Die cutting defect detection method and system

Abstract

The invention relates to the technical field of die cutting machines, and particularly discloses a die cutting defect detection method and system, and the method comprises the steps: obtaining a bright field image and a polarization image of a to-be-detected film; correcting, registering and integrating the bright field image and the polarization image to obtain multi-channel image input information; utilizing a pre-trained defect positioning model to analyze and obtain a defect area according to the multi-channel image input information, and positioning coordinates of the defect area; extracting defect characteristic information in the defect area; associating the defect characteristic information with production parameter information of the to-be-detected film, and storing the defect characteristic information in a defect database; analyzing the defect information in the defect database regularly or in real time to generate an analysis result; according to the method, a complete defect detection, analysis and feedback closed loop is formed through multi-modal imaging, image processing, model positioning, feature extraction and data association and analysis, and the accuracy and robustness of defect detection are improved.

Description

Technical Field

[0001] The present application relates to the technical field of die-cutting machines, and in particular to a die-cutting defect detection method and system. Background Art

[0002] In high-end manufacturing fields such as precision electronic devices, flexible circuit boards, and optical films, high-speed film die-cutting processes use dies to precisely cut and shape multi-layer composite films. Due to the high light transmittance, strong surface reflectivity, and sensitive interlayer stresses of film materials, defects such as edge burrs, delamination warping, and microcracks can easily occur during the production process due to factors such as die wear, tension fluctuations, and temperature changes.

[0003] Traditional inspection methods often use brightfield imaging with a single-light source linear scan camera, identifying obvious damage through threshold segmentation or edge detection algorithms. However, in high-speed, dynamic scenarios, defects can be missed due to interference from material reflections and the halo effect of transmission light. Polarization-sensitive stress defects (such as internal delamination) are particularly difficult to detect. Furthermore, existing technologies lack real-time correlation analysis between defect characteristics and process parameters, resulting in defect tracing relying on manual experience and difficulty in locating process anomalies, hindering production yield improvement and process iteration efficiency.

[0004] There is currently no effective technical solution to the above problems. Summary of the Invention

[0005] The purpose of this application is to provide a die-cutting defect detection method and system to extract and analyze defects through multimodal image data, and to perform data association and analysis in combination with production parameters to provide data support for production process optimization.

[0006] In a first aspect, the present application provides a die-cutting defect detection method for use in a high-speed film die-cutting production line, the method comprising the following steps: S1, obtaining a bright field image and a polarization image of the film to be tested; S2. Correcting, registering, and integrating the bright field image and the polarization image to obtain multi-channel image input information; S3. Analyzing the multi-channel image input information using a pre-trained defect localization model to obtain a defect area and locate the coordinates of the defect area; S4, extracting defect characteristic information in the defect area; S5, associating the defect characteristic information with the production parameter information of the film to be inspected, and storing the information in a defect database; S6. Perform statistical analysis on the defect information in the defect database regularly or in real time, and perform correlation analysis on the defect information and the production parameter information to generate analysis results.

[0007] The die-cutting defect detection method of the present application obtains bright field and polarization multi-channel images, performs correction and registration, and then inputs the defect positioning model, extracts defect features and correlates them with production parameters for analysis, overcomes the influence of the diversity of material optical properties, effectively integrates multimodal imaging information, and deeply correlates defect features with production parameters for analysis, providing data support for production process optimization.

[0008] The die-cutting defect detection method, wherein step S2 comprises: S21, calculating the displacement deviation value and the angle deviation value of the bright field image and the polarization image using a phase correlation method; S22. Performing geometric correction on the bright field image and the polarization image using a bilinear interpolation algorithm according to the displacement deviation value and the angle deviation value to obtain a corrected bright field image and a corrected polarization image; S23. Performing image registration on the corrected bright field image and the corrected polarization image based on a Fourier transform phase correlation method to calculate and obtain a sub-pixel level offset; S24, using the corrected bright field image as a reference image, and resampling the corrected polarization image according to the offset, so that the corrected polarization image and the corrected bright field image are spatially aligned; S25 , performing channel merging on the registered and aligned bright field image and polarization image to construct the multi-channel image input information.

[0009] The above processing method effectively solves the image distortion and spatial position deviation that may exist in the acquisition process of bright field images and polarization images, so that the information of different imaging modes can be accurately integrated, providing reliable input data for subsequent defect location and analysis, and improving the accuracy of defect detection.

[0010] The die-cutting defect detection method, wherein step S23 includes: S231, performing Fourier transform on the corrected bright field image and the corrected polarization image, respectively, to obtain corresponding bright field image spectrum and polarization image spectrum; S232, calculating a cross power spectrum between the bright field image spectrum and the polarization image spectrum, and normalizing the cross power spectrum to obtain a normalized cross power spectrum; S233, performing an inverse Fourier transform on the normalized cross power spectrum to obtain a phase correlation function; S234 , searching for a peak in the phase correlation function, and obtaining a sub-pixel level offset according to a coordinate position corresponding to the peak.

[0011] In the die-cutting defect detection method, the step of normalizing the cross power spectrum to obtain the normalized cross power spectrum includes: S2321. Calculate the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarization image spectrum; S2322. Divide the cross power spectrum by the product of the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarization image spectrum to obtain a normalized cross power spectrum.

[0012] The die-cutting defect detection method, wherein step S3 comprises: S31, inputting the multi-channel image input information into a pre-trained defect localization model, and outputting a predicted probability map and a predicted bounding box of the defect area; S32 . Based on the prediction probability map, a non-maximum suppression algorithm with an adaptive threshold is used to screen overlapping prediction bounding boxes, retain the prediction bounding box with the highest confidence as the final defect area, and extract the coordinates of the defect area.

[0013] The die-cutting defect detection method, wherein the step of screening overlapping prediction bounding boxes using a non-maximum suppression algorithm with an adaptive threshold based on the prediction probability map and retaining the prediction bounding box with the highest confidence as the final defect area includes: S321, determining an initial threshold value according to the predicted probability map and the type of thin film material; S322. Adjusting the initial threshold according to the defect size to obtain an adaptive threshold, wherein the defect size is determined by analyzing the area and aspect ratio of the predicted bounding box; S323, screening the prediction bounding boxes in the prediction probability map whose confidence scores are greater than the adaptive threshold according to the adaptive threshold to obtain candidate defect areas; S324. For the candidate defect area, calculate the overlap between each predicted bounding box and its adjacent predicted bounding boxes. If the overlap is greater than a preset overlap threshold, retain the predicted bounding box with the highest confidence and suppress other overlapping predicted bounding boxes to obtain the final defect area.

[0014] The die-cutting defect detection method, wherein the defect characteristic information includes multiple ones of defect boundary, area, aspect ratio, direction, texture characteristics, edge sharpness, and contrast difference and / or polarization response difference in bright field image and polarization image.

[0015] The die-cutting defect detection method, wherein step S5 comprises: S51. Obtain the material batch, equipment operating status, die-cutting process parameters, and ambient temperature and humidity of the film to be tested from the management system of the high-speed film die-cutting production line, and construct a production parameter information set; S52, extracting the physical position information of the film corresponding to the image acquisition time according to the coordinates of the defect area, and constructing a related data set by combining the defect characteristic information, physical position information and production parameter information set of the defect area; S53: Store the associated data set into the defect database according to a preset data format.

[0016] The die-cutting defect detection method, wherein step S6 comprises: S61. Regularly perform statistical analysis, spatial distribution analysis, time trend analysis on the defect information in the defect database, and analyze the correlation between the defect information and the production parameter information to generate analysis results, and present the analysis results through an interface.

[0017] In a second aspect, the present application further provides a die-cutting defect detection system for use in a high-speed film die-cutting production line, the system comprising: An image acquisition module, used to acquire a bright field image and a polarization image of the film to be inspected; a preprocessing module, configured to correct, register, and integrate the bright field image and the polarization image to obtain multi-channel image input information; A defect localization module is configured to obtain a defect area based on the multi-channel image input information using a pre-trained defect localization model and locate the coordinates of the defect area; A feature extraction module is used to extract defect feature information in the defect area; A storage module, configured to associate the defect characteristic information with production parameter information of the film to be inspected and store the information in a defect database; The analysis module is used to perform statistical analysis on the defect information in the defect database regularly or in real time and to perform correlation analysis on the defect information and the production parameter information to generate analysis results.

[0018] The die-cutting defect detection system of the present application obtains bright field and polarization multi-channel images, performs correction and registration, and then inputs the defect positioning model, extracts defect features and correlates them with production parameters for analysis, overcomes the influence of the diversity of material optical properties, effectively integrates multimodal imaging information, and conducts in-depth correlation analysis between defect features and production parameters to provide data support for production process optimization.

[0019] From the above, it can be seen that the present application provides a die-cutting defect detection method and system, wherein the die-cutting defect detection method of the present application obtains bright field and polarization multi-channel images, performs correction and registration, and then inputs the defect positioning model, extracts defect features and correlates them with production parameters for analysis, overcomes the influence of the diversity of material optical properties, effectively integrates multimodal imaging information, and deeply correlates defect features with production parameters for analysis, providing data support for production process optimization. Through multimodal imaging, image processing, model positioning, feature extraction, data association and analysis, it forms a complete defect detection, analysis and feedback closed loop, improves the accuracy and robustness of defect detection, and supports tracing the root cause of defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 Flowchart of the die-cutting defect detection method provided in an embodiment of the present application.

[0021] Figure 2 A schematic diagram of the structure of the die-cutting defect detection system provided in an embodiment of the present application.

[0022] Reference numerals: 201, image acquisition module; 202, pre-processing module; 203, defect location module; 204, feature extraction module; 205, storage module; 206, analysis module. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0024] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0025] First, please refer to Figure 1 Some embodiments of the present application provide a die-cutting defect detection method for use in a high-speed film die-cutting production line. The method comprises the following steps: S1, obtaining a bright field image and a polarization image of the film to be tested; S2, correcting, registering, and integrating the bright field image and the polarization image to obtain multi-channel image input information; S3. Analyze the multi-channel image input information using a pre-trained defect localization model to obtain the defect area and locate the coordinates of the defect area; S4, extracting defect characteristic information in the defect area; S5, associating the defect characteristic information with the production parameter information of the film to be inspected, and storing the information in a defect database; S6. Perform statistical analysis on the defect information in the defect database regularly or in real time, and perform correlation analysis on the defect information and production parameter information to generate analysis results.

[0026] Specifically, step S1 utilizes at least two sets of high-speed cameras and accompanying light sources to capture images of the high-speed moving film. One set of cameras and light sources captures brightfield images, while the other set, equipped with polarization optics, captures polarized images. The high-speed cameras ensure clear images at production line speeds, while the accompanying light sources provide stable lighting conditions.

[0027] More specifically, step S2 preprocesses the acquired brightfield and polarization images. Correction eliminates geometric distortion introduced by the image acquisition system. Registration ensures precise spatial alignment of the brightfield and polarization images. Integration combines the aligned images from different modalities into a single multi-channel data structure (i.e., multi-channel image input information) for subsequent analysis.

[0028] More specifically, step S3 uses a pre-trained defect localization model to process the multi-channel image input information. The model analyzes the fused image features of the multi-channel image input information, identifies potential defect areas, and outputs the precise coordinates of these areas in the image.

[0029] More specifically, step S4 extracts information that quantitatively describes the nature of the defect from the defect area located in step S3. This information may include various attributes of the defect, such as size, shape, texture, and optical response.

[0030] More specifically, step S5 correlates the extracted defect characteristic information with the production parameter information corresponding to the film under inspection at the time of image acquisition. This production parameter information can be obtained from the production line control system and may include material type, equipment status, process settings, etc. The correlated data is then structured and stored in the defect database.

[0031] More specifically, step S6 analyzes the defect data accumulated in the defect database and the associated production parameters. Statistical analysis can reveal the frequency and distribution of defects. Correlation analysis aims to discover which production parameters are associated with specific defects, thereby identifying the potential causes of the defects. The analysis results are visualized through the user interface.

[0032] The die-cutting defect detection method of the embodiment of the present application simultaneously acquires a brightfield image and a polarized image of the film in step S1. The brightfield image provides general surface information, while the polarized image is sensitive to the material's internal stress, surface microstructure, or contaminants. The combination of the two provides a more comprehensive description of the film's state, helping to distinguish real defects from background textures. The method performs geometric correction on the acquired brightfield and polarized images in step S2 to eliminate distortion, then performs precise spatial registration to ensure pixel-level alignment of the different modal images, and finally integrates the aligned images into multi-channel input data. The method inputs the multi-channel data into a pre-trained defect localization model in step S3. The model analyzes and fuses features, identifies and locates the defect area in the image, and outputs the precise coordinates of the defect. The method extracts various defect characteristic information from the located defect area in steps S4 and S5, and then obtains production parameter information corresponding to the defect location and time, including material batch, equipment operating status, process parameters, etc. The extracted defect characteristic information is associated with the corresponding production parameter information to form structured data and store it in a defect database. In step S6, the method periodically or in real time performs statistical and correlation analysis on the data in the database, analyzing, for example, whether a particular defect is highly correlated with the operating parameters of a particular device or a particular material batch. The analysis results are presented on an interface, providing data support for process adjustments and production optimization, thereby achieving root cause control and prevention of defects.

[0033] From the above, it can be seen that the die-cutting defect detection method of the embodiment of the present application obtains bright field and polarization multi-channel images, performs correction and alignment, and then inputs the defect positioning model, extracts defect features and correlates them with production parameters for analysis, overcomes the influence of the diversity of material optical properties, effectively integrates multimodal imaging information, and deeply correlates defect features with production parameters for analysis, providing data support for production process optimization. It forms a complete defect detection, analysis and feedback closed loop through multimodal imaging, image processing, model positioning, feature extraction, data association and analysis, improves the accuracy and robustness of defect detection, and supports tracing the root cause of defects.

[0034] In some preferred embodiments, step S2 includes: S21, calculating the displacement deviation value and the angle deviation value of the bright field image and the polarization image using a phase correlation method; S22. Performing geometric correction on the bright field image and the polarization image using a bilinear interpolation algorithm according to the displacement deviation value and the angle deviation value to obtain a corrected bright field image and a corrected polarization image; S23. Performing image registration on the corrected bright field image and the corrected polarization image based on a Fourier transform phase correlation method to calculate and obtain a sub-pixel level offset; S24, using the corrected bright field image as a reference image, and resampling the corrected polarization image according to the offset, so that the corrected polarization image and the corrected bright field image are spatially aligned; S25. Merge the channels of the registered and aligned bright field image and polarization image to construct multi-channel image input information.

[0035] Specifically, the displacement deviation value and the angle deviation value obtained in step S21 represent the initial global alignment information between the images. Step S22 corrects the image distortion and achieves preliminary alignment. Steps S23 and S24 obtain the sub-pixel level offset that represents the spatial offset between the images by calculation, and resample and correct the corrected polarization image based on the offset, so that the two images are aligned in spatial position, further improving the alignment accuracy. Step S25 merges the channels of the aligned bright field image and polarization image to construct multi-channel image input information, wherein the corrected bright field image is used as the first channel and the aligned polarization image is used as the second channel, thereby integrating information from different imaging modes.

[0036] More specifically, the above processing method effectively solves the image distortion and spatial position deviation that may exist in the acquisition process of bright field images and polarized images, so that the information of different imaging modes can be accurately integrated, providing reliable input data for subsequent defect location and analysis, and improving the accuracy of defect detection.

[0037] In some preferred embodiments, step S23 includes: S231, performing Fourier transform on the corrected bright field image and the corrected polarization image, respectively, to obtain corresponding bright field image spectrum and polarization image spectrum; S232, calculating the cross power spectrum of the bright field image spectrum and the polarization image spectrum, and normalizing the cross power spectrum to obtain a normalized cross power spectrum; S233, performing an inverse Fourier transform on the normalized cross power spectrum to obtain a phase correlation function; S234 , searching for a peak in the phase correlation function, and obtaining a sub-pixel level offset according to a coordinate position corresponding to the peak.

[0038] Specifically, in practical applications, images may contain noise or contrast variations. Directly calculating the phase correlation function and searching for peaks may be subject to interference, affecting the accuracy and stability of the sub-pixel offset calculation. To address this issue, the above-mentioned processing method uses a Fourier transform-based phase correlation method for image registration. First, the corrected brightfield image and the corrected polarization image are Fourier transformed separately to convert the images from the spatial domain to the frequency domain. In the frequency domain, the cross-power spectrum of the two image spectra is calculated. The cross-power spectrum can highlight the common translation information in the two images and suppress the influence of noise. Furthermore, normalizing the cross-power spectrum can enhance the sharpness of the correlation peak, reduce the impact of image content differences on the correlation, and improve the stability of the registration. Then, the normalized cross-power spectrum is inverse Fourier transformed to obtain the phase correlation function. The phase correlation function produces a peak in the spatial domain, and the position coordinates of this peak correspond to the amount of translation between the two images. Finally, the peak is searched for in the phase correlation function. The position coordinates of the peak represent the pixel-level offset. To obtain a sub-pixel offset, techniques such as interpolation can be used to refine the area near the peak, thereby determining the more precise peak coordinates. This sub-pixel offset is critical information for subsequent image registration and alignment, ensuring precise overlay of brightfield and polarization images, providing accurate multi-channel input for subsequent defect detection.

[0039] More specifically, the cross-power spectrum reflects the phase difference between the two signals in the frequency domain, while the inverse Fourier transform converts this phase information back to the spatial domain. This phase information is reflected in the peak position of the phase correlation function. Therefore, the coordinates of the peak position actually correspond to the sub-pixel offset between the two images in the spatial domain, where the horizontal and vertical coordinates of the peak position represent the magnitude of the image offset in the horizontal and vertical directions, respectively.

[0040] In some preferred embodiments, the step of normalizing the cross power spectrum and obtaining the normalized cross power spectrum includes: S2321. Calculate the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarization image spectrum; S2322. Divide the cross power spectrum by the product of the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarization image spectrum to obtain a normalized cross power spectrum.

[0041] Specifically, step S2321 is used to obtain the amplitude information of the two images in the frequency domain. Then, step S2322 divides the cross power spectrum by the product of the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarized image spectrum. This normalizes the cross power spectrum, effectively removing the influence of amplitude variations on the cross power spectrum and highlighting the phase difference information. The phase difference information directly reflects the amount of image translation.

[0042] More specifically, the normalized cross-power spectrum obtained after this normalization process produces a sharp peak in the phase correlation function obtained by its inverse Fourier transform at the location corresponding to the image offset. This normalization method enhances the signal-to-noise ratio of the phase correlation function peak, making the subsequent peak search and sub-pixel offset extraction more accurate and reliable. This processing method solves the problem that the peak may not be prominent or affected by noise when not using a specific normalization method, thereby improving the accuracy of image registration.

[0043] In some preferred embodiments, step S3 includes: S31, inputting the multi-channel image input information into a pre-trained defect localization model, and outputting a predicted probability map and a predicted bounding box of the defect area; S32. Based on the prediction probability map, a non-maximum suppression algorithm with an adaptive threshold is used to filter overlapping prediction bounding boxes, retain the prediction bounding box with the highest confidence as the final defect area, and extract the coordinates of the defect area.

[0044] Specifically, in step S31, the defect localization model is a deep learning model, preferably including multiple convolutional layers, pooling layers, and fully connected layers. Among them, the convolutional layer is used to extract features at different levels from the input image to capture the local and global patterns of defects. The pooling layer is used to reduce the spatial size of the feature map, reduce computational complexity, and enhance the model's robustness to position changes. The fully connected layer is used to perform classification and regression tasks based on the extracted features. The classification part outputs a predicted probability map, which represents the probability that each pixel or area in the image is a defect. The regression part outputs a predicted bounding box corresponding to the predicted probability map, which frames the possible defect location and range. In this way, the model can identify potential defect areas in the image and provide preliminary location information.

[0045] More specifically, the adaptive threshold is preferably dynamically adjusted based on the type of film material and the size of the defect. For example, a base threshold can be set for a specific material; for defects predicted to be larger, the threshold can be appropriately lowered to avoid mistaking real defects for background noise; for defects predicted to be smaller, the threshold can be appropriately increased to reduce false detections.

[0046] More specifically, the defect localization model may generate multiple overlapping predicted bounding boxes for the same defect, each with different confidence levels (typically related to the probability values ​​in the predicted probability map). Therefore, step S32 uses the non-maximum suppression (NMS) algorithm to process these overlapping bounding boxes. The NMS algorithm uses the predicted probability map and predicted bounding boxes to iteratively select the optimal bounding box. Specifically, the algorithm selects a series of bounding boxes with the highest confidence level based on an adaptive threshold, then suppresses (removes) all other bounding boxes whose overlap exceeds a certain threshold. This process is repeated until all bounding boxes have been processed. The adaptive threshold is dynamically set based on the material type of the film being inspected and the defect size indicated by the predicted bounding box. Different material types (such as polyester, polyimide, and polypropylene) may have different optical properties and background textures, which can affect the model's predicted probability distribution. Defects of different sizes may also require different screening criteria. By adjusting the NMS threshold based on material type and defect size, bounding boxes representing true defects can be more accurately selected, reducing false detections and missed detections. The bounding box that is finally retained is determined as the final defect area, and its corresponding coordinate information is extracted.

[0047] More specifically, the above processing method solves the problem of insufficient positioning accuracy or limited recognition ability of the general defect localization model when facing different materials and defect types, and improves the accuracy and robustness of defect localization.

[0048] In some preferred embodiments, the step of screening overlapping prediction bounding boxes using a non-maximum suppression algorithm with an adaptive threshold according to the prediction probability map and retaining the prediction bounding box with the highest confidence as the final defect area includes: S321, determining an initial threshold value based on the predicted probability map and the type of film material; S322, adjusting the initial threshold according to the defect size to obtain an adaptive threshold, wherein the defect size is determined by analyzing the area and aspect ratio of the predicted bounding box; S323, screening the prediction bounding boxes in the prediction probability map whose confidence scores are greater than the adaptive threshold according to the adaptive threshold to obtain candidate defect areas; S324. For the candidate defect area, calculate the overlap between each predicted bounding box and its adjacent predicted bounding boxes. If the overlap is greater than a preset overlap threshold, retain the predicted bounding box with the highest confidence and suppress other overlapping predicted bounding boxes to obtain the final defect area.

[0049] Specifically, step S321 determines an initial threshold based on the material type of the film to be detected. Different material types, such as polyester film, polyimide film, or polypropylene film, have different optical properties and background noise levels, thus requiring different initial sensitivities. This initial threshold can be pre-stored in a lookup table or obtained using model parameters associated with the material type.

[0050] More specifically, step S322 adjusts the initial threshold based on the defect size obtained from the predicted bounding box analysis to generate an adaptive threshold. The defect size is determined by calculating the area and aspect ratio of the predicted bounding box. Generally, larger defects are more noticeable in the image and require a lower threshold. Therefore, the larger the defect size, the smaller the adaptive threshold. This adjustment can be achieved using a preset function or mapping relationship, which takes the defect size as input and outputs the threshold adjustment.

[0051] More specifically, the adaptive threshold is actually a confidence score threshold. Step S323 uses the generated adaptive threshold to perform a preliminary screening of the predicted probability map. All predicted bounding boxes in the predicted probability map with confidence scores greater than the adaptive threshold are retained to form candidate defect regions. This step effectively eliminates a large number of low-confidence background regions, reducing the computational effort in subsequent processing.

[0052] More specifically, step S324 applies a non-maximum suppression algorithm to the candidate defect region. This algorithm calculates the degree of overlap between any two overlapping predicted bounding boxes within the candidate region, using, for example, the intersection-over-union (IoU) ratio as a metric. If the overlap between two bounding boxes exceeds a preset overlap threshold, their confidence scores are compared, the bounding box with the highest confidence score is retained, and the remaining overlapping bounding boxes are suppressed. By iterating this process, a set of non-overlapping or low-overlapping predicted bounding boxes is ultimately obtained, each corresponding to a separate defect instance, thereby obtaining the final defect region.

[0053] By combining adaptive thresholds and non-maximum suppression for material type and defect size, this scheme can more robustly extract accurate defect areas from the prediction results, effectively solving the problem of difficult screening under different materials and defect sizes, and improving the accuracy of defect location.

[0054] In some preferred embodiments, the defect characteristic information includes defect boundary, area, aspect ratio, orientation, texture characteristics, edge sharpness, and multiple of the contrast difference and polarization response difference in bright field image and polarization image.

[0055] Specifically, the defect boundary can be determined based on a predicted bounding box of the defect area, or can be determined by performing image recognition based on the position of the defect area.

[0056] More specifically, in the above-mentioned defect characteristic information, the defect boundary defines the scope of the defect, the area quantifies the size of the defect, the aspect ratio and direction describe the shape characteristics of the defect, the texture characteristics reflect the grayscale change pattern inside the defect area, the edge sharpness characterizes the clarity of the defect boundary, the contrast difference measures the difference in brightness or grayscale between the defect and the background in the bright field image, and the polarization response difference measures the difference in the defect's ability to change polarized light in the polarized image and the background. Among them, the contrast difference and polarization response difference indicate the difference in sensitivity of different imaging modes to defects.

[0057] More specifically, this multi-dimensional, multi-modal characteristic information provides richer, more discriminative input data for defect correlation and statistical analysis. This improves the effective association between defect information and production parameters, supports accurate identification and tracing of defect causes, and addresses the issue of general defect information being insufficient to fully characterize defects.

[0058] In some preferred embodiments, step S4 includes: S41, extracting a predicted bounding box of the defect area as a defect boundary; S42, fitting the defect boundary using a polygonal approximation algorithm, and calculating the area of ​​the region enclosed by the defect boundary after fitting and the aspect ratio of the minimum circumscribed rectangle; S43. Based on the gray-level co-occurrence matrix method, four texture feature parameters, namely energy, contrast, correlation and entropy, of the defect area in the bright field image and the polarization image are extracted respectively; S44, calculating the gradient amplitude of the edge pixels of the defect area, and calculating the ratio of the number of pixels with a gradient amplitude greater than a preset sharpness threshold to the total number of edge pixels as the sharpness of the defect edge; S45. Calculate the difference between the average grayscale values ​​of the defect area in the bright field image and the polarization image as the contrast difference. Simultaneously, calculate the difference between the maximum polarization angle and the minimum polarization angle of the defect area in the polarization image as the polarization response difference.

[0059] Specifically, in step S41 , the predicted bounding box is output by the defect localization model.

[0060] More specifically, in step S42, the polygonal approximation algorithm can be implemented using the Ramer-Douglas-Peucker algorithm, with an epsilon parameter controlling the approximation accuracy. The area of ​​the fitted region can be calculated by calculating the vertex coordinates of the approximated polygon using the Shoelace formula. The minimum bounding rectangle can be calculated by analyzing the vertex set of the fitted polygon. The defect boundary can be determined using either a predicted bounding box or a fitted defect boundary.

[0061] More specifically, in step S43, the calculation of the gray-level co-occurrence matrix (GLCM) requires setting the number of gray levels, spatial distance, and angle (e.g., 0 degrees, 45 degrees, 90 degrees, 135 degrees); energy, contrast, correlation, and entropy are standard statistics based on the calculated GLCM matrix.

[0062] More specifically, in step S44, the gradient magnitude can be calculated using the Sobel, Prewitt, or Roberts operators; edge pixels can be determined using an edge detection algorithm (e.g., Canny edge detection) or based on a gradient magnitude threshold; and the sharpness ratio is calculated by counting the number of edge pixels whose gradient magnitude exceeds a preset threshold and dividing the result by the total number of edge pixels. These four texture feature parameters can be used directly as texture features.

[0063] More specifically, in step S45, the average grayscale value is obtained by summing the grayscale values ​​of all pixels in the defect area and dividing it by the number of pixels; the calculation of the polarization angle depends on the specific configuration of the polarization imaging system, for example, it is calculated by analyzing the intensity of images in different polarization directions or Stokes parameters; the maximum and minimum polarization angles are obtained by searching in the defect area.

[0064] More specifically, in the above processing method, step S41 uses the predicted bounding box output by the positioning model to quickly obtain the basic geometric information of the defect as the basic area for subsequent analysis. Step S42 uses a polygonal approximation algorithm to fit the boundary more accurately, calculate the fitting area and aspect ratio, thereby obtaining more accurate shape and size information than a simple bounding box, which helps to distinguish defects of different shapes. Step S43 is based on the grayscale co-occurrence matrix method to extract texture features from bright field and polarization images respectively, quantify the spatial relationship of pixels inside the defect area, capture the microstructure and surface characteristics of the defect, especially the texture performance under different imaging modes, which is crucial for distinguishing real defects from background textures or different types of defects. Step S44 quantifies the sharpness of the defect edge by calculating the gradient amplitude of the edge pixels and counting the proportion exceeding the threshold, which helps to evaluate how the defect is formed and provides information about the physical morphology of the defect. Finally, step S45 calculates the average grayscale difference between the defect area in the brightfield and polarized images, as well as the polarization response difference in the polarized image. This directly quantifies the difference in the defect's appearance under the two different imaging modes, fully utilizing multi-channel image information. This is important for identifying defects that are not obvious or easily confused in a single mode. Thus, this processing method obtains a variety of key information, including geometry, texture, edge, and multimodal response. This provides a data foundation for subsequent defect classification, root cause tracing, and production parameter correlation analysis, improving the performance and practicality of the defect detection method.

[0065] In some preferred embodiments, the following steps are further included between step S43 and step S44: S4A. Calculate the difference between the two sets of texture feature parameters corresponding to the defect area as the texture feature of the defect area.

[0066] Specifically, this implementation uses the difference in texture feature parameters as a texture signature. This difference directly quantifies the difference in texture appearance of the defect area under different imaging modes. This differential information helps distinguish true defects from background texture or variations in the optical properties of the material itself. By calculating the difference, the differences between modes can be highlighted, enhancing the differentiation of defect features. Using this difference as a texture signature for the defect area provides more discriminative information for subsequent defect analysis and correlation, helping to improve the accuracy and robustness of defect detection.

[0067] In some preferred embodiments, step S5 includes: S51. Obtain the material batch, equipment operating status, die-cutting process parameters, and ambient temperature and humidity of the film to be tested from the management system of the high-speed film die-cutting production line, and construct a production parameter information set; S52, extracting the physical position information of the film corresponding to the image acquisition time according to the coordinates of the defect area, and constructing a related data set by combining the defect characteristic information, physical position information and production parameter information set of the defect area; S53: Store the associated data set into the defect database according to a preset data format.

[0068] Specifically, step S51 constructs a comprehensive set of production parameter information by acquiring information such as material batches, equipment operating status, die-cutting process parameters, and ambient temperature and humidity from the production line's management system. Step S52 uses the coordinates of the defect area to determine the physical location of the defect on the film, accurately associates the characteristic information of the specific defect with the production parameter information set corresponding to the physical location, and constructs an associated data set containing defects, locations, and production parameters. Step S53 stores the constructed associated data set in the defect database according to a preset data format, providing a structured and traceable data foundation for subsequent statistical analysis and correlation analysis. Through these steps, it is ensured that the defect data not only contains the characteristics of the defect itself, but also contains information on the production environment and process conditions when the defect occurs, so that the causes of the defects can be deeply analyzed in the future, and the production process can be optimized and improved.

[0069] In some preferred embodiments, step S6 includes: S61. Regularly perform statistical analysis, spatial distribution analysis, time trend analysis on the defect information in the defect database, and analyze the correlation between the defect information and production parameter information to generate analysis results, and present the analysis results through the interface.

[0070] Specifically, statistical analysis can provide basic information such as the number and type distribution of defects. Spatial distribution analysis can identify areas where defects cluster on the film surface, helping to locate specific locations or equipment where problems may exist during the production process. Time trend analysis can reveal how the number of defects changes over time, helping to discover periodic fluctuations or abnormal events in the production process. Correlation analysis between defect information and production parameter information can quantify the degree of influence of different production parameters on defects, providing data support for process optimization. These analysis results are integrated into a report and presented through an interface, allowing users to intuitively understand the defect status and identify key influencing factors, thereby guiding production process adjustments and reducing the occurrence of defects.

[0071] More specifically, the above-mentioned processing method regularly processes data in the defect database, performing statistical analysis, spatial distribution analysis, temporal trend analysis, and correlation analysis between defect information and production parameter information. First, basic statistics are performed on the defect data in the database to obtain information such as the total number of defects and the distribution of defects by type. Next, spatial distribution analysis methods, such as kernel density estimation, are used to analyze the location distribution of defects on the film surface, generating a heat map. The color depth of the heat map represents the density of defects, thereby identifying areas of defect concentration. Simultaneously, the number of defects is statistically analyzed by production batch or time period, and a trend chart of the defect number over time is plotted. The trend chart can be smoothed using a sliding average method to identify time points when the defect number fluctuates abnormally. Furthermore, correlation analysis methods, such as Spearman rank correlation analysis, are used to calculate the correlation coefficient between defect characteristics (such as area, aspect ratio, texture characteristics, etc.) and production parameters (such as die-cutting speed, pressure, temperature, etc.). Production parameters with absolute values ​​of the correlation coefficient exceeding a preset threshold are screened out to analyze their impact on defect generation. Finally, the statistical analysis results, spatial distribution heat map, time trend diagram, correlation coefficient and impact degree analysis results are integrated to generate an analysis report, which is presented in the form of charts and text through the user interface, thereby providing data support for production process monitoring, problem diagnosis and process improvement.

[0072] In some preferred embodiments, step S61 includes: S611. Determine the defect data range to be analyzed based on the defect database according to the time period or real-time analysis instructions; S612. Based on the defect data range, a kernel density estimation method is used to analyze the spatial distribution of defects on the film surface and generate a heat map. The color depth of the heat map represents the defect density. The defect cluster area is identified based on the heat map. S613. Count the number of defects according to the defect data range, production batch or time period, draw a trend chart of the number of defects over time, smooth the trend chart using a sliding average method, and identify the time points when the number of defects fluctuates abnormally; S614. Based on the defect data range, use the Spearman rank correlation analysis method to calculate the correlation coefficient between the defect characteristics and the production parameters. Screen out the production parameters whose absolute value of the correlation coefficient is greater than the preset threshold and analyze their impact on the defect generation. S615. Integrate the heat map, trend map, correlation coefficient, and impact analysis results to generate an analysis report, and present it in the form of charts and text through the user interface.

[0073] Specifically, the above solution refines the execution process of step S61 and provides an analysis method and result presentation format: First, the defect data range to be analyzed is determined based on a time period or real-time instructions, limiting the data set to be analyzed. Next, kernel density estimation is used to process the defect location data on the film surface to generate a heat map. The heat map displays the spatial density of defects through color changes, identifying areas where defects are most likely to occur, which helps locate potentially problematic production equipment or process steps. Simultaneously, the number of defects is counted by time (production batch or time period), and a trend chart is plotted. The trend chart displays the change in the number of defects over time. The trend is smoothed using a sliding average method, allowing observation of the changing trend of the defect count and identification of time points of defect fluctuation. These time points may correspond to process parameter adjustments, raw material changes, or changes in equipment status. Furthermore, Spearman rank correlation analysis is used to calculate the correlation between defect characteristics (such as size, shape, texture, etc.) and production parameters (such as speed, temperature, and pressure). Spearman rank correlation analysis is a non-parametric method suitable for analyzing nonlinear relationships or non-normally distributed data. By calculating correlation coefficients and setting thresholds for screening, production parameters with strong correlations with defect characteristics can be identified and their impact quantified, providing data support for tracing the causes of defects. Finally, the spatial distribution heat map, temporal trend graph, and correlation analysis results between defect characteristics and production parameters are integrated to generate an analysis report. The report is presented in graphical and text form on the user interface, allowing operators or engineers to understand the occurrence patterns of defects and their relationship to the production process, thereby guiding the improvement and optimization of production processes and reducing defects.

[0074] Second, please refer to Figure 2 Some embodiments of the present application further provide a die-cutting defect detection system for use in a high-speed film die-cutting production line. The system comprises: An image acquisition module 201 is used to acquire a bright field image and a polarization image of the film to be inspected; A pre-processing module 202 is used to correct, register and integrate the bright field image and the polarization image to obtain multi-channel image input information; The defect localization module 203 is used to obtain the defect area based on the multi-channel image input information analysis using a pre-trained defect localization model and locate the coordinates of the defect area; The feature extraction module 204 is used to extract defect feature information in the defect area; The storage module 205 is used to associate the defect characteristic information with the production parameter information of the film to be inspected and store it in the defect database; The analysis module 206 is used to perform statistical analysis on the defect information in the defect database regularly or in real time and to perform correlation analysis on the defect information and production parameter information to generate analysis results.

[0075] The die-cutting defect detection system of the embodiment of the present application obtains bright field and polarization multi-channel images, performs correction and registration, and then inputs the defect positioning model, extracts defect features and analyzes the correlation with production parameters, overcomes the influence of the diversity of material optical properties, effectively integrates multimodal imaging information, and deeply correlates the defect features with production parameters to provide data support for production process optimization.

[0076] In addition, the units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0077] Furthermore, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0078] In this document, relational terms such as first and second, etc. are used merely to distinguish one entity or operation from another entity or operation, but do not necessarily require or imply any actual relationship or order between these entities or operations.

[0079] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A die-cutting defect detection method, applied to a high-speed film die-cutting production line, characterized in that: The method comprises the following steps: S1, obtaining a bright field image and a polarization image of the film to be tested; S2. Correcting, registering, and integrating the bright field image and the polarization image to obtain multi-channel image input information; S3. Analyzing the multi-channel image input information using a pre-trained defect localization model to obtain a defect area and locate the coordinates of the defect area; S4, extracting defect characteristic information in the defect area; S5, associating the defect characteristic information with the production parameter information of the film to be inspected, and storing the information in a defect database; S6. Perform statistical analysis on the defect information in the defect database regularly or in real time, and perform correlation analysis on the defect information and the production parameter information to generate analysis results.

2. The die-cutting defect detection method according to claim 1, characterized in that: Step S2 includes: S21, calculating the displacement deviation value and the angle deviation value of the bright field image and the polarization image using a phase correlation method; S22. Performing geometric correction on the bright field image and the polarization image using a bilinear interpolation algorithm according to the displacement deviation value and the angle deviation value to obtain a corrected bright field image and a corrected polarization image; S23. Performing image registration on the corrected bright field image and the corrected polarization image based on a Fourier transform phase correlation method to calculate and obtain a sub-pixel level offset; S24, using the corrected bright field image as a reference image, and resampling the corrected polarization image according to the offset, so that the corrected polarization image and the corrected bright field image are spatially aligned; S25 , performing channel merging on the registered and aligned bright field image and polarization image to construct the multi-channel image input information.

3. The die-cutting defect detection method according to claim 2, characterized in that: Step S23 includes: S231, performing Fourier transform on the corrected bright field image and the corrected polarization image, respectively, to obtain corresponding bright field image spectrum and polarization image spectrum; S232, calculating a cross power spectrum between the bright field image spectrum and the polarization image spectrum, and normalizing the cross power spectrum to obtain a normalized cross power spectrum; S233, performing an inverse Fourier transform on the normalized cross power spectrum to obtain a phase correlation function; S234 , searching for a peak in the phase correlation function, and obtaining a sub-pixel level offset according to a coordinate position corresponding to the peak.

4. The die-cutting defect detection method according to claim 3, characterized in that: The step of normalizing the cross power spectrum to obtain a normalized cross power spectrum includes: S2321. Calculate the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarization image spectrum; S2322. Divide the cross power spectrum by the product of the amplitude spectrum of the bright field image spectrum and the amplitude spectrum of the polarization image spectrum to obtain a normalized cross power spectrum.

5. The die-cutting defect detection method according to claim 1, characterized in that: Step S3 includes: S31, inputting the multi-channel image input information into a pre-trained defect localization model, and outputting a predicted probability map and a predicted bounding box of the defect area; S32 . Based on the prediction probability map, a non-maximum suppression algorithm with an adaptive threshold is used to screen overlapping prediction bounding boxes, retain the prediction bounding box with the highest confidence as the final defect area, and extract the coordinates of the defect area.

6. The die-cutting defect detection method according to claim 5, characterized in that: The step of screening overlapping prediction bounding boxes using a non-maximum suppression algorithm with an adaptive threshold according to the prediction probability map and retaining the prediction bounding box with the highest confidence as the final defect area includes: S321, determining an initial threshold value according to the predicted probability map and the type of thin film material; S322. Adjusting the initial threshold according to the defect size to obtain an adaptive threshold, wherein the defect size is determined by analyzing the area and aspect ratio of the predicted bounding box; S323, screening the prediction bounding boxes in the prediction probability map whose confidence scores are greater than the adaptive threshold according to the adaptive threshold to obtain candidate defect areas; S324. For the candidate defect area, calculate the overlap between each predicted bounding box and its adjacent predicted bounding boxes. If the overlap is greater than a preset overlap threshold, retain the predicted bounding box with the highest confidence and suppress other overlapping predicted bounding boxes to obtain the final defect area.

7. The die-cutting defect detection method according to claim 1, characterized in that: The defect characteristic information includes multiple ones of defect boundary, area, aspect ratio, direction, texture characteristics, edge sharpness, and contrast difference and / or polarization response difference in bright field image and polarization image.

8. The die-cutting defect detection method according to claim 1, characterized in that: Step S5 includes: S51. Obtain the material batch, equipment operating status, die-cutting process parameters, and ambient temperature and humidity of the film to be tested from the management system of the high-speed film die-cutting production line, and construct a production parameter information set; S52, extracting the physical position information of the film corresponding to the image acquisition time according to the coordinates of the defect area, and constructing a related data set by combining the defect characteristic information, physical position information and production parameter information set of the defect area; S53: Store the associated data set into the defect database according to a preset data format.

9. The die-cutting defect detection method according to claim 1, characterized in that: Step S6 includes: S61. Regularly perform statistical analysis, spatial distribution analysis, time trend analysis on the defect information in the defect database, and analyze the correlation between the defect information and the production parameter information to generate analysis results, and present the analysis results through an interface.

10. A die-cutting defect detection system, used in a high-speed film die-cutting production line, characterized in that: The system comprises: An image acquisition module, used to acquire a bright field image and a polarization image of the film to be inspected; a preprocessing module, configured to correct, register, and integrate the bright field image and the polarization image to obtain multi-channel image input information; A defect localization module is configured to utilize a pre-trained defect localization model to analyze and obtain a defect area based on the multi-channel image input information, and locate the coordinates of the defect area; A feature extraction module is used to extract defect feature information in the defect area; A storage module, configured to associate the defect characteristic information with production parameter information of the film to be inspected and store the information in a defect database; The analysis module is used to perform statistical analysis on the defect information in the defect database regularly or in real time and to perform correlation analysis on the defect information and the production parameter information to generate analysis results.

Images