This is the number of the candidate attachment defect region.

[0071] Processor 100 removes candidate attachment defect regions with an area smaller than the projected area of ​​one optical resolution unit. The area of ​​a candidate attachment defect region is equal to the number of pixels within that region, and the area threshold is... .when At that time, processor 100 deletes candidate attachment defect areas. .

[0072] Processor 100 extracts the region skeleton and region boundary from the deleted candidate attachment defect regions. The region skeleton is obtained using the Zhang-Suen morphological thinning algorithm, which iteratively performs thinning operations until the region pixels no longer change, resulting in a region skeleton with a single pixel width while maintaining region connectivity. The region boundary is the outer contour of the candidate attachment defect region, representing the boundary line between the candidate attachment defect region pixels and the background pixels. Candidate attachment defect region The regional skeleton is denoted as The regional boundary is denoted as .

[0073] Processor 100 generates region boundary neighborhoods at the optical resolution unit scale. The width of the region boundary neighborhood is equal to the projection width of one optical resolution unit in the image plane. The region boundary neighborhood is denoted as... The calculation is as follows: ; in, Candidate adhesion defect region The regional boundary neighborhood.

[0074] The processor 100 searches for peak response pixels in the micro-bend attachment defect response map and peak edge pixels in the micro-bend edge variation map within a neighborhood measured in optical resolution units. For candidate attachment defect regions... The peak response pixel in the candidate attachment defect region The candidate attachment defect region is determined by the maximum response value of the micro-bend attachment defect, with the edge peak pixel located in the candidate attachment defect region. The value is determined by the maximum intensity of change at the micro-bend edge. The peak response pixel and peak edge pixel are calculated as follows: ; ; in, Candidate adhesion defect region peak response pixels, Candidate adhesion defect region Edge peak pixels.

[0075] Processor 100 determines whether the region skeleton contains response peak pixels and whether the region boundary neighborhood intersects with the edge peak pixels. This determination logic is based on the morphological characteristics of water-based ink adhesion defects. The strongest response point of a real water-based ink adhesion defect is usually located on the defect center line, corresponding to the region skeleton position; the strongest edge point is usually located at the defect boundary, corresponding to the region boundary neighborhood. The response peaks of interface micro-voids and ink layer micro-lifting are located in the defect center skeleton, and the edge peaks are located at the defect outer boundary; the response peaks and edge peaks of narrow peeling edges and micro-cracks are distributed along the defect extension direction, with the peak points falling within the region skeleton and the region boundary neighborhood, respectively. The determination is as follows: ; in, Candidate adhesion defect region The confirmation flag value. At that time, candidate attachment defect area The area was identified as having adhesion defects. At that time, candidate attachment defect area The area is uncertain and is considered a confirmed adhesion defect area. If the determination fails and the candidate adhesion defect area is... The area is greater than If the projected area of ​​the optical resolution unit is doubled, then the processor 100 will select candidate attachment defect areas. The system is divided into multiple sub-regions and the confirmation judgment is re-executed to avoid missing large-area water-based ink adhesion defects due to single peak shift.

[0076] Processor 100 generates a confirmed attachment defect region map from all confirmed attachment defect regions. Each connected region in the confirmed attachment defect region map simultaneously exhibits a spatial co-location relationship of polarization specular abrupt change and diffuse reflection edge enhancement.

[0077] Example 9: Processor 100 performs morphological closure and hole filling on the confirmed attachment defect region map to obtain an attachment defect region map. Morphological closure is used to connect adjacent fracture regions and fill gaps, while hole filling is used to fill holes inside the defect region formed by pixel sampling or response unevenness.

[0078] Morphological closure operation is performed first. The second expansion will be executed. Secondary corrosion uses morphological structural elements with side lengths equal to Square structural element The expansion operation is used to connect adjacent fracture defect segments and fill gaps smaller than the size of the structural element; the erosion operation is used to restore the boundary of the expanded region to its original size, preventing the defect region from being excessively enlarged.

[0079] The hole filling uses a flood filling algorithm, filling only holes completely surrounded within the defect area. The maximum area of ​​the holes that can be filled is set to... The projected area of ​​the optical resolution unit is doubled, and internal holes exceeding this size are retained to avoid misfilling the recessed structure of the defect itself.

[0080] Processor 100 extracts the contour of the attached defect area map and overlays the contour onto the cross-polarized image in the planar workstation correction image. Contour extraction uses the Suzuki 85 contour tracking algorithm, extracting only the outer contour of the defect area and excluding the contour of internal holes. A closed curve of pixel width is superimposed on the cross-polarized image in the planar workstation correction image, with the outline color set to a highly saturated red to distinguish it from the background image. The cross-polarized image serves as a base map to display the actual state of the water-based ink printing layer 220 and the morphological details of adhesion defects; the outline is used to indicate the boundary shape and location of each adhesion defect.

[0081] The defect classification step occurs after the generation of the confirmed attachment defect area map and before the output of the attachment defect marker map. For each confirmed attachment defect area, the processor 100 extracts four types of feature parameters for classification, namely the mean value of the micro-bending mirror surface change. Mean value of slight bend edge variation Aspect ratio of the defective area Circularity of the defect area .

[0082] ; ; ; ; in, For the first One confirmed area of ​​adhesion defect. For the first The average change in the micro-bending mirror surface in each confirmed area of ​​adhesion defects. For the first The mean change in the micro-bend edge of each confirmed adhesion defect area. For the first The aspect ratio of the confirmed attachment defect area. For the first The circularity of the defect area in the confirmed adhesion defect area. For the first The longest side of the smallest bounding rectangle of the confirmed attachment defect area. For the first The shorter side of the smallest bounding rectangle of the confirmed attachment defect area. For the first The area of ​​the defective region in each confirmed attachment defect area. For the first The perimeter of the defect area in the confirmed attachment defect area.

[0083] The classification rule is set as follows: when the defect area has a high degree of circularity... Greater than And the average change of the slightly curved edge Below the average change of slightly curved mirror surface When the defect area is circular, it is determined to be a micro-void at the interface; when the defect area is circular... Between to Between and the aspect ratio of the defect area Less than When the ink layer is slightly raised, it is determined to be a defective area with a slight aspect ratio. Greater than And the average change of the slightly curved edge Higher than the average change of the micro-curved mirror surface When the aspect ratio of the defect area is narrow, it is judged as a thin peeling edge; when the aspect ratio of the defect area is... Greater than And the average change of the slightly curved edge Not less than Mean value of micro-bending mirror surface At that time, it was determined to be a microcrack.

[0084] The processor 100 maintains the planar workstation coordinates and outputs an adhesion defect marker image through the adhesion defect marker image output interface 120. The pixel coordinates in the adhesion defect marker image correspond to the physical positions on the environmentally friendly substrate 210, and the adhesion defect marker image indicates the location and boundary of the adhesion defect. The classification result follows the adhesion defect marker. Figure 1 The same output is used to mark the corresponding defect type next to each defect outline, so as to distinguish and detect four types of water-based ink adhesion defects: micro-voids at the interface, micro-lifting of the ink layer, narrow peeling edge, and micro-cracks.

[0085] Example 10: For different types of environmentally friendly substrates 210, the processor 100 has a built-in corresponding parameter set. The parameter set includes five categories of parameters: light source driving current, camera exposure time, polarization angle fine-tuning value, defect detection threshold, and micro-bending guide roller diameter matching mark. When switching the type of environmentally friendly substrate 210, the processor 100 automatically calls the corresponding parameter set to complete the hardware parameter configuration and algorithm threshold adjustment. If the environmentally friendly substrate 210 is a new and unknown substrate, the processor 100 executes a one-click calibration process, sequentially acquiring the planar polarization image and the micro-bending polarization image of the blank substrate, automatically calculating the light source brightness, exposure parameters, and detection threshold, and saving them as a new substrate parameter set.

[0086] The overall calibration of the visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates consists of four steps: polarization calibration, geometric calibration, illuminance calibration, and accuracy verification. Polarization calibration confirms the polarization orthogonality of parallel and cross-polarization channels; geometric calibration confirms the image proportions and registration accuracy; illuminance calibration confirms the accuracy of illumination correction; and accuracy verification uses standard defect samples for inspection, requiring a defect detection rate of no less than [percentage missing]. The false positive rate is no higher than The calibration cycle is once per quarter. A complete calibration is performed again after equipment maintenance or replacement of core components.

[0087] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates, applied to a visual inspection system comprising a planar support imaging station and a micro-bending guide roller imaging station, wherein the two imaging stations are arranged sequentially along a continuous conveying direction, and each imaging station is equipped with a light source, a camera, a parallel polarization channel, and a cross polarization channel, characterized in that, Configured for execution: Collect planar polarization images, micro-bending polarization images, and edge reference strip images of the water-based ink printed on the environmentally friendly substrate to be tested; The planar workstation polarization image and the micro-bending workstation polarization image are respectively illuminated and corrected using the edge reference strip image. The illuminated and corrected micro-bending workstation polarization image is then registered to the planar workstation coordinates to obtain the planar workstation corrected image and the micro-bending workstation registered image. A water-based ink printing area map is generated from the cross-polarization image in the planar workstation correction image; Based on the planar station correction image, the micro-bending station registration image, and the water-based ink printing area map, a micro-bending adhesion defect response map and a micro-bending edge change map are generated. Within the water-based ink printing area map, an adhesion defect candidate area map is generated based on the micro-bending adhesion defect response map; By combining the microbend attachment defect response map and the microbend edge change map, the connectivity of the attachment defect candidate region map is confirmed to obtain the confirmed attachment defect region map. Output an attachment defect marking map based on the confirmed attachment defect area map.

2. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 1, characterized in that, Collect planar polarization images, micro-bending polarization images, and edge reference strip images of the water-based ink printed on the environmentally friendly substrate to be tested, including: The water-based ink printed material on the environmentally friendly substrate to be tested is passed sequentially through the planar support imaging station and the micro-bent guide roller imaging station. In the planar support imaging station, the area to be photographed is brought into contact with the planar support surface to form a planar workstation imaging state; In the micro-bending guide roller imaging station, a micro-bending station imaging state is formed by a cylindrical guide roller that does not contact the printing surface. The encoder triggers a linear or area color camera, enabling the same printed micro-area to form paired images at two imaging stations. Parallel polarized images, cross-polarized images, and edge reference strip images located at the edge of the substrate and not overlapping with the pattern to be detected are acquired from the same row of images or the same image.

3. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 2, characterized in that, The planar workstation polarization image and the micro-bending workstation polarization image are respectively illuminated and corrected using the edge reference strip image. The illuminated and corrected micro-bending workstation polarization image is then registered to the planar workstation coordinates to obtain a planar workstation corrected image and a micro-bending workstation registered image, including: Perform dark field subtraction for each imaging station, each polarization channel, and each color channel; The brightness of the edge reference strip image is read along the continuous conveying direction, and the planar station reference lighting map and the micro-bend station reference lighting map are obtained by interpolation along the horizontal direction of the image. The image after dark field subtraction is illuminated using the planar workstation reference lighting diagram and the micro-bend workstation reference lighting diagram to obtain an illuminated corrected image. Calculate the coordinate transformation relationship of the workstation based on the printing boundary, registration mark, and pattern texture area; The illumination correction image of the micro-bending guide roller imaging station is resampled according to the workstation coordinate transformation relationship to obtain the micro-bending workstation registration image, and the illumination correction image of the planar support imaging station is used as the planar workstation correction image.

4. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 3, characterized in that, Generating a water-based ink printing area map from the cross-polarization image in the planar workstation correction image includes: Select a planar workstation cross-polarization image from the planar workstation correction image; Based on the brightness of the planar workstation reference lighting map and the pixel brightness of each color channel in the planar workstation cross-polarization image, an aqueous ink absorption map for each color channel is formed. In the water-based ink absorption maps of the red, green, and blue channels, the channel value with the largest absorption value is selected pixel by pixel to form an ink absorption brightness map; Perform inter-class variance threshold segmentation on the ink absorption brightness map to obtain the water-based ink printing area map; The printing area pixels in the water-based ink printing area map are used to generate the micro-bending adhesion defect response map and the adhesion defect candidate area map.

5. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 4, characterized in that, Based on the planar station correction image, the micro-bending station registration image, and the water-based ink printing area map, a micro-bending adhesion defect response map and a micro-bending edge change map are generated, including: In the planar workstation correction image and the micro-bending workstation registration image, channel difference processing and reflection brightness normalization processing are performed on the parallel polarization image and the cross polarization image respectively to obtain the planar workstation mirror reflection image and the micro-bending workstation mirror reflection image. The difference between the micro-bending station mirror reflection image and the planar station mirror reflection image is used to obtain the micro-bending mirror change image. The cross-polarization images in the planar station correction image and the micro-bending station registration image are respectively averaged in three color channels and edge extracted to obtain diffuse reflection edge maps of the two imaging stations; The difference between the two diffuse edge maps is corrected by using the background value of the neighboring edge within the water-based ink printing area map to obtain the micro-curved edge variation map; By combining the micro-bend mirror surface change map, the micro-bend edge change map, and the neighborhood mirror surface fluctuation value within the water-based ink printing area map, the micro-bend adhesion defect response map is obtained.

6. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 5, characterized in that, Within the water-based ink printing area map, an adhesion defect candidate area map is generated based on the micro-bending adhesion defect response map, including: The optical resolution unit is determined based on the camera pixel size, lens magnification, and the system modulation transfer function cutoff spatial frequency. According to the optical resolution unit, the printing design boundary pixels are removed from the water-based ink printing area map to obtain the printing main body area map; Within the printed main area map, the quantile judgment threshold of the micro-bending adhesion defect response map is obtained to obtain the adhesion defect response threshold. Pixels located within the water-based ink printing area map and whose micro-bending adhesion defect response map is greater than the adhesion defect response threshold are identified as candidate defect pixels. The candidate region map of the attachment defect is generated from the candidate defect pixels.

7. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 6, characterized in that, By combining the microbend adhesion defect response map and the microbend edge change map, the connectivity of the candidate adhesion defect region map is confirmed to obtain a confirmed adhesion defect region map, including: Connected region marking is performed on the candidate region map of adhesion defects to obtain candidate adhesion defect regions; Remove candidate attachment defect regions whose area is smaller than the projected area of ​​one of the optical resolution units; Extract the region skeleton and region boundary from the deleted candidate attachment defect regions; Within a neighborhood measured by the optical resolution unit, the peak response pixel of the microbend attachment defect response map and the peak edge pixel of the microbend edge variation map are found respectively. The candidate attachment defect region whose region skeleton includes the response peak pixel and whose region boundary neighborhood intersects with the edge peak pixel is determined as the confirmed attachment defect region; The confirmed attachment defect region map is generated from the confirmed attachment defect region, wherein the width of the region boundary neighborhood is equal to the projection width of one of the optical resolution units in the image plane.

8. The visual inspection system for adhesion defects of water-based inks on environmentally friendly substrates according to claim 7, characterized in that, Based on the confirmed attachment defect area map, an attachment defect marking map is output, including: Morphological closure and void filling are performed on the confirmed attachment defect area map to obtain the attachment defect area map; The projection shape of one of the optical resolution units in the planar workstation coordinates is used as a morphological structural element; Extract the contour of the adhesion defect area map; superimpose the contour onto the cross-polarized image in the planar workstation correction image; Maintaining the coordinates of the plane workstation, output the attachment defect marking diagram that indicates the location and boundary of the attachment defect.

Images