Electrostatic Dust Identification System and Method for PE Film Surface

Abstract

This invention relates to the field of surface defect detection technology, and in particular to a system and method for identifying electrostatic dust on the surface of PE film. The system includes: an image acquisition module that acquires multi-angle, multi-wavelength images of the PE film surface using a ring LED array and a multi-wavelength illumination system; a multi-manifold mapping background suppression module that treats the PE film surface as a composite manifold, separating background and dust features; a differential geometric feature extraction module that extracts geometric features such as curvature and shape index, calculates the geodesic distance matrix, and generates a noise-dust probability map; and a topology feature segmentation module that performs edge detection, morphological processing, and topology-preserving clustering to classify image regions and determine dust locations. This method, combining multi-manifold mapping and differential geometric features, effectively overcomes the detection difficulties caused by the highly reflective surface of PE film and can accurately identify dust particles at the 0.5-10 μm level.

Description

Technical Field

[0001] This invention relates to the field of surface defect detection technology, and in particular to a system and method for identifying electrostatic dust particles on the surface of PE films, used to accurately detect tiny electrostatically adsorbed dust particles on the surface of PE films. Background Technology

[0002] As an important packaging material and protective layer for optoelectronic components, the surface quality of PE film directly affects the performance and appearance of the product. During the production and application of PE film, due to electrostatic effects, tiny dust particles easily adhere to the surface of the PE film, forming contaminants that are difficult to remove.

[0003] Existing PE film surface inspection technologies mainly include manual visual inspection, ordinary image processing, and traditional machine vision methods. Manual visual inspection is inefficient and prone to fatigue, ordinary image processing methods have limited ability to detect tiny dust particles (especially at the 0.5-10μm level), and traditional machine vision methods often have difficulty effectively distinguishing between dust particles and background noise when dealing with highly reflective surfaces such as PE films.

[0004] Furthermore, existing technologies typically employ simple threshold segmentation or basic morphological processing methods, which struggle to handle the complex light reflection characteristics of PE film surfaces. This is especially true for PE films with varying illumination angles and surface materials, where detection accuracy significantly decreases. Therefore, developing a system and method capable of accurately identifying minute electrostatic dust particles on PE film surfaces has significant practical value. Summary of the Invention

[0005] The purpose of this invention is to provide an electrostatic dust identification system and method for PE film surfaces, which can overcome the problems existing in the prior art and achieve accurate identification of dust particles at the 0.5-10μm level on the surface of PE films.

[0006] This invention proposes a PE film surface electrostatic dust identification system, comprising:

[0007] The image acquisition module is used to acquire multi-angle, multi-wavelength images of the PE film surface;

[0008] A multi-manifold mapping background suppression module, connected to the image acquisition module, is used to receive the multi-angle, multi-wavelength images, construct a manifold representation of the PE film surface, separate the PE film background and potential micro-dust features, and generate a background suppression image.

[0009] The differential geometric feature extraction module is connected to the multi-manifold mapping background suppression module and is used to receive the background suppression image, extract differential geometric features, calculate the geodesic distance matrix, and generate a noise-dust probability map.

[0010] The topological feature segmentation module, connected to the differential geometric feature extraction module, is used to receive the noise-dust probability map, perform edge detection and morphological processing, perform topological preservation clustering, divide the image region into speckle noise, blurred noise, dust and background, and determine the location and feature information of dust on the PE film surface.

[0011] Preferably, the image acquisition module includes:

[0012] A ring-shaped LED array is used to illuminate the surface of the PE film from multiple angles;

[0013] The multi-wavelength illumination system includes three sets of dark field illumination modules with different wavelengths, which are used to sequentially activate and generate illumination light of different wavelengths;

[0014] Multiple photoelectric imaging detectors are connected to the multi-wavelength illumination system to acquire images of the PE film surface from different angles.

[0015] Preferably, the multi-manifold mapping background suppression module includes:

[0016] Local structure preservation units are used to construct the local geometry of an image and generate local geometric feature maps.

[0017] A global structure mapping unit, connected to the local structure holding unit, is used to receive the local geometric feature map, construct a global similarity map, optimize the manifold embedding objective function, and generate a low-dimensional representation of the PE film and dust.

[0018] An adaptive threshold adjustment unit, connected to the global structure mapping unit, is used to receive the low-dimensional representation, calculate the manifold entropy, determine the optimal segmentation threshold, and generate a background suppression image.

[0019] Preferably, the differential geometric feature extraction module includes:

[0020] A differential geometric feature extractor is used to convert an image into a height function, calculate the Gaussian curvature, mean curvature, and shape index, and generate a differential geometric feature map.

[0021] The geodesic distance calculation unit is connected to the differential geometric feature extractor and is used to receive the differential geometric feature map, define a feature-based Riemannian metric tensor, calculate the geodesic distance using the fast traversal method, and construct a distance matrix.

[0022] The manifold probability distribution estimator, connected to the geodesic distance calculation unit, is used to receive the distance matrix and the differential geometric feature map, estimate the probability density of the feature space, calculate the probability that each point belongs to noise or dust, and generate a noise-dust probability map.

[0023] Preferably, the topological feature segmentation module includes:

[0024] An edge detector is used to compute the Laplacian Gaussian operator, detect zero-crossing points, and extract edge feature maps.

[0025] A morphological processing unit, connected to the edge detector, is used to receive the edge feature map, perform multi-scale morphological dilation and erosion operations, and generate a morphologically enhanced feature map.

[0026] A topology-preserving clusterer, connected to the morphological processing unit, is used to receive the morphological enhancement feature map, construct a high-dimensional feature vector, perform topology-preserving clustering, and cluster pixels into speckle noise, blurred noise, dust, and background.

[0027] The feature space fusion unit, connected to the topology-preserving clusterer, is used to receive multi-level features and clustering results, calculate feature weights, perform nonlinear feature fusion, and generate the final dust location and feature information.

[0028] Preferably, the local structure preservation unit constructs the local geometric structure by: determining the set of k nearest neighbors for each pixel, calculating the similarity between the pixel and its nearest neighbors, constructing a local similarity matrix, and extracting feature vectors to represent the local geometric structure.

[0029] Preferably, the differential geometric features extracted by the differential geometric feature extractor include: principal curvature, Gaussian curvature, mean curvature, shape index, and rate of change of curvature, and generate multi-scale curvature representations in multiple scale spaces.

[0030] Preferably, the topology-preserving clusterer performs topology-preserving clustering by: constructing a high-dimensional feature vector containing spatial, morphological, and statistical features; defining a topology-preserving objective function; iteratively optimizing through gradient descent to preserve the local neighborhood structure; automatically determining the optimal number of categories; and dividing the pixels into four categories.

[0031] Preferably, the feature space fusion processor performs nonlinear feature fusion by: calculating the reliability of each feature layer, assigning adaptive weights, designing a nonlinear combination function, integrating background suppression layer features, noise detection layer features, and morphological processing features to generate the final dust representation.

[0032] A method for identifying electrostatic dust particles on the surface of a PE film includes the following steps:

[0033] Acquire multi-angle, multi-wavelength images of the PE film surface;

[0034] A manifold representation of the PE film surface is constructed to separate the PE film background and potential micro-dust features, generating a background-suppressed image.

[0035] Extract the differential geometric features of the background-suppressed image, calculate the geodesic distance matrix, and generate a noise-dust probability map;

[0036] Edge detection and morphological processing are performed on the noise-dust probability map, and topology-preserving clustering is performed to divide the image region into speckle noise, blurred noise, dust and background.

[0037] By integrating multi-level features, the location and characteristic information of micro-dust on the PE film surface can be determined.

[0038] The beneficial effects of this invention include:

[0039] 1. Improved accuracy of micro-dust detection: Through multi-manifold mapping and differential geometric feature extraction technology, it can detect micro-dust at the 0.5μm level that is difficult to detect by traditional methods, with a detection rate of 85% in the 0.5-2μm range and 98% in the 2-10μm range.

[0040] 2. Enhanced anti-interference capability: The topological feature-based differentiation mechanism effectively distinguishes between dust and other interfering factors on the PE film surface (such as reflection, scratches, etc.), reducing the false alarm rate to below 3%.

[0041] 3. Improved system adaptability: The adaptive parameter adjustment mechanism enables the system to automatically adapt to PE films with different materials and surface properties, significantly improving its versatility.

[0042] 4. Improved production efficiency: The system can process 1080p resolution images at a speed of 15 frames per second, which greatly improves the detection efficiency and reduces the manual inspection process.

[0043] 5. Reduced production costs: High-precision dust detection reduces the defect rate, decreases losses in subsequent processing stages, and improves overall economic efficiency. Attached Figure Description

[0044] Figure 1 This is a schematic diagram of the overall structure of the electrostatic dust identification system on the PE film surface of the present invention;

[0045] Figure 2 This is a schematic diagram of the multi-manifold mapping background suppression module of the present invention;

[0046] Figure 3 This is a schematic diagram of the differential geometric feature extraction module of the present invention;

[0047] Figure 4 This is a schematic diagram of the topology feature segmentation module of the present invention;

[0048] Figure 5 This is a flowchart of the electrostatic dust identification method for the PE film surface of the present invention. Detailed Implementation

[0049] Please refer to Figures 1-5 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. These embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0050] like Figure 1 As shown, the PE film surface electrostatic dust identification system provided by the present invention includes an image acquisition module 1, a multi-manifold mapping background suppression module 2, a differential geometric feature extraction module 3, and a topological feature segmentation module 4.

[0051] Image acquisition module 1 is used to acquire multi-angle, multi-wavelength images of the PE film surface. Multi-manifold mapping background suppression module 2, connected to image acquisition module 1, receives multi-angle, multi-wavelength images, constructs a manifold representation of the PE film surface, separates the PE film background and potential dust features, and generates a background-suppressed image. Differential geometric feature extraction module 3, connected to multi-manifold mapping background suppression module 2, receives the background-suppressed image, extracts differential geometric features, calculates the geodesic distance matrix, and generates a noise-dust probability map. Topological feature segmentation module 4, connected to differential geometric feature extraction module 3, receives the noise-dust probability map, performs edge detection and morphological processing, performs topology-preserving clustering, and divides the image region into speckle noise, blurred noise, dust, and background, determining the location and feature information of dust on the PE film surface.

[0052] like Figure 1 As shown, in a preferred embodiment of the present invention, the image acquisition module 1 includes a ring LED array 11, a multi-wavelength illumination system 12, and multiple photoelectric imaging detectors 13.

[0053] The annular LED array 11 is used to illuminate the PE film surface from multiple angles. Preferably, the annular LED array 11 includes 16 LED light sources arranged uniformly in a ring, which can illuminate the PE film surface from different angles to produce a multi-angle illumination effect. The wavelength range of the LED light sources is preferably 400nm-700nm, covering the visible spectrum range to obtain rich spectral information.

[0054] The multi-wavelength illumination system 12 includes three sets of dark-field illumination modules with different wavelengths, which are sequentially activated to generate illumination light of different wavelengths. Preferably, the center wavelengths of the three sets of dark-field illumination modules are 450nm (blue light), 550nm (green light), and 650nm (red light), respectively, with a wavelength interval of 100nm, to obtain the reflection characteristics of the PE film surface at different wavelengths. Each set of dark-field illumination modules works in conjunction with the ring LED array 11 to ensure multi-angle, multi-wavelength illumination effects.

[0055] Multiple photoelectric imaging detectors 13 are connected to a multi-wavelength illumination system 12 to acquire images of the PE film surface from different angles. Preferably, the system is configured with three high-resolution cameras (at least 1080p resolution) evenly distributed at a 120° angle to obtain omnidirectional images of the PE film surface. The camera exposure time is adjustable from 1ms to 50ms to adapt to different lighting conditions.

[0056] In actual operation, the ring-shaped LED array 11 first illuminates the PE film surface from multiple angles. Then, the multi-wavelength illumination system 12 sequentially activates three sets of dark-field illumination modules with different wavelengths. Finally, multiple photoelectric imaging detectors 13 simultaneously acquire images of the PE film surface. This multi-angle, multi-wavelength acquisition method can obtain richer image information of the PE film surface, providing sufficient data support for subsequent processing.

[0057] like Figure 2 As shown, in a preferred embodiment of the present invention, the multi-manifold mapping background suppression module 2 includes a local structure preservation unit 21, a global structure mapping unit 22, and an adaptive threshold adjustment unit 23.

[0058] The local structure preservation unit 21 is used to construct the local geometric structure of the image and generate a local geometric feature map. Specifically, the local structure preservation unit 21 constructs the local geometric structure in the following way: determining the set of k nearest neighbors for each pixel, calculating the similarity between the pixel and its nearest neighbors, constructing a local similarity matrix, and extracting feature vectors to represent the local geometric structure.

[0059] In practical implementation, for each pixel p in the image, its k-nearest neighbor set is first determined. The optimal value for k is 8-12 to balance the integrity of local information and computational complexity. Then, the similarity between point p and its nearest neighbors is calculated using a Gaussian kernel function.

[0060] ,

[0061] in: This represents the similarity between pixels i and j; and These represent the feature vectors (including position coordinates and grayscale values) of pixels i and j, respectively. This is the bandwidth parameter of the Gaussian kernel, expressed in pixels, with an optimal value of 0.1-0.5 times the standard deviation of the image grayscale. Representing the eigenvector and The Euclidean distance between them.

[0062] In practical applications of electrostatic dust identification on PE film surfaces, when detecting smaller dust particles (0.5-2μm), The value can be set to a smaller value (e.g., 0.2 times the standard deviation) to enhance local details; when detecting larger dust particles (5-10 μm), The value can be set to a larger value (such as 0.4 times the standard deviation) to obtain a more stable representation of the feature.

[0063] Based on the similarity matrix W, a feature representation of the local geometric structure is constructed:

[0064] ,

[0065] Where: L is the Laplacian matrix, with dimensions n×n, where n is the total number of pixels in the image; D is a diagonal matrix, with dimensions n×n; diagonal elements This represents the sum of similarities between the i-th pixel and all its nearest neighbors.

[0066] By solving the generalized eigenvalue problem:

[0067] ,

[0068] Where: f is the feature vector with dimension n×1; Let be the eigenvalues. The eigenvector corresponding to the smallest non-zero eigenvalue is taken as the representation of the local geometric structure.

[0069] The global structure mapping unit 22 is connected to the local structure preservation unit 21 to receive local geometric feature maps, construct a global similarity map, optimize the manifold embedding objective function, and generate low-dimensional representations of PE film and dust.

[0070] In practical implementation, the global structure mapping unit 22 first constructs a global adjacency graph G based on local features, and then defines an optimization objective function:

[0071] ,

[0072] in: and These represent the representations of pixels i and j in a low-dimensional space, with a dimension of d×1 (d is usually 2 or 3). These are the elements of the similarity matrix calculated earlier; This represents the Euclidean distance between points i and j in the low-dimensional space.

[0073] This optimization problem can be transformed into solving:

[0074] ,

[0075] Where: Y is the low-dimensional representation matrix of all points, with dimension n x d; tr represents the trace operation of the matrix, that is, the sum of the elements on the main diagonal; This represents the transpose of matrix Y.

[0076] In the application of micro-dust identification on PE film surfaces, the optimization objective is to keep similar points (such as those belonging to the PE film background or those belonging to the same micro-dust) close together in a low-dimensional space, while separating points of different categories (such as PE film background points and micro-dust points). By solving this optimization problem, a low-dimensional representation that preserves the topological relationships can be obtained, thereby separating the PE film background and micro-dust in a low-dimensional space.

[0077] The adaptive threshold adjustment unit 23 is connected to the global structure mapping unit 22 and is used to receive the low-dimensional representation, calculate the manifold entropy, determine the optimal segmentation threshold, and generate a background suppression image.

[0078] In practical implementation, the adaptive threshold adjustment unit 23 calculates the manifold entropy under different thresholds t:

[0079] ,

[0080] in: This represents the manifold entropy at a threshold t; This represents the probability distribution of the i-th region under threshold t, calculated as the number of pixels in that region divided by the total number of pixels; It represents the natural logarithm.

[0081] Also consider contrast measurement:

[0082] ,

[0083] in: This represents the contrast measure at a threshold t; and These represent the average grayscale values ​​of the foreground (dust) and background (PE film), respectively. and These represent the standard deviations of gray levels for the foreground and background, respectively. This represents the absolute value of the difference between the average gray values ​​of the foreground and background.

[0084] Construct a comprehensive evaluation function:

[0085] ,

[0086] in: For comprehensive evaluation function; and The weighting parameters are dimensionless, with preferred values ​​of 0.6 and 0.4 to balance the effects of entropy and contrast.

[0087] Through search Maximum threshold This is applied to low-dimensional representation to generate background-suppressed images. In applications involving the detection of micro-dust on PE film surfaces, This typically corresponds to the optimal separation point between the PE film background and microparticles, and its value is related to the PE film material, surface smoothness, and microparticle characteristics. For example, for PE films with high smoothness... Typically, the grayscale histogram is located between two peaks in a low-dimensional representation; for PE films with rougher surfaces, It may be necessary to bias towards background peaks to reduce false positives.

[0088] like Figure 3 As shown, in a preferred embodiment of the present invention, the differential geometric feature extraction module 3 includes a differential geometric feature extractor 31, a geodesic distance calculation unit 32, and a manifold probability distribution estimator 33.

[0089] The differential geometric feature extractor 31 is used to convert the image into a height function, calculate the Gaussian curvature, mean curvature, and shape index, and generate a differential geometric feature map. Specifically, the differential geometric features extracted by the differential geometric feature extractor 31 include principal curvature, Gaussian curvature, mean curvature, shape index, and rate of change of curvature, and generate a multi-scale curvature representation in multiple scale spaces.

[0090] In practical implementation, the background-suppressed image I(x,y) is considered as a height function h(x,y) = I(x,y) on a two-dimensional manifold. The first derivative (gradient) is calculated as follows:

[0091] ,

[0092] in: The gradient vector of the height function h is 2×1. and Let represent the partial derivatives of h with respect to x and y, respectively, which can be calculated using the finite difference method.

[0093] Calculate the second derivative (Hessian matrix):

[0094] ,

[0095] Where: H is the Hessian matrix with a dimension of 2×2; and Let represent the second-order partial derivatives of h with respect to x and y, respectively; and Let represent the cross partial derivatives of h. According to the properties of continuous functions, they are equal.

[0096] Calculate principal curvature based on the Hessian matrix. and (i.e., the eigenvalues ​​of the Hessian matrix). Then calculate the Gaussian curvature K and the mean curvature H:

[0097] ,

[0098] ,

[0099] Where: K is the Gaussian curvature, representing the intrinsic curvature of the surface at that point; H is the mean curvature, representing the extrinsic curvature of the surface at that point; and The principal curvature is the eigenvalue of the Hessian matrix.

[0100] Calculate the shape index S:

[0101] ,

[0102] Where: S is the shape index, with a value range of [-1, 1]; arctan is the arctangent function; The ratio of the sum of the principal curvatures to the difference of the principal curvatures is given when... When the limit value is reached, the limit value is taken.

[0103] Calculate the rate of change of curvature C:

[0104] ,

[0105] Where: C is the rate of change of curvature; This represents the gradient vector of the Gaussian curvature K; This represents the dot product of the gradient vectors, i.e., the squared magnitude of the gradient vector.

[0106] In PE film surface dust detection applications, these differential geometric features can effectively distinguish different types of surface structures. For example, dust particles typically appear as localized protrusions or depressions, with a large average curvature H and a small shape index S variation; while the texture of the PE film surface typically exhibits wavy undulations, with a small average curvature H and a large shape index S variation. This can be achieved through multiple scales. Repeat the above calculations to generate a multi-scale curvature representation. The preferred scale is selected as follows. , i=0,1,2, covering micro-dust features of different sizes.

[0107] The geodesic distance calculation unit 32 is connected to the differential geometric feature extractor 31 to receive differential geometric feature maps, define feature-based Riemannian metric tensors, calculate geodesic distances using the fast traversal method, and construct a distance matrix.

[0108] In practical implementation, we first define the feature-based Riemannian metric tensor M:

[0109] ,

[0110] Where: M(p) is the Riemannian metric tensor at point p, with a dimension of 2×2; I is a 2×2 identity matrix; Let p be the feature gradient vector at point p, with a dimension of 2×1; α is the transpose of the feature gradient vector, with a dimension of 1×2; α and β are weight parameters, dimensionless, with preferred values ​​of 0.2 and 0.8, respectively, used to balance the effects of isotropy and anisotropy. This represents the outer product of vectors, resulting in a 2×2 matrix.

[0111] In the application of micro-dust detection on PE film surfaces, the Riemannian metric tensor M defines a distance metric in the feature space, which can adaptively adjust the distance calculation based on the feature gradient. For example, in the edge region of micro-dust, the feature gradient is large, and M mainly consists of anisotropic components. The contribution of the gradient causes the geodesic distance to increase along the gradient direction; in flat regions, the characteristic gradient is small, and M is mainly composed of isotropic components. The contribution is close to the Euclidean distance.

[0112] Based on the Riemannian metric tensor, the fast travel method is used to calculate geodesic distances. The basic steps of the fast travel method are as follows:

[0113] 1. Initialization: Set the distance to the starting point to 0, and the distances to other points to infinity, and construct a priority queue of candidate points.

[0114] 2. Iterative calculation: Each time, process the point with the smallest distance in the queue, update the distances of its neighboring points, and readjust the priority queue.

[0115] 3. Termination condition: All points have been processed or the preset distance threshold has been reached.

[0116] Finally, the complete geodesic distance matrix D is constructed, where This represents the geodesic distance from point i to point j.

[0117] The manifold probability distribution estimator 33 is connected to the geodesic distance calculation unit 32. It is used to receive the distance matrix and the differential geometric feature map, estimate the probability density of the feature space, calculate the probability of each point belonging to noise or dust, and generate a noise-dust probability map.

[0118] In practical implementation, representative noise and dust samples are first selected, and the probability density function is estimated based on these samples. A kernel density estimation method is used, but geodesic distance is employed instead of Euclidean distance.

[0119] ,

[0120] in: This indicates that feature F belongs to a category. The probability density; For category The number of samples; For the kernel function (preferably a Gaussian kernel) h is the bandwidth parameter; This is the distance between geodesic lines; For category Sample characteristics; Indicates category Summing is performed on all samples.

[0121] For each pixel p, calculate its probability of belonging to noise. And the probability of belonging to dust particles :

[0122] ,

[0123] ,

[0124] in: This represents the posterior probability that point p belongs to noise. Let p represent the posterior probability that point p belongs to the dust particle; and Let represent the conditional probability density of the feature of point p under the noise and dust categories, respectively; and These are the prior probabilities of noise and dust, respectively. They are dimensionless and can be set empirically. The preferred values ​​are 0.7 and 0.3, respectively, reflecting the fact that noise is usually more common than dust on the surface of PE film. The normalization factor is calculated as follows:

[0125] ,

[0126] In the application of dust detection on PE film surfaces, the kernel density estimation method can adaptively estimate the probability of each point belonging to noise or dust based on the sample distribution in the feature space, without assuming a specific parameter distribution. Furthermore, using geodesic distance instead of Euclidean distance can better capture the nonlinear structure in the feature space, improving classification accuracy. Finally, a noise-dust probability map is generated, where the value of each pixel represents its probability of belonging to dust. The value range is [0,1].

[0127] like Figure 4 As shown, in a preferred embodiment of the present invention, the topological feature segmentation module 4 includes an edge detector 41, a morphological processing unit 42, a topology-preserving clusterer 43, and a feature space fusion unit 44.

[0128] Edge detector 41 is used to compute the Gaussian Laplacian operator, detect zero-crossing points, and extract edge feature maps.

[0129] In practical implementation, Gaussian smoothing is first applied to the noise-dust probability map P:

[0130] $ ,

[0131] in: This represents the image after Gaussian smoothing; * indicates a convolution operation. The standard deviation of the Gaussian kernel is expressed in pixels, with an optimal value of 1.5-2.5 pixels to balance edge detection accuracy and noise reduction capability. It is a two-dimensional Gaussian function; The normalization coefficients ensure that the sum of the Gaussian kernels is 1.

[0132] Then calculate the Laplace operator:

[0133] ,

[0134] in: express The Laplace operator; and They represent The second-order partial derivatives with respect to x and y.

[0135] Zero-crossing points are detected, i.e., locations where the Laplacian operator value changes from positive to negative or vice versa; these locations correspond to edges in the image. Simultaneously, gradient magnitude is used as a threshold condition to filter for significant edges.

[0136] ,

[0137] in: express The gradient magnitude is calculated as follows: ; The gradient threshold is preferably 1.5 times the average gradient value of the image.

[0138] In the application of micro-dust detection on PE film surfaces, the Gaussian Laplacian operator can effectively detect the edges of micro-dust particles, especially for circular or elliptical micro-dust particles, whose edges appear as closed zero-crossing loops in the Laplacian operator image. By adjusting... Values ​​that can accommodate dust particles of different sizes: smaller ones... Values ​​(e.g., 1.5 pixels) are suitable for detecting the edges of small dust particles (0.5-2μm); larger values... Values ​​(e.g., 2.5 pixels) are suitable for detecting the edges of large dust particles (5-10 μm). The final edge feature map E is generated.

[0139] The morphological processing unit 42 is connected to the edge detector 41 and is used to receive edge feature maps, perform multi-scale morphological dilation and erosion operations, and generate morphological enhancement feature maps.

[0140] In practical implementation, a multi-scale structural element set is designed. The graph includes structural elements of different sizes and shapes to accommodate dust particles of varying dimensions. A sequence of morphological operations is performed on the edge feature map E:

[0141] Expansion operation:

[0142] The expansion operation is defined as follows:

[0143] ,

[0144] in: This represents the dilated image; Represents the edge feature map; Represents a structural element; This indicates the operation of retrieving the maximum value; This represents the coordinate offset within a structuring element. The dilation operation expands the edges, filling tiny gaps.

[0145] Corrosion operation:

[0146] The corrosion operation is defined as follows:

[0147] ,

[0148] in: This represents the image after erosion. This represents the dilated image; Represents a structural element; This indicates the minimum value operation. The erosion operation removes minor noise while preserving the main structure.

[0149] Opening and closing operations: The closed operation is defined as expansion followed by corrosion: The opening operation is defined as corrosion followed by expansion: Where Close and Open represent the closing and opening operations, respectively; and This represents a structural element. Opening and closing operations further smooth edges and enhance region integrity.

[0150] In the application of micro-dust detection on PE film surfaces, morphological manipulation can effectively enhance the morphological features of micro-dust, fill edge gaps, remove stray noise, and make the micro-dust area more complete and prominent. The preferred structuring element size is 3×3 to 7×7 pixels to cover micro-dust features of different sizes: 3×3 structuring elements are suitable for handling small dust particles (0.5-2μm); 5×5 and 7×7 structuring elements are suitable for handling medium to large dust particles (2-10μm). By combining multi-scale results, a morphologically enhanced feature map M is generated, in which the micro-dust area is represented as a connected, bright region.

[0151] The topology-preserving clusterer 43 is connected to the morphological processing unit 42 and is used to receive morphological enhancement feature maps, construct high-dimensional feature vectors, and perform topology-preserving clustering to cluster pixels into speckle noise, blurred noise, dust, and background. Specifically, the topology-preserving clusterer 43 performs topology-preserving clustering in the following ways: constructing high-dimensional feature vectors containing spatial, morphological, and statistical features; defining a topology-preserving objective function; iteratively optimizing through gradient descent to preserve the local neighborhood structure; automatically determining the optimal number of categories; and classifying pixels into four categories.

[0152] In practical implementation, the high-dimensional feature vector F is first constructed:

[0153] ,

[0154] Where: F(p) represents the high-dimensional feature vector of point p; Spatial features include pixel location (x, y) and local neighborhood statistics; These are morphological features, including edge intensity, orientation, and curvature; These are statistical features, including local mean, variance, and histogram features; These are differential geometric features, including the previously calculated Gaussian curvature, mean curvature, etc.

[0155] Define the topology-preserving objective function:

[0156] ,

[0157] Where: J is the objective function value; and Let i and j represent the clustering space respectively, with a dimension of d×1 (d is usually 2 or 3). λ represents the elements of the similarity matrix; λ is the balance parameter, dimensionless, with an optimal value of 0.3-0.7; K is the number of categories; This represents the k-th category; For category The center; This represents summing over all pairs of points (i,j); This represents the summation over all categories; Indicates category Sum all points in the given information.

[0158] The first term of the objective function ensures the preservation of the local neighborhood structure, keeping similar points close in the cluster space; the second term promotes intra-cluster aggregation, bringing similar points closer to the cluster center.

[0159] Iterative optimization via gradient descent:

[0160] ,

[0161] in: and Let represent the cluster space representations for the t-th and t+1-th iterations, respectively; η is the learning rate, with an optimal value of 0.01-0.05; The objective function J represents the pair of The gradient.

[0162] In the application of micro-dust detection on PE film surfaces, topology-preserving clustering can classify pixels into different categories based on multidimensional features while maintaining the topological structure of the data. During the iteration process, the number of categories K is automatically adjusted, with an optimal range of 3-5, and finally determined to be 4, corresponding to four categories: speckle noise, blurred noise, micro-dust, and background. Speckle noise typically manifests as small, bright isolated points; blurred noise manifests as diffuse areas with unclear boundaries; micro-dust manifests as connected regions with clear boundaries and uniform internal structure; and the background manifests as flat areas with low grayscale.

[0163] The feature space fusion unit 44 is connected to the topology-preserving clusterer 43 to receive multi-level features and clustering results, calculate feature weights, perform nonlinear feature fusion, and generate the final dust location and feature information. Specifically, the feature space fusion unit 44 performs nonlinear feature fusion by: calculating the reliability of each feature layer, assigning adaptive weights, designing a nonlinear combination function, and integrating background suppression layer features, noise detection layer features, and morphological processing features to generate the final dust representation.

[0164] In practical implementation, the reliability R of each feature layer is first calculated:

[0165] ,

[0166] in: The reliability of the i-th feature layer is represented by inter-class variance; inter-class variance represents the variance between classes; intra-class variance represents the variance within classes. For category The number of samples; For category The center; The global center is used, and the average value is calculated as the mean of all samples. Features of sample j; This represents the summation over all categories; Indicates category Summing all samples in the dataset.

[0167] Based on reliability, adaptive weights are assigned. :

[0168] ,

[0169] in: represents the weight of the i-th feature layer, which is dimensionless; It represents the sum of the reliability of all feature layers.

[0170] Design a nonlinear combination function to integrate multi-level features:

[0171] ,

[0172] in: This represents the final feature representation of point p; The weights of the i-th feature layer; Let p be the representation of point p in the i-th feature layer; For non-linear mapping functions, the sigmoid function is preferred; This represents summing over all feature layers.

[0173] The sigmoid function is defined as:

[0174] ,

[0175] in: The value of the sigmoid function is (0, 1). and These are adjustable parameters that control the steepness and offset of the function, with the optimal value determined adaptively based on the feature distribution. is the base of the natural logarithm.

[0176] In the application of micro-dust detection on PE film surfaces, the feature fusion unit can adaptively adjust weights based on the reliability of each feature layer, integrating multi-level feature information. Through a nonlinear mapping function, the complementarity between different feature layers can be enhanced, improving overall detection performance. Ultimately, it generates micro-dust location and feature information, including the micro-dust's location coordinates (x, y), size (area and perimeter), shape (circularity and eccentricity), and reliability score.

[0177] like Figure 5 As shown, the present invention also provides a method for identifying electrostatic dust particles on the surface of a PE film, comprising the following steps:

[0178] Step S1: Acquire multi-angle, multi-wavelength images of the PE film surface;

[0179] Step S2: Construct a manifold representation of the PE film surface, separate the PE film background and potential micro-dust features, and generate a background-suppressed image;

[0180] Step S3: Extract the differential geometric features of the background suppression image, calculate the geodesic distance matrix, and generate a noise-dust probability map;

[0181] Step S4: Perform edge detection and morphological processing on the noise-dust probability map, and perform topology-preserving clustering to divide the image region into speckle noise, blurred noise, dust and background;

[0182] Step S5: Integrate multi-level features to determine the location and characteristic information of micro-dust on the PE film surface.

[0183] In practical applications, step S1 is implemented by image acquisition module 1, step S2 is implemented by multi-manifold mapping background suppression module 2, step S3 is implemented by differential geometric feature extraction module 3, and steps S4 and S5 are implemented by topological feature segmentation module 4.

[0184] Preferably, in step S2, the process of constructing the manifold representation of the PE film surface includes: constructing the local geometric structure of the image and generating a local geometric feature map; constructing a global similarity map based on the local geometric feature map, optimizing the manifold embedding objective function, and generating a low-dimensional representation of the PE film and dust particles; calculating the manifold entropy, determining the optimal segmentation threshold, and generating a background suppression image.

[0185] Preferably, in step S3, the process of extracting differential geometric features includes: converting the image into a height function, calculating the first derivative (gradient) and the second derivative (Hessian matrix); calculating the principal curvature, Gaussian curvature, mean curvature, shape exponent and rate of change of curvature based on the Hessian matrix; and generating a multi-scale curvature representation in multiple scale spaces.

[0186] Preferably, in step S4, the topology-preserving clustering process includes: constructing a high-dimensional feature vector containing spatial, morphological, statistical, and differential geometric features; defining a topology-preserving objective function, iteratively optimizing it through gradient descent while preserving the local neighborhood structure; and automatically determining the optimal number of categories to classify pixels into four categories: speckle noise, blurred noise, dust, and background.

[0187] Preferably, in step S5, the multi-level feature fusion process includes: calculating the reliability of each feature layer and assigning adaptive weights; designing a nonlinear combination function to integrate background suppression layer features, noise detection layer features, and morphological processing features; and generating the final dust representation, including the dust's location coordinates, size, shape, and reliability score.

[0188] The technical solution of the present invention is illustrated below through specific embodiments, but the scope of protection of the present invention is not limited thereto.

[0189] In this embodiment, the electrostatic dust identification system for the PE film surface of the present invention is used to detect a batch of PE film samples in production. The PE film is 50 μm thick and has dimensions of 300 mm × 400 mm.

[0190] First, multi-angle, multi-wavelength images of the PE film surface are acquired by the image acquisition module 1. A ring-shaped LED array 11 contains 16 LED light sources with a wavelength of 550nm, illuminating the PE film surface from different angles. A multi-wavelength illumination system 12 sequentially activates three sets of dark-field illumination modules with center wavelengths of 450nm, 550nm, and 650nm, respectively. Three photoelectric imaging detectors 13 are uniformly distributed at a 120° angle, with a resolution of 1920×1080 pixels and an exposure time set to 10ms.

[0191] Then, the multi-manifold mapping background suppression module 2 processes the acquired image. The local structure preservation unit 21 determines the 10 nearest neighbor set for each pixel and calculates the similarity using a Gaussian kernel function, with the Gaussian kernel's bandwidth parameter... The standard deviation of image grayscale is set to 0.3 times. The global structure mapping unit 22 constructs a global adjacency graph based on local features and solves the optimization problem to obtain a low-dimensional representation. The adaptive threshold adjustment unit 23 calculates the manifold entropy and contrast measure under different thresholds, with weight parameters... and The optimal segmentation thresholds were determined by setting them to 0.6 and 0.4 respectively, and background-suppressed images were generated.

[0192] Next, the differential geometric feature extraction module 3 processes the background-suppressed image. The differential geometric feature extractor 31 converts the image into a height function, calculates the first and second derivatives, and then calculates the principal curvature, Gaussian curvature, mean curvature, shape index, and rate of change of curvature. This is done at three scales. Repeated calculations are performed at (i=0,1,2) to generate a multi-scale curvature representation. The geodesic distance calculation unit 32 defines a feature-based Riemannian metric tensor with weight parameters. and The values ​​are set to 0.2 and 0.8 respectively, and the geodesic distance is calculated using the fast travel method to construct the distance matrix. The manifold probability distribution estimator 33 estimates the probability density function based on representative noise and dust samples. The prior probabilities of noise and dust are set to 0.7 and 0.3 respectively. The probability of each point belonging to noise or dust is calculated, and a noise-dust probability map is generated.

[0193] Finally, the topological feature segmentation module 4 processes the noise-dust probability map. The edge detector 41 applies Gaussian smoothing to the probability map, and the standard deviation of the Gaussian kernel is... Set to 2.0 pixels, calculate the Laplacian operator, detect zero-crossing points, and apply gradient thresholding. Edge feature maps are extracted by setting the gradient to 1.5 times the average gradient of the image. Morphological processing unit 42 uses multi-scale structuring elements ranging from 3×3 to 7×7 pixels to perform dilation, erosion, and opening / closing operations to generate morphologically enhanced feature maps. Topology-preserving clusterer 43 constructs high-dimensional feature vectors containing spatial, morphological, statistical, and differential geometric features, defines a topology-preserving objective function, and balances parameters. Set the learning rate to 0.5. Setting it to 0.03, pixels are categorized into four types: speckle noise, blurred noise, dust, and background. The feature space fusion unit 44 calculates the reliability of each feature layer, assigns adaptive weights, uses the sigmoid function as a non-linear mapping function, integrates multi-level features, and generates the final dust location and feature information.

[0194] In this embodiment, the system successfully detected micro-dust on the surface of the PE film, including micro-dust particles of various sizes ranging from 0.5 to 10 μm. Detection results showed a detection rate of 86.3% in the 0.5-2 μm range and 98.7% in the 2-10 μm range, with a false alarm rate of only 2.1%. The processing speed was 14.5 frames per second (1080p resolution), meeting the requirements for real-time detection.

[0195] In this embodiment, the electrostatic dust identification method for PE film surfaces of the present invention is used to detect a batch of PE film samples with different surface properties. PE films include three types: ordinary PE films, antistatic PE films, and optical-grade PE films.

[0196] First, step S1 is performed to acquire multi-angle, multi-wavelength images of the three types of PE film surfaces. Sixteen LED light sources with a wavelength of 550nm are used to illuminate the PE film surface from different angles. Three sets of dark field illumination modules with center wavelengths of 450nm, 550nm, and 650nm are activated sequentially. Three cameras with a resolution of 1920×1080 pixels are used to acquire images from different angles. The exposure time is adaptively adjusted within the range of 5-15ms according to the type of PE film.

[0197] Then, step S2 is executed to construct a manifold representation of the PE film surface. For each type of PE film, the 10 nearest neighbors set for each pixel is determined, similarity is calculated, and local geometry is constructed. A global adjacency graph is constructed based on local features, the manifold embedding objective function is optimized, and a low-dimensional representation is generated. The manifold entropy and contrast metric are calculated to determine the optimal segmentation threshold and generate a background-suppressed image. Parameters are adaptively adjusted according to the surface characteristics of different types of PE films: the Gaussian kernel bandwidth parameter for ordinary PE films... Set the standard deviation to 0.3 times the image grayscale standard deviation, 0.4 times for antistatic PE film, and 0.25 times for optical grade PE film.

[0198] Next, step S3 is executed to extract differential geometric features. The background suppression image is converted into a height function, and the first and second derivatives are calculated to obtain differential geometric features. This calculation is repeated at three scales to generate a multi-scale representation. A feature-based Riemannian metric tensor is defined to calculate the geodesic distance. The probability density function is estimated, and the probability that each point belongs to noise or dust is calculated to generate a noise-dust probability map. Parameters are adaptively adjusted based on the surface characteristics of different types of PE films: the weighting parameters for ordinary PE films... and The values ​​are set to 0.2 and 0.8 respectively, the values ​​for antistatic PE film are set to 0.25 and 0.75, and the values ​​for optical grade PE film are set to 0.15 and 0.85.

[0199] Then, step S4 is performed to conduct topology-preserving clustering. Gaussian smoothing is applied to the probability map, the Laplacian operator is calculated, zero-crossing points are detected, and edge feature maps are extracted; morphological operations are performed to generate morphologically enhanced feature maps; high-dimensional feature vectors are constructed, a topology-preserving objective function is defined, and clustering is performed to classify pixels into four categories. Parameters are adaptively adjusted based on the surface characteristics of different types of PE films: the Gaussian kernel standard deviation for ordinary PE films... Set to 2.0 pixels, antistatic PE film to 2.2 pixels, and optical grade PE film to 1.8 pixels.

[0200] Finally, step S5 is executed to fuse multi-level features. The reliability of each feature layer is calculated, and adaptive weights are assigned. A nonlinear combination function is designed to integrate multi-level features, generating the final dust location and feature information. For detected dust particles, their location coordinates, size, shape, and reliability score are recorded.

[0201] In this embodiment, the method successfully adapted to the surface characteristics of three different types of PE films. The detection results showed that the detection rate for ordinary PE films was 85.7% in the 0.5-2μm range and 98.2% in the 2-10μm range; the detection rate for antistatic PE films was 84.1% in the 0.5-2μm range and 97.8% in the 2-10μm range; and the detection rate for optical-grade PE films was 87.9% in the 0.5-2μm range and 99.3% in the 2-10μm range. The false alarm rate for all three types of PE films was controlled below 3%, and the processing speed was within the range of 12-15 frames / second, meeting the requirements for real-time detection.

[0202] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A PE film surface electrostatic dust identification system, characterized in that, include: The image acquisition module is used to acquire multi-angle, multi-wavelength images of the PE film surface; A multi-manifold mapping background suppression module, connected to the image acquisition module, is used to receive the multi-angle, multi-wavelength images, construct a manifold representation of the PE film surface, separate the PE film background and potential micro-dust features, and generate a background suppression image. The differential geometric feature extraction module is connected to the multi-manifold mapping background suppression module and is used to receive the background suppression image, extract differential geometric features, calculate the geodesic distance matrix, and generate a noise-dust probability map. The topological feature segmentation module, connected to the differential geometric feature extraction module, is used to receive the noise-dust probability map, perform edge detection and morphological processing, perform topological preservation clustering, divide the image region into speckle noise, blurred noise, dust and background, and determine the location and feature information of dust on the PE film surface. The multi-manifold mapping background suppression module includes: Local structure preservation units are used to construct the local geometry of an image and generate local geometric feature maps; A global structure mapping unit, connected to the local structure holding unit, is used to receive the local geometric feature map, construct a global similarity map, optimize the manifold embedding objective function, and generate a low-dimensional representation of the PE film and dust. An adaptive threshold adjustment unit, connected to the global structure mapping unit, is used to receive the low-dimensional representation, calculate the manifold entropy, determine the optimal segmentation threshold, and generate a background suppression image. The differential geometric feature extraction module includes: A differential geometric feature extractor is used to convert an image into a height function, calculate the Gaussian curvature, mean curvature, and shape index, and generate a differential geometric feature map. The geodesic distance calculation unit is connected to the differential geometric feature extractor and is used to receive the differential geometric feature map, define a feature-based Riemannian metric tensor, calculate the geodesic distance using the fast traversal method, and construct a distance matrix. The manifold probability distribution estimator, connected to the geodesic distance calculation unit, is used to receive the distance matrix and the differential geometric feature map, estimate the probability density of the feature space, calculate the probability that each point belongs to noise or dust, and generate a noise-dust probability map.

2. The electrostatic dust identification system for PE film surface according to claim 1, characterized in that, The image acquisition module includes: A ring-shaped LED array is used to illuminate the surface of the PE film from multiple angles; The multi-wavelength illumination system includes three sets of dark field illumination modules with different wavelengths, which are used to sequentially activate and generate illumination light of different wavelengths; Multiple photoelectric imaging detectors are connected to the multi-wavelength illumination system to acquire images of the PE film surface from different angles.

3. The electrostatic dust identification system for PE film surface according to claim 1, characterized in that, The topology feature segmentation module includes: An edge detector is used to compute the Laplacian Gaussian operator, detect zero-crossing points, and extract edge feature maps. A morphological processing unit, connected to the edge detector, is used to receive the edge feature map, perform multi-scale morphological dilation and erosion operations, and generate a morphologically enhanced feature map. A topology-preserving clusterer, connected to the morphological processing unit, is used to receive the morphological enhancement feature map, construct a high-dimensional feature vector, perform topology-preserving clustering, and cluster pixels into speckle noise, blurred noise, dust, and background. The feature space fusion unit, connected to the topology-preserving clusterer, is used to receive multi-level features and clustering results, calculate feature weights, perform nonlinear feature fusion, and generate the final dust location and feature information.

4. The electrostatic dust identification system for PE film surface according to claim 1, characterized in that, The local structure preservation unit constructs the local geometry by: determining the set of k nearest neighbors for each pixel, calculating the similarity between the pixel and its nearest neighbors, constructing a local similarity matrix, and extracting feature vectors to represent the local geometry.

5. The electrostatic dust identification system for PE film surface according to claim 1, characterized in that, The differential geometric features extracted by the differential geometric feature extractor include principal curvature, Gaussian curvature, mean curvature, shape index, and rate of change of curvature, and generate multi-scale curvature representations in multiple scale spaces.

6. The electrostatic dust identification system for PE film surface according to claim 3, characterized in that, The topology-preserving clusterer performs topology-preserving clustering in the following way: it constructs a high-dimensional feature vector containing spatial, morphological, and statistical features, defines a topology-preserving objective function, iteratively optimizes it through gradient descent, preserves the local neighborhood structure, automatically determines the optimal number of categories, and divides the pixels into four categories.

7. The electrostatic dust identification system for PE film surface according to claim 3, characterized in that, The feature space fusion unit performs nonlinear feature fusion by calculating the reliability of each feature layer, assigning adaptive weights, designing a nonlinear combination function, integrating background suppression layer features, noise detection layer features, and morphological processing features to generate the final dust representation.

8. A method for identifying electrostatic dust particles on the surface of a PE film, using the electrostatic dust particle identification system for the surface of a PE film as described in any one of claims 1-7, characterized in that, Includes the following steps: Acquire multi-angle, multi-wavelength images of the PE film surface; A manifold representation of the PE film surface is constructed to separate the PE film background and potential micro-dust features, generating a background-suppressed image. Extract the differential geometric features of the background-suppressed image, calculate the geodesic distance matrix, and generate a noise-dust probability map; Edge detection and morphological processing are performed on the noise-dust probability map, and topology-preserving clustering is performed to divide the image region into speckle noise, blurred noise, dust and background. By integrating multi-level features, the location and characteristic information of micro-dust on the PE film surface can be determined.

Images