A method for detecting surface uniformity of impregnated paper based on texture image analysis

Abstract

The present invention discloses a method for detecting the surface uniformity of impregnated paper based on texture image analysis, specifically relating to the fields of image enhancement and texture analysis. The method comprises obtaining an image of impregnated paper, constructing an edge map based on grayscale differences, extracting image regions in the edge map with edge values ​​below a first threshold, and defining these regions as fuzzy regions; calculating the edge value differences between adjacent pixels in the fuzzy regions to form a difference map, and extracting pixel regions from the difference map with edge change values ​​greater than a second threshold, defining these regions as breakage regions. This method identifies fuzzy regions with low edge values ​​and tracks their directional paths, thereby extracting the weak boundary between the film and the fiber. This method addresses the problem of traditional methods being unable to identify low-contrast edges and improves the detection accuracy of non-uniform regions.

Description

Technical Field

[0001] The present invention relates to the technical field of image enhancement and texture analysis, and more particularly to a method for detecting surface uniformity of gummed paper based on texture image analysis. Background Art

[0002] In the surface structure of impregnated paper, the natural fiber area and the glued area usually show a low-contrast, strong-transition fuzzy boundary layer in optical imaging. This phenomenon is due to the uncertainty of the diffusion and penetration of the glue on the fiber surface and the microscopic difference in surface tension.

[0003] In current image detection practices, surface structure segmentation is mainly performed using gradient-based edge detection algorithms (such as Sobel and Canny). However, these methods rely on grayscale mutation responses and have limited ability to identify blurred edges and gradient transition areas.

[0004] In particular, in the technical path composed of common image edge detection and region segmentation methods, once the image boundary loses significant grayscale slope changes, it will be judged as having no edge. As a result, the entire film and fiber interface area will be classified as a textured continuous surface. This misclassification directly causes the system to treat the edge area as part of the overall uniformity, thereby missing key areas where there is uneven glue coating or abnormal material transition.

[0005] In addition, when fuzzy boundaries are distributed over a large area in the image, they will interfere with the system's normal recognition of texture details, causing overall judgment deviations in the subsequent analysis of the entire image and distorted detection results.

[0006] In summary, the problem with the existing technology for analyzing the surface of impregnated paper is that there is a highly discontinuous but low-contrast boundary structure at the junction of the film and the fibers, which cannot be identified and extracted by conventional image analysis methods. As a result, local uneven areas are systematically masked, and the boundary distortion ultimately becomes a blind spot that affects the overall judgment accuracy. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a method for detecting the surface uniformity of impregnated paper based on texture image analysis. By constructing a directional map-guided path tracking and boundary closure mechanism, structural abnormal areas in fuzzy boundaries are identified to solve the problems raised in the above-mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: a method for detecting the surface uniformity of impregnated paper based on texture image analysis, comprising:

[0009] S1. Obtain an image of the gummed paper, construct an edge map based on grayscale differences, and extract image regions in the edge map where the edge value is lower than a first threshold value, defining the regions as fuzzy regions.

[0010] S2. Calculating the edge value difference between adjacent pixels in the fuzzy area to form a difference map, and extracting pixel areas with edge change values ​​greater than a second threshold from the difference map, defining them as break areas;

[0011] S3. Based on the position of each pixel in the fracture area, the gradient direction is calculated in the original impregnated paper image, and a direction map containing the entire image is constructed based on the gradient direction. The direction value corresponding to the fracture area in the direction map is used as a guide to generate a set of continuous paths, which are defined as derivation paths;

[0012] S4. Comparing the derived paths with the directional map, calculating a change in directional value within each derived path, and marking the corresponding path as a boundary path if the change in directional value exceeds a third threshold;

[0013] S5. Perform a contour closing operation on all boundary paths to generate closed areas, and output an image containing the closed areas as a non-uniform area detection result, which is used to indicate the structural abnormality position caused by boundary blur in the image.

[0014] In a preferred embodiment, it is defined in S1 Represents the pixel point in the edge map The edge value of

[0015]

[0016]

[0017]

[0018]

[0019] The fuzzy region judgment formula is defined as:

[0020]

[0021] in Indicates the original soaked paper image at pixel point Gray value at ; Represents the two-dimensional pixel coordinate position in the horizontal and vertical directions of the image; Indicated by pixels The pixel coordinate index in the horizontal and vertical directions within the neighborhood centered at ; Represents the pixel points in the neighborhood Gray value of Represents pixel points Grayscale difference in the horizontal direction; Represents pixel points Grayscale difference in the vertical direction; Indicated by pixels is the center and the radius is The neighborhood pixel set of Representing a collection The total number of pixels; Represents pixel points Grayscale deviation value; is the edge fusion function; represents the first threshold; is a set of fuzzy regions.

[0022] In a preferred embodiment, in S2, for each pixel in the fuzzy area extracted in S1, its edge value in the edge map is retrieved, and with the pixel as the center, four adjacent pixels above, below, left, and right are selected as reference neighborhoods, and the difference between the edge value of the current pixel and the edge value of each adjacent pixel is calculated, and the absolute value is taken as the local difference term;

[0023] The values ​​of the four local difference items are calculated by arithmetic average to obtain the local edge variation of the pixel in the blurred area. All local edge variations are recorded in sequence according to the pixel position to generate a difference map corresponding to the pixel position in the blurred area.

[0024] In the difference map, based on the set second threshold, the pixels whose local edge variation is greater than the second threshold are screened, and all pixel sets that meet the condition are defined as the fracture area.

[0025] In a preferred embodiment, in S3, for each pixel position in the fracture area, the grayscale value of the position is obtained in the original impregnated paper image, and the grayscale change of the pixel in the horizontal direction and the vertical direction is calculated respectively, and the gradient component is obtained using the central difference method; the gradient direction of the pixel is calculated based on the horizontal gradient component and the vertical gradient component, and the gradient direction is used as the direction value of the position;

[0026] The direction value of each pixel position in the fracture area is mapped into a two-dimensional structure with the same size as the original image according to its coordinate mapping in the original soaked paper image, and a direction map containing the entire image is constructed; in the direction map, only the positions corresponding to the fracture area are recorded with valid direction values, and the remaining positions are left blank or unassigned.

[0027] Taking each valid direction value in the fracture area of ​​the directional map containing the entire image as a guide, the adjacent pixel positions are iteratively tracked along the angular direction indicated by the valid direction value, and a continuous path is constructed by connecting point by point. The step-by-step pixel tracking and point-by-point connection operations guided by the direction value are repeated for all valid direction values ​​in the fracture area to generate a set of continuous paths, which are defined as the derivation path.

[0028] In a preferred embodiment, in calculating the change in direction value within each derivation path in S4, first, the direction values ​​corresponding to the pixel positions sequentially arranged in the derivation path in the direction map are extracted to form a direction value sequence; a first-order difference is performed on the direction value sequence to calculate the angle difference between any two adjacent direction values ​​to form a direction difference value sequence, wherein the direction difference value sequence represents the point-by-point change in the direction value within the path;

[0029] The directional difference sequence is divided into multiple non-overlapping sub-segments according to a fixed preset length. The standard deviation of the directional difference within each sub-segment is calculated as the directional fluctuation intensity of the sub-segment. The directional fluctuation intensity of all sub-segments is averaged to obtain the average directional fluctuation rate of the current derivation path. The average directional fluctuation rate is used to measure the local consistency of the directional change of the entire path.

[0030] Traversing the direction difference sequence of the derivation path, counting the positions in the direction difference sequence where the angle difference exceeds a preset angle threshold, and counting the number of jump points, which is used to quantify the degree of discontinuity of the direction change in the derivation path; combining the number of jump points with a preset jump penalty factor to generate a direction jump term, which is used to reflect the intensity of the discontinuous directional behavior in the path;

[0031] The absolute value sum of the directional differences, the average directional volatility and the directional jump term are weighted and superimposed to form a complete directional value change index, which is used to comprehensively measure the degree of directional change in the current derivation path and serves as the basis for judging whether to mark it as a boundary path.

[0032] In a preferred embodiment, in S5, a boundary connection determination is performed on each path marked as a boundary path in S4, and path pairs are screened for which the pixel Euclidean distance between the end points of the two paths is less than a preset distance threshold and the difference between the direction values ​​of the two path ends does not exceed a preset angle threshold, thereby constructing a candidate connection relationship graph;

[0033] According to the endpoint distance and direction consistency between the path pairs in the candidate connection relationship graph, a path merging operation is performed to form a continuous closed path set as a closed contour candidate set;

[0034] Performing a pixel filling operation on each closed path in the closed contour candidate set to form a closed region map, wherein the closed region map represents the image region enclosed by the boundary path;

[0035] The closed area map is superimposed on the coordinate space of the original image to generate a non-uniform area detection result map, in which the closed areas in the detection result map are used to mark the locations of structural abnormalities caused by boundary blur.

[0036] Technical effects and advantages of the present invention:

[0037] 1. This case identifies fuzzy areas with low edge values ​​and tracks their directional paths to extract the weak boundary between the film and the fiber. This solves the problem that traditional methods cannot identify low-contrast edges and improves the detection accuracy of non-uniform areas.

[0038] 2. By introducing edge difference map construction and multi-threshold judgment mechanism, the boundary fracture location in the fuzzy area can be identified, and the structural perception ability of transition anomalies in complex textures can be enhanced.

[0039] 3. Build a continuous path by guiding the direction of the fractured area, establish a direction map for the entire image, enhance the coherence of boundary path tracking, and be suitable for structure extraction in large fuzzy areas.

[0040] 4. By using the quantitative statistics of the direction changes within the path and combining it with the angle jump judgment method, texture interference path can be eliminated.

[0041] 5. Generate a closed area map through a closed path and map it back to the original image as the detection result, effectively marking the image abnormal areas caused by blurred boundaries, facilitating subsequent structural analysis and defect identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The figure is a flow chart of the method steps of the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] Refer to the instruction manual Figure 1 A method for detecting surface uniformity of impregnated paper based on texture image analysis according to an embodiment of the present invention includes:

[0045] S1. Obtain an image of the gummed paper, construct an edge map based on grayscale differences, and extract image regions in the edge map where the edge value is lower than a first threshold value, defining the regions as fuzzy regions.

[0046] S2. Calculating the edge value difference between adjacent pixels in the fuzzy area to form a difference map, and extracting pixel areas with edge change values ​​greater than a second threshold from the difference map, defining them as break areas;

[0047] S3. Based on the position of each pixel in the fracture area, the gradient direction is calculated in the original impregnated paper image, and a direction map containing the entire image is constructed based on the gradient direction. The direction value corresponding to the fracture area in the direction map is used as a guide to generate a set of continuous paths, which are defined as derivation paths;

[0048] S4. Comparing the derived paths with the directional map, calculating a change in directional value within each derived path, and marking the corresponding path as a boundary path if the change in directional value exceeds a third threshold;

[0049] S5. Perform a contour closing operation on all boundary paths to generate closed areas, and output an image containing the closed areas as a non-uniform area detection result, which is used to indicate the structural abnormality position caused by boundary blur in the image.

[0050] It should be noted that in the formula structure involved in this solution, dimensionless terms can serve as proportionality or structural adjustment factors. When combined with quantities with units, they only play a numerical scaling role and do not introduce new physical dimensions. Therefore, they will not change or confuse the overall unit system of expression. This combination of "dimensionless terms and units" can be understood as a composite structural expression commonly used in mathematical and physical modeling, conforming to the principle of dimensional consistency and having a clear physical interpretation basis.

[0051] Secondly, in the formula structure of this scheme, if multiple variables with different physical units are involved, including but not limited to time, mass or energy variables, their joint appearance is to express the collaborative modeling relationship of multiple physical mechanisms. Each variable can be formed into a unified structure through function mapping, ratio combination or normalization adjustment. The units and meanings are clear, and the overall expression conforms to the principle of dimensional consistency and the common formula of engineering modeling.

[0052] Any constants, weights, adjustment factors, threshold parameters, and proportional coefficients involved in this solution are all adjustable control parameters for different application environments. Their values ​​depend on the target device configuration, data input characteristics, and performance optimization goals. During the implementation phase, they are set within a reasonable range through model verification, performance constraints, or engineering calibration. Although these parameters do not have preset unique values, they have clear adjustment logic and calculation paths and are part of the deterministic setting process in engineering implementation. The purpose of such setting is to ensure that the solution is both universally adaptable, reproducible, and operable, without affecting its technical clarity and feasibility.

[0053] Defined in S1 Represents the pixel point in the edge map The edge value represents the composite response intensity that combines the multi-directional grayscale difference and the structural grayscale deviation;

[0054]

[0055]

[0056]

[0057]

[0058] The fuzzy region judgment formula is defined as:

[0059]

[0060] in Indicates the original soaked paper image at pixel point Gray value at ; Represents the two-dimensional pixel coordinate position in the horizontal and vertical directions of the image; Indicated by pixels The pixel coordinate index in the horizontal and vertical directions within the neighborhood centered at , Used to traverse each pixel position in the neighborhood; Represents the pixel points in the neighborhood Gray value of Represents pixel points Grayscale difference in the horizontal direction, the grayscale difference in the horizontal direction is used to define the absolute grayscale difference between adjacent left and right pixels; Represents pixel points Grayscale difference in the vertical direction, the grayscale difference in the vertical direction is used to define the absolute grayscale difference between adjacent upper and lower pixels; Indicated by pixels is the center and the radius is The neighborhood pixel set of Used to count the grayscale of local areas; Representing a collection The total number of pixels; Represents pixel points Grayscale deviation value, which represents the absolute deviation of the current pixel relative to the average grayscale of its neighborhood; is the edge blending function, and the definition of the edge blending function includes: ; where the variable They represent the grayscale difference of the pixel in the horizontal direction, the grayscale difference in the vertical direction, and the deviation of the pixel relative to the grayscale mean of its neighborhood in the input soaked paper image. The three together serve as The input variable is used to construct the pixel response value in the edge map; The grayscale deviation enhancement coefficient is used to adjust its weight in the edge map. It can be manually preset or empirically set based on labeled samples. The larger the grayscale deviation enhancement coefficient, the stronger the response to local grayscale disturbances. The smaller the grayscale deviation enhancement coefficient, the more dependent it is on directional grayscale differences, which is suitable for images with clear boundaries. Its value can be set in combination with image contrast and texture detail density. represents a first threshold, which is used to identify areas with lower edge values ​​in the edge map; is a fuzzy region set, which consists of all regions that satisfy of pixels.

[0061] In S2, for each pixel in the fuzzy area extracted in S1, its edge value in the edge map is retrieved, and with the pixel as the center, the four adjacent pixels above, below, left and right are selected as reference neighborhoods. The difference between the edge value of the current pixel and the edge value of each adjacent pixel is calculated, and the absolute value is taken as the local difference term;

[0062] Taking the arithmetic average of the values ​​of the four local difference items to obtain the local edge variation of the pixel in the blurred area, recording all local edge variations in sequence according to the pixel position, and generating a difference map corresponding to the pixel position in the blurred area. The difference map is used to characterize the edge response fluctuation distribution within the blurred area;

[0063] In the difference map, according to the set second threshold, the pixels whose local edge change is greater than the second threshold are screened, and all pixel sets that meet the conditions are defined as the fracture area, which is used to indicate the location where the edge discontinuity is significant in the fuzzy area, and serves as the input basis for the subsequent construction of the directional map.

[0064] In S3, for each pixel position in the fracture area, the grayscale value of the position is obtained in the original impregnated paper image, and the grayscale change of the pixel in the horizontal direction and the vertical direction is calculated respectively, and the gradient component is obtained by using the central difference method; the gradient direction of the pixel is calculated based on the horizontal gradient component and the vertical gradient component, and the gradient direction is used as the direction value of the position;

[0065] The direction value of each pixel position in the fracture area is mapped into a two-dimensional structure with the same size as the original image according to its coordinate mapping in the original soaked paper image, and a direction map containing the entire image is constructed; in the direction map, only the positions corresponding to the fracture area are recorded with valid direction values, and the remaining positions are left blank or unassigned.

[0066] Taking each valid direction value in the fracture area of ​​the directional map containing the entire image as a guide, the adjacent pixel positions are iteratively tracked along the angular direction indicated by the valid direction value, and a continuous path is constructed by connecting point by point. The step-by-step pixel tracking and point-by-point connection operations guided by the direction value are repeated for all valid direction values ​​in the fracture area to generate a set of continuous paths, which are defined as the derivation path.

[0067] In calculating the change in direction value within each derivation path in S4, first, the direction values ​​corresponding to the pixel positions arranged sequentially in the derivation path in the direction map are extracted to form a direction value sequence; a first-order difference is performed on the direction value sequence to calculate the angle difference between any two adjacent direction values ​​to form a direction difference value sequence, wherein the direction difference value sequence represents the point-by-point change in the direction value within the path;

[0068] The directional difference sequence is divided into multiple non-overlapping sub-segments according to a fixed preset length. The standard deviation of the directional difference within each sub-segment is calculated as the directional fluctuation intensity of the sub-segment. The directional fluctuation intensity of all sub-segments is averaged to obtain the average directional fluctuation rate of the current derivation path. The average directional fluctuation rate is used to measure the local consistency of the directional change of the entire path.

[0069] The directional difference sequence of the derived path is traversed, and the positions in the directional difference sequence where the angle difference exceeds a preset angle threshold are counted to form the number of jump points. The number of jump points is used to quantify the degree of discontinuity of the directional change in the derived path. The number of jump points is combined with a preset jump penalty factor to generate a directional jump term, which is used to reflect the intensity of the discontinuous directional behavior in the path. The jump penalty factor is a weight parameter used to control the intensity of the influence of the number of jump points on the directional value change index. Its value can be adjusted on the annotated samples to optimize the indicator differentiation between boundary paths and non-boundary paths. It is usually set to a positive number of the same magnitude as the average directional volatility.

[0070] The absolute value sum of the directional differences, the average directional volatility and the directional jump term are weighted and superimposed to form a complete directional value change index, which is used to comprehensively measure the degree of directional change in the current derivation path and serves as the basis for judging whether to mark it as a boundary path.

[0071] In S5, a boundary connection determination is performed on each path marked as a boundary path in S4, and path pairs are screened for which the pixel Euclidean distance between the end points of the two paths is less than a preset distance threshold, and the difference between the direction values ​​of the two path ends does not exceed a preset angle threshold, thereby constructing a candidate connection relationship graph. This screening is used to exclude connection relationships with obvious spatial gaps or inconsistent directions between the paths, ensuring that connections are only established between path pairs that meet the requirements of geometric proximity and continuous boundary extension directions, thereby providing a judgment basis for structural connectivity and directional consistency for the subsequent construction of closed contours.

[0072] According to the endpoint distance and direction consistency between the path pairs in the candidate connection relationship graph, a path merging operation is performed to form a continuous closed path set as a closed contour candidate set;

[0073] Performing a pixel filling operation on each closed path in the closed contour candidate set to form a closed region map, wherein the closed region map represents the image region enclosed by the boundary paths; wherein the pixel filling operation refers to determining the inner region of the image region enclosed by each closed path according to the path boundary, and assigning a fill mark value to all pixel positions within the boundary to form a completely closed binary region map; the closed region map refers to the image structure obtained after performing the pixel filling operation, wherein the filled pixel region corresponds to the inner image region enclosed by the closed path formed by the boundary paths;

[0074] The closed area map is superimposed on the coordinate space of the original image to generate a non-uniform area detection result map, wherein the closed areas in the detection result map are used to mark the locations of structural abnormalities caused by boundary blur. The process of generating the non-uniform area detection result map includes: in the coordinate space of the original image, mapping all pixel positions marked as closed areas in the closed area map to the original image, and superimposing a unified identification mark (such as a fixed grayscale value or color channel value) on these positions to highlight the area enclosed by the boundary path, thereby forming a non-uniform area detection result map for characterizing the locations of structural abnormalities.

[0075] It should be noted that in images of impregnated paper, there is often a low-contrast, highly transitional, fuzzy boundary between the glued area and the natural fiber area. This makes it impossible for traditional edge detection methods based on grayscale gradient mutations to effectively segment such areas. As a result, actual defects such as uneven glue coating and abnormal material transitions are systematically masked during image processing. The design of this solution is based on this recognition loss problem to eliminate the recognition blind spots caused by fuzzy boundaries.

[0076] During the image acquisition phase, by constructing an edge map and extracting regions with edge values ​​below a set threshold, image regions with potential structural ambiguity are identified and defined as fuzzy regions. This operation avoids the drawback of over-reliance on gradient strength in traditional methods and lays the foundation for subsequent differential structure extraction. Next, the solution further performs difference analysis of edge values ​​between pixels within the fuzzy region, forming a difference map and extracting pixel regions with edge change values ​​exceeding a second threshold, defining them as fracture regions. This step strengthens the judgment logic for potential discontinuities within the fuzzy boundary by introducing a local edge volatility indicator, increasing sensitivity to actual physical faults or abnormal locations of glue penetration.

[0077] On this basis, the solution calculates the gradient direction of each pixel in the fractured area in the original impregnated paper image, constructing a directional map for the entire image. Using the directional values ​​within the fractured area as a guide, a set of directionally consistent paths, defined as inferred paths, is formed through continuous tracing and connection operations. This path construction mechanism takes into account both the extension trend of the image structure and the directional constraints of the fractured area, making the extension judgment of the fuzzy boundary interpretable and controllable.

[0078] The solution then uses the derived paths as units, counting the changes in directional values ​​within each path in the directional map to determine whether they exceed a preset angle threshold. If so, the path is marked as a boundary path. This process introduces angle jump detection and a weighted penalty mechanism, enabling path screening to identify changes, thereby preventing continuous textures or normal transition areas from being mistakenly identified as boundary structures.

[0079] Finally, the scheme performs a contour closing operation based on the boundary path set. By constructing a connection relationship graph between closed path pairs and a pixel filling mechanism, a closed region map is generated. This map is then mapped onto the original image to form a non-uniform region detection result map, which is used to accurately mark structural abnormalities caused by blurred boundaries in the image.

[0080] This scheme uses the structural characteristics of the actual impregnated paper image as the judgment basis in each step, avoiding the problem of insufficient adaptability of general image processing algorithms to images of specific materials.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for detecting surface uniformity of impregnated paper based on texture image analysis, characterized in that: include: S1. Obtain an image of the gummed paper, construct an edge map based on grayscale differences, and extract image regions in the edge map where the edge value is lower than a first threshold value, defining the regions as fuzzy regions. S2. Calculating the edge value difference between adjacent pixels in the fuzzy area to form a difference map, and extracting pixel areas with edge change values ​​greater than a second threshold from the difference map, defining them as break areas; S3. Based on the position of each pixel in the fracture area, the gradient direction is calculated in the original impregnated paper image, and a direction map containing the entire image is constructed based on the gradient direction. The direction value corresponding to the fracture area in the direction map is used as a guide to generate a set of continuous paths, which are defined as derivation paths; S4. Comparing the derived paths with the directional map, calculating a change in directional value within each derived path, and marking the corresponding path as a boundary path if the change in directional value exceeds a third threshold; S5. Perform a contour closing operation on all boundary paths to generate closed areas, and output an image containing the closed areas as a non-uniform area detection result, which is used to indicate the structural abnormality position caused by boundary blur in the image.

2. The method for detecting surface uniformity of impregnated paper based on texture image analysis according to claim 1, characterized in that: Defined in S1 Represents the pixel point in the edge map The edge value of ； ； ； ； The fuzzy region determination formula is defined as; ； in Indicates the original soaked paper image at pixel point Gray value at ; Represents the two-dimensional pixel coordinate position in the horizontal and vertical directions of the image; Indicated by pixels The pixel coordinate index in the horizontal and vertical directions within the neighborhood centered at ; Represents the pixel points in the neighborhood Gray value of Represents pixel points Grayscale difference in the horizontal direction; Represents pixel points Grayscale difference in the vertical direction; Indicated by pixels is the center and the radius is The neighborhood pixel set of Representing a collection The total number of pixels; Represents pixel points Grayscale deviation value; is the edge fusion function; represents the first threshold; is a set of fuzzy regions.

3. The method for detecting surface uniformity of impregnated paper based on texture image analysis according to claim 2, characterized in that: In S2, for each pixel in the fuzzy area extracted in S1, its edge value in the edge map is retrieved, and with the pixel as the center, the four adjacent pixels above, below, left and right are selected as reference neighborhoods. The difference between the edge value of the current pixel and the edge value of each adjacent pixel is calculated, and the absolute value is taken as the local difference term; The values ​​of the four local difference items are calculated by arithmetic average to obtain the local edge variation of the pixel in the blurred area. All local edge variations are recorded in sequence according to the pixel position to generate a difference map corresponding to the pixel position in the blurred area. In the difference map, based on the set second threshold, the pixels whose local edge variation is greater than the second threshold are screened, and all pixel sets that meet the condition are defined as the fracture area.

4. The method for detecting surface uniformity of impregnated paper based on texture image analysis according to claim 3, characterized in that: In S3, for each pixel position in the fracture area, the grayscale value of the position is obtained in the original impregnated paper image, and the grayscale change of the pixel in the horizontal and vertical directions is calculated respectively, and the gradient component is obtained by using the central difference method; Calculate the gradient direction of the pixel based on the horizontal gradient component and the vertical gradient component, and use the gradient direction as the direction value of the position; The direction value of each pixel position in the fracture area is mapped into a two-dimensional structure with the same size as the original image according to its coordinate mapping in the original soaked paper image, and a direction map containing the entire image is constructed; in the direction map, only the positions corresponding to the fracture area are recorded with valid direction values, and the remaining positions are left blank or unassigned. Using each valid direction value in the fracture area of ​​the directional map containing the entire image as a guide, iteratively traces the adjacent pixel positions along the angle direction indicated by the valid direction value, and connects them point by point to construct a continuous path; The step-by-step pixel tracking and point-by-point connection operations guided by the direction values ​​are repeated for all valid direction values ​​in the fracture area to generate a set of continuous paths, which are defined as the derived paths.

5. The method for detecting surface uniformity of impregnated paper based on texture image analysis according to claim 4, characterized in that: In calculating the change in direction value within each derivation path in S4, first, the direction values ​​corresponding to the pixel positions arranged sequentially in the derivation path in the direction map are extracted to form a direction value sequence; a first-order difference is performed on the direction value sequence to calculate the angle difference between any two adjacent direction values ​​to form a direction difference value sequence, wherein the direction difference value sequence represents the point-by-point change in the direction value within the path; The directional difference sequence is divided into multiple non-overlapping sub-segments according to a fixed preset length. The standard deviation of the directional difference within each sub-segment is calculated as the directional fluctuation intensity of the sub-segment. The directional fluctuation intensity of all sub-segments is averaged to obtain the average directional fluctuation rate of the current derivation path. The average directional fluctuation rate is used to measure the local consistency of the directional change of the entire path. Traversing the direction difference sequence of the derivation path, counting the positions in the direction difference sequence where the angle difference exceeds a preset angle threshold, and counting the number of jump points, which is used to quantify the degree of discontinuity of the direction change in the derivation path; combining the number of jump points with a preset jump penalty factor to generate a direction jump term, which is used to reflect the intensity of the discontinuous directional behavior in the path; The absolute value sum of the directional differences, the average directional volatility and the directional jump term are weighted and superimposed to form a complete directional value change index, which is used to comprehensively measure the degree of directional change in the current derivation path and serves as the basis for judging whether to mark it as a boundary path.

6. The method for detecting surface uniformity of impregnated paper based on texture image analysis according to claim 5, characterized in that: In S5, a boundary connection determination is performed on each path marked as a boundary path in S4, and a pair of paths is selected in which the pixel Euclidean distance between the end points of the two paths is less than a preset distance threshold, and the difference between the direction values ​​of the two end points does not exceed a preset angle threshold, and a candidate connection relationship graph is constructed; According to the endpoint distance and direction consistency between the path pairs in the candidate connection relationship graph, a path merging operation is performed to form a continuous closed path set as a closed contour candidate set; Performing a pixel filling operation on each closed path in the closed contour candidate set to form a closed region map, wherein the closed region map represents the image region enclosed by the boundary path; The closed area map is superimposed on the coordinate space of the original image to generate a non-uniform area detection result map, in which the closed areas in the detection result map are used to mark the locations of structural abnormalities caused by boundary blur.

Images