A pavement disease feature extraction method based on disease mask image

Abstract

The present invention discloses a method for extracting features of road surface diseases based on a disease mask image, including using a connected component analysis method to preprocess the disease mask image of the road surface; performing a refinement algorithm and edge calculation on the preprocessed disease mask image to obtain the disease pixel width and the disease pixel area of ​​the disease mask image; performing a size transformation on the disease pixel width and the disease pixel area to obtain the actual width and actual area of ​​the road surface disease, and classifying the disease according to a severity classification standard; performing vertex judgment on the edge pixels of the disease to obtain polygonal information of the disease. The present invention extracts geometric quantitative indicators of road surface diseases from the disease mask image, and evaluates and judges the severity of the disease according to the set disease classification standard; at the same time, the present invention extracts polygonal vertices of the disease area based on the disease mask image, and records the shape characteristics of the disease.

Description

Technical Field

[0001] The invention belongs to the technical field of pavement defect feature extraction, and in particular relates to a pavement defect feature extraction method based on a defect mask image. Background Art

[0002] With the continuous development of modern transportation, roads, as the main channels for people to travel, are the basis for vehicle traffic and traffic operation. Once the road conditions are abnormal, it will not only directly affect the safety and comfort of driving, but also hinder the guarantee of road traffic efficiency and the maintenance of traffic order. Therefore, scientific evaluation and monitoring of road conditions are of great significance to ensure the safety and smoothness of roads and maintain traffic order.

[0003] Pavement damage is one of the manifestations of poor road conditions. It is caused by a variety of factors such as road design, materials, and traffic load. Common pavement damage phenomena include cracks, fissures, potholes, etc. Accurately identifying and evaluating the characteristics of pavement damage is a basic step in road maintenance and repair. Therefore, extracting the characteristic information of pavement damage is very important for achieving efficient road maintenance and management.

[0004] Based on computer vision and image processing technologies, a disease mask image can be extracted from a road surface image. However, the current mask image only records the extracted disease pixels and background pixels, and the severity and shape characteristics of the disease still need further processing and analysis to obtain.

[0005] Prior art 1, Image-based crack quantification, uses a defect image, a defect centerline and set kernels in different directions, calculates the defect width in pixels along the defect centerline, and then scales the calculated width to the actual width according to the actual scale.

[0006] Step 1: Use the morphological thinning method to thin the disease, and finally retain the pixels as the center line of the disease.

[0007] Step 2: To identify the direction perpendicular to the disease at each centerline pixel, use several kernels with different directions, and the corresponding kernel directions are set in equal increments.

[0008] Step 3: For each centerline pixel, the kernel corresponding to the maximum correlation value represents the tangent direction of the centerline. Therefore, the width direction of the disease is determined as the direction perpendicular to the tangent direction.

[0009] Step 4: For each center line pixel, count the original segmented defect pixels and the corresponding pixels in the width direction in the horizontal and vertical directions to obtain the defect width.

[0010] Step 5: Use the camera object distance and camera focal length information to obtain the scale parameter between the pixel size and the actual size, so as to convert the disease width into the actual size.

[0011] The disadvantages of this prior art are:

[0012] 1. This scheme uses the directional kernel of equal increments to query the tangent direction of the center line of the disease, which can only obtain the relevant values ​​of specific angles, resulting in large errors.

[0013] 2. This scheme is based on the calculation of correlation values ​​of multiple directional kernels to obtain the directional kernel corresponding to the maximum correlation value, which increases the computational cost and redundancy.

[0014] Prior art 2: A method for extracting geometric features of hidden road surface defect areas: This method uses defect ground penetrating radar (GPR) data to construct a reconstructed energy spectrum and then extract the geometric features of the defect area.

[0015] Step 1: Collect raw GPR data.

[0016] Step 2: Preprocess the raw data.

[0017] Step 3: Perform continuous wavelet transform on the preprocessed information and reconstruct the energy spectrum.

[0018] Step 4: Reduce the dimension of the energy spectrum and perform discrete wavelet decomposition on the reduced-dimensional data to remove interference.

[0019] Step 5: Obtain the probability density function of the energy function of the diseased area and the expected value of the function.

[0020] Step 6: Use the expected value to set the threshold, filter out the extreme points caused by interference, and find the two minimum points closest to the maximum point to determine the boundary width of the disease.

[0021] Step 7: Calculate the improved generalized S transform spectrum of the diseased area, reduce its dimension, find the maximum value with the smallest index in the dimension reduction function, and calculate the depth.

[0022] The disadvantages of the prior art 2 are:

[0023] 1. This solution uses ground penetrating radar to detect diseases, so it places high demands on professional equipment and technical personnel. At the same time, data acquisition also requires a large amount of time. Therefore, in practical applications, this solution has limitations in obtaining a large amount of disease data.

[0024] 2. This solution includes complex algorithms such as continuous wavelet transform, dimensionality reduction, and discrete wavelet decomposition, which hinder the effectiveness of algorithm implementation and computational efficiency in practical applications. Summary of the invention

[0025] The purpose of the present invention is to address the above-mentioned deficiencies in the prior art and to provide a method for extracting pavement disease features based on a disease mask image, so as to solve the problem that the existing mask image only records the extracted disease pixels and background pixels, and cannot obtain the severity and shape characteristics of the road disease.

[0026] In order to achieve the above object, the technical solution adopted by the present invention is:

[0027] A method for extracting pavement disease features based on a disease mask image comprises the following steps:

[0028] S1. Use the connected component analysis method to preprocess the road surface disease mask image;

[0029] S2, performing a thinning algorithm and edge calculation on the preprocessed disease mask image to obtain the disease pixel width and disease pixel area of ​​the disease mask image;

[0030] S3, transforming the pixel width and pixel area of ​​the damage to obtain the actual width and area of ​​the road damage, and classifying the damage according to the severity classification standard;

[0031] S4. Perform vertex judgment on edge pixels of the disease to obtain polygonal information of the disease.

[0032] Furthermore, step S1 includes the following steps:

[0033] S1.1. Using the connected component analysis algorithm, the connected region labels and connected region areas of the disease mask image are obtained;

[0034] S1.2. Arrange the connected regions in descending order and calculate the cumulative percentage of the connected regions ;

[0035] S1.3. According to the cumulative percentage, set the threshold to 85% and find the threshold index that satisfies the cumulative percentage greater than the threshold. R t :

[0036]

[0037] in, R i A sorted index corresponding to the connected area label;

[0038] S1.4. Keep indexes less than the threshold R t The corresponding connected region labels are obtained to obtain the preprocessed image connected region label set :

[0039]

[0040] Preserve the label set of the connected regions of the image The original pixel value of the connected area corresponding to the label in the middle is set, and the pixel values ​​of the remaining connected areas are set to 0 to obtain the pixel value of the preprocessed disease mask image:

[0041]

[0042] in, is the disease mask image ( m , n ) pixels at the position, is the value of the pixel; For label i The corresponding connected area pixel set.

[0043] Furthermore, the cumulative percentage of the connected area is calculated in step S1.2 ,for:

[0044]

[0045]

[0046] in, Sorting index R i The area size corresponding to the connected region.

[0047] Further, step S2 includes the following steps:

[0048] S2.1. Search for the defective pixels in the defective mask image, record the location of the defective pixels and their neighboring pixel information, and retain the defective pixels with connectivity of 1. I As the initial skeleton pixels;

[0049] S2.2, screening and marking the initial skeleton pixels to be removed according to the number of non-diseased pixels in the neighborhood of the initial skeleton pixels;

[0050] S2.3, removing the initial skeleton pixels marked as to be removed, and assigning the removed initial skeleton pixels to 0, to obtain a preliminarily thinned disease mask image;

[0051] S2.4, for the initially thinned disease mask image, repeat steps S2.1 to S2.3 until there are no initial skeleton pixels to be removed in the disease mask image, so as to obtain skeleton information of the disease mask image;

[0052] S2.5, taking the center pixels of the continuous defective pixels on the defective pixels of the first row and the last row for thinning;

[0053] S2.6, using the nearest neighbor search method to obtain the position information of the edge pixel closest to the skeleton pixel, and calculating the average width of the defect pixel of the defect mask image;

[0054] S2.7. Query the number of non-zero pixels in the preprocessed disease mask image to obtain the disease pixel area.

[0055] Furthermore, step S2.2 screens and marks the initial skeleton pixels to be removed according to the number of non-defective pixels in the neighborhood of the initial skeleton pixels, including:

[0056] Determine the number of non-diseased pixels in the neighborhood of the initial skeleton pixel. When the number of non-diseased pixels in the neighborhood does not exceed 1 and there is at least one defective pixel in the neighborhood pixels in the horizontal and vertical directions of the skeleton pixel, mark the skeleton pixel as the initial skeleton pixel to be removed.

[0057] When the number of non-defective pixels in the neighborhood is between 2 and 6, and there are at most two defective pixels in the neighborhood pixels in the horizontal and vertical directions of the skeleton pixel, the skeleton pixel is marked as the initial skeleton pixel to be removed.

[0058] Furthermore, the average width of the defect pixels of the defect mask image is calculated in step S2.6, including:

[0059] Calculate the Euclidean distance between two pixels :

[0060]

[0061] in, x 1 is the row number of pixel 1, y 1 is the column number of pixel 1; x 2 is the number of rows of pixel 2, y 2 is the number of columns of pixel 2;

[0062] Calculate the lesion width at the skeleton pixel :

[0063]

[0064] in, is the edge pixel closest to the skeleton pixel, is the skeleton pixel;

[0065] Calculate the average width of the diseased pixels in the disease mask image :

[0066]

[0067] in, n is the number of skeleton pixels, For the i The pixel width at skeleton pixels.

[0068] Further, step S3 includes the following steps:

[0069] S3.1. Acquisition equipment parameters and acquisition methods based on the road surface disease mask image to obtain width transformation parameters and area transformation parameters , and calculate the actual width of the disease and actual area :

[0070]

[0071] S3.2. According to the actual width of the disease and actual area , and combined with the severity classification standards, the disease is classified.

[0072] Further, step S4 includes the following steps:

[0073] S4.1. Based on the connected regions of the acquired disease image, according to the disease edge pixels Corresponding to the connected area , and the diseased edge pixels The value is :

[0074]

[0075] S4.2. Find the diseased edge pixels in a connected area, and obtain the diseased edge pixels and The angle formed by the edge pixels in the neighborhood. When a defective edge pixel has at least two adjacent edge pixels with the same angle, the defective edge pixel is marked as a defective edge pixel to be removed.

[0076] S4.3, remove the diseased edge pixels to be removed, and the diseased edge pixels that are not removed are the polygon vertices of the connected area, that is, the coordinates of the polygon vertices of the connected area are obtained;

[0077] S4.4. For other connected areas of the disease, repeat steps S4.2 and S4.3 until polygon information of different connected areas of the disease is obtained.

[0078] The pavement disease feature extraction method based on disease mask image provided by the present invention has the following beneficial effects:

[0079] 1. The present invention extracts geometric quantitative indicators of pavement defects from the defect mask image, and evaluates and judges the severity of the defects according to the set defect classification standard, which involves three types of defects: cracks, fissures, and potholes. At the same time, based on the defect mask image, the present invention extracts the polygon vertices of the defect area and records the shape characteristics of the defect.

[0080] 2. The present invention can filter the pixels of the mask image for disease segmentation, screen out isolated pixels to reduce the influence of background pixels; obtain the disease skeleton and edge from the mask image, and obtain the disease width and area based on nearest neighbor search and non-zero query. The set method has good portability for similar image-based quantization requirements.

[0081] 3. The present invention extracts vertex information of the defective polygon based on vertex discrimination of edge pixels, thereby reducing the cost of storage, transmission and application of the defective information.

[0082] 4. The present invention uses the nearest neighbor search to find the nearest disease edge pixel position of the disease skeleton, and uses the improved Euclidean distance to obtain the disease width at the disease skeleton, thereby obtaining the average disease width.

[0083] 5. When obtaining the width of the defect, the present invention reduces the steps of determining the tangent and vertical directions of the defect compared to the traditional method of using kernels in several directions; taking the kernels in four directions (i.e., 0° / 45° / 90° / 135° / ) as an example, when obtaining the width of the defect, the calculation efficiency is improved by 80.6% on average.

[0084] 6. The present invention converts the disease into polygons for storage and transmission, and uses the disease edge pixels obtained by contour detection to set a vertex screening method, that is, using the angle between the edge pixel point and the neighboring edge pixel, it is determined that the edge pixel is a vertex or a point that constitutes an edge, and then the vertex is included in the disease feature set to extract the characteristic information of the disease. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] Figure 1 The present invention is a flow chart of a method for extracting pavement disease features based on a disease mask image.

[0086] Figure 2 FIG. 1 is a schematic diagram of the numbering of a defective pixel I and its neighboring pixels according to the present invention.

[0087] Figure 3 This is an example diagram of connectivity judgment of the present invention.

[0088] Figure 4 This is an example of skeleton refinement of the first row of pixels of a disease image according to the present invention. DETAILED DESCRIPTION

[0089] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0090] Example 1

[0091] This embodiment provides a method for extracting pavement disease features based on a disease mask image. This embodiment extracts geometric quantitative indicators of pavement diseases from the disease mask image, and evaluates and judges the severity of the disease according to the set disease classification standard. This embodiment extracts polygonal vertices in the disease area to obtain the shape characteristics of the disease. Figure 1 , which specifically includes the following steps:

[0092] Step S1: Due to the influence of noise points, some background pixels in the disease mask image obtained based on image segmentation are mistakenly marked as disease areas. Based on this, this embodiment uses connected component analysis to obtain connected area information of the disease mask image, and deletes isolated pixel areas according to the area size of each connected area and a set threshold to reduce wrong markings, which specifically includes the following steps:

[0093] Step S1.1: Use the connected component analysis algorithm to obtain the connected region labels and connected region areas of the disease mask image, and define A i The connected region label is i The area of ​​the connected region;

[0094] Step S1.2: Sort the connected regions in descending order so that the label of each connected region corresponds to a sorting index. R i , and each R i The values ​​of are not equal, and the cumulative percentage of the connected area is calculated ;

[0095]

[0096]

[0097] in, Sorting index R i The area size corresponding to the connected region.

[0098] Step S1.3: According to the cumulative percentage, set the threshold to 85% and find the threshold index that satisfies the cumulative percentage greater than the threshold.R t :

[0099]

[0100] in, R i A sorted index corresponding to the connected area label;

[0101] Step S1.4: retain indexes less than the threshold R t The corresponding connected region labels are obtained to obtain the preprocessed image connected region label set :

[0102]

[0103] Preserve the label set of the connected regions of the image The original pixel value of the connected area corresponding to the label in the middle is set, and the pixel values ​​of the remaining connected areas are set to 0 to obtain the pixel value of the preprocessed disease mask image:

[0104]

[0105] in, is the disease mask image ( m , n ) pixels at the position, is the value of the pixel; For label i The corresponding connected area pixel set.

[0106] Step S2: Performing a thinning algorithm and edge calculation on the preprocessed disease mask image to obtain the disease pixel width and disease pixel area of ​​the disease mask image, which specifically includes the following contents:

[0107] Step S2.1, search for diseased pixels in the disease mask image, record the location of the diseased pixels and their Neighborhood pixel information, reference Figure 2 , P I For the diseased pixel I The value of

[0108] refer to Figure 3 For the defective pixel I, its connectivity v is checked, and the defective pixel I with connectivity of 1 is retained as the initial skeleton pixel, that is, only one of its neighboring defective pixels has a different value from the corresponding sequential neighboring pixel.

[0109] Step S2.2, screening and marking the initial skeleton pixels to be removed according to the number of non-diseased pixels in the neighborhood of the initial skeleton pixels;

[0110] Specifically, in this embodiment, for the initial skeleton pixels,

[0111] Determine the number of non-diseased pixels in the neighborhood of the initial skeleton pixel. When the number of non-diseased pixels in the neighborhood does not exceed 1 and there is at least one defective pixel in the neighborhood pixels in the horizontal and vertical directions of the skeleton pixel, mark the skeleton pixel as the initial skeleton pixel to be removed.

[0112] When the number of non-defective pixels in the neighborhood is between 2 and 6, and there are at most two defective pixels in the neighborhood pixels in the horizontal and vertical directions of the skeleton pixel, the skeleton pixel is marked as the initial skeleton pixel to be removed.

[0113] Step S2.3, remove the initial skeleton pixels marked as to be removed, and assign the removed initial skeleton pixels to 0, so as to obtain a preliminarily thinned disease mask image.

[0114] Step S2.4, for the initially thinned disease mask image, repeat steps S2.1 to S2.3 until there are no initial skeleton pixels to be removed in the disease mask image, so as to obtain skeleton information of the disease mask image;

[0115] Step S2.5: In the skeleton thinning process of this embodiment, the determination of the defective pixels in the first row and the last row of the image has not been considered;

[0116] Based on this, reference Figure 4 , take the center pixels of the continuous defective pixels on the defective pixels in the first row and the last row for refinement.

[0117] Step S2.6, use the contour detection algorithm to detect and obtain the edge information of the disease mask image. For each pixel in the skeleton, use the nearest neighbor search method to obtain the edge pixel position information closest to the skeleton pixel. The Euclidean distance calculates the distance between two points, and the pixel is not an abstract point, it has its own size and area. Based on this, this embodiment improves the Euclidean distance algorithm to obtain a more accurate pixel distance, which is as follows: the distance between two pixels is calculated using the improved Euclidean distance algorithm :

[0118]

[0119] in, x 1 is the row number of pixel 1, y 1 is the column number of pixel 1; x 2 is the number of rows of pixel 2, y 2 is the number of columns of pixel 2;

[0120] Calculate the lesion width at the skeleton pixel :

[0121]

[0122] in, is the edge pixel closest to the skeleton pixel, is the skeleton pixel;

[0123] Calculate the average width of the diseased pixels in the disease mask image :

[0124]

[0125] in, n is the number of skeleton pixels, For the i The pixel width at skeleton pixels.

[0126] Step S2.7, query the number of non-zero pixels in the preprocessed disease mask image to obtain the disease pixel area.

[0127] Step S3, transforming the pixel width and pixel area of ​​the defect to obtain the actual width and area of ​​the road defect, and classifying the defect according to the severity classification standard;

[0128] Step S3.1: Obtain width transformation parameters based on acquisition equipment parameters and acquisition methods of the road surface disease mask image and area transformation parameters , and calculate the actual width of the disease and actual area :

[0129]

[0130] Step S3.2: According to the actual width of the disease and actual area , and combined with the severity classification standards, the disease is classified.

[0131] The severity classification standards are shown in Table 1:

[0132] Table 1 Severity grading criteria

[0133]

[0134] Step S4, performing vertex determination on edge pixels of the defect to obtain polygon information of the defect;

[0135] Step S4.1: Based on the connected regions of the acquired disease image, the disease edge pixels are Corresponding to the connected area , and the diseased edge pixels The value is :

[0136]

[0137] Step S4.2: Find the diseased edge pixels in a connected area, and obtain the diseased edge pixels and The angle formed by the edge pixels in the neighborhood. When a defective edge pixel has at least two adjacent edge pixels with the same angle, the defective edge pixel is marked as a defective edge pixel to be removed.

[0138] Step S4.3, remove the diseased edge pixels to be removed, and the diseased edge pixels that are not removed are the polygon vertices of the connected area, that is, the coordinates of the polygon vertices of the connected area are obtained;

[0139] Step S4.4: Repeat steps S4.2 and S4.3 for other connected areas of the disease until polygon information of different connected areas of the disease is obtained.

[0140] Although the specific implementation of the invention is described in detail in conjunction with the drawings, it should not be understood as limiting the scope of protection of this patent. Within the scope described in the claims, various modifications and variations that can be made by those skilled in the art without creative work still fall within the scope of protection of this patent.

Claims

1. A pavement disease feature extraction method based on disease mask image, It is characterized in that The following steps are involved: S1. Use the connected component analysis method to preprocess the road surface disease mask image; S2, performing a thinning algorithm and edge calculation on the preprocessed disease mask image to obtain the disease pixel width and disease pixel area of ​​the disease mask image; S3, transforming the pixel width and pixel area of ​​the damage to obtain the actual width and area of ​​the road damage, and classifying the damage according to the severity classification standard; S4, performing vertex judgment on edge pixels of the disease to obtain polygonal information of the disease; The step S1 comprises the following steps: S1.

1. Using the connected component analysis algorithm, the connected region labels and connected region areas of the disease mask image are obtained; S1.

2. Arrange the connected regions in descending order and calculate the cumulative percentage of the connected regions ; S1.

3. According to the cumulative percentage, set the threshold to 85% and find the threshold index that satisfies the cumulative percentage greater than the threshold. R t : in, R i A sorted index corresponding to the connected area label; S1.

4. Keep indexes less than the threshold R t The corresponding connected region labels are obtained to obtain the preprocessed image connected region label set : Preserve the label set of the connected regions of the image The original pixel value of the connected area corresponding to the label in the middle is set, and the pixel values ​​of the remaining connected areas are set to 0 to obtain the pixel value of the preprocessed disease mask image: in, In the disease mask image ( m , n ) pixels at the position, is the value of the pixel; For label i The corresponding connected area pixel set; The cumulative percentage of the connected area calculated in step S1.2 ,for: in, Sorting index R i The area size corresponding to the connected region.

2. The method for extracting pavement damage features based on damage mask images according to claim 1, It is characterized in that The step S2 comprises the following steps: S2.

1. Search for the defective pixels in the defective mask image, record the location of the defective pixels and their neighboring pixel information, and retain the defective pixels with connectivity of 1. I As the initial skeleton pixels; S2.2, screening and marking the initial skeleton pixels to be removed according to the number of non-diseased pixels in the neighborhood of the initial skeleton pixels; S2.3, removing the initial skeleton pixels marked as to be removed, and assigning the removed initial skeleton pixels to 0, to obtain a preliminarily thinned disease mask image; S2.4, for the initially thinned disease mask image, repeat steps S2.1 to S2.3 until there are no initial skeleton pixels to be removed in the disease mask image, so as to obtain skeleton information of the disease mask image; S2.5, taking the center pixels of the continuous defective pixels on the defective pixels of the first row and the last row for thinning; S2.6, using the nearest neighbor search method to obtain the position information of the edge pixel closest to the skeleton pixel, and calculating the average width of the defect pixel of the defect mask image; S2.

7. Query the number of non-zero pixels in the preprocessed disease mask image to obtain the disease pixel area.

3. The method for extracting pavement damage features based on damage mask images according to claim 2, It is characterized in that The step S2.2 screens and marks the initial skeleton pixels to be removed according to the number of non-defective pixels in the neighborhood of the initial skeleton pixels, including: Determine the number of non-diseased pixels in the neighborhood of the initial skeleton pixel. When the number of non-diseased pixels in the neighborhood does not exceed 1 and there is at least one defective pixel in the neighborhood pixels in the horizontal and vertical directions of the skeleton pixel, mark the skeleton pixel as the initial skeleton pixel to be removed. When the number of non-defective pixels in the neighborhood is between 2 and 6, and there are at most two defective pixels in the neighborhood pixels in the horizontal and vertical directions of the skeleton pixel, the skeleton pixel is marked as the initial skeleton pixel to be removed.

4. The method for extracting pavement disease features based on disease mask images according to claim 2, It is characterized in that The step S2.6 calculates the average width of the defect pixels of the defect mask image, including: Calculate the Euclidean distance between two pixels : in, x 1 is the number of rows of pixel 1, y 1 is the number of columns of pixel 1; x 2 is the number of rows of pixel 2, y 2 is the number of columns of pixel 2; Calculate the lesion width at the skeleton pixel : in, is the edge pixel closest to the skeleton pixel, is the skeleton pixel; Calculate the average width of the diseased pixels in the disease mask image : in, n is the number of skeleton pixels, For the i The pixel width at skeleton pixels.

5. The method for extracting pavement disease features based on disease mask images according to claim 4, It is characterized in that The step S3 comprises the following steps: S3.

1. Acquisition equipment parameters and acquisition methods based on the road surface disease mask image to obtain width transformation parameters and area transformation parameters , and calculate the actual width of the disease and actual area : S3.

2. According to the actual width of the disease and actual area , and combined with the severity classification standards, the disease is classified.

6. The method for extracting pavement disease features based on disease mask images according to claim 1, It is characterized in that The step S4 comprises the following steps: S4.

1. Based on the connected regions of the acquired disease image, according to the disease edge pixels Corresponding to the connected area , and the diseased edge pixels The value is : S4.

2. Find the diseased edge pixels in a connected area, and obtain the diseased edge pixels and The angle formed by the edge pixels in the neighborhood. When a defective edge pixel has at least two adjacent edge pixels with the same angle, the defective edge pixel is marked as a defective edge pixel to be removed. S4.3, remove the diseased edge pixels to be removed, and the diseased edge pixels that are not removed are the polygon vertices of the connected area, that is, the coordinates of the polygon vertices of the connected area are obtained; S4.

4. For other connected areas of the disease, repeat steps S4.2 and S4.3 until polygon information of different connected areas of the disease is obtained.

Images