A road disease detection method and system

Abstract

This invention relates to the field of image recognition technology, specifically to a method and system for detecting road defects. In this invention, the original color image of the road surface is converted into a two-dimensional grayscale matrix, and candidate region matrices are divided according to fixed pixel dimensions, thus establishing a unified grayscale coordinate basis for the road surface image. Then, the coordinates of suspected defects are temporarily stored using Radon transform angle traversal sequences and parallel line projection values. Combined with the aspect ratio of the connected regions of suspected defects, a set of coordinates for linear defects is determined, allowing linear road defects with directional stability and spatial continuity to be separated first. Subsequently, an initial circular closed contour of the defect is constructed in the remaining two-dimensional grayscale matrix region based on the Chan-Vese level set algorithm. A set of coordinates for regional defects is selected based on the change in the total distribution difference, allowing regional road defects to be further identified within the grayscale region after excluding interference from linear road defects.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology, and in particular to a method and system for detecting road defects. Background Technology

[0002] Road defect detection methods refer to the process of inspecting and judging abnormalities such as damage and cracks on the surface and inside of highways or urban roads. This is used to assess the health status and structural integrity of roads. Visual inspection or machine vision algorithms are usually combined with technologies such as ground penetrating radar to achieve the purpose of identifying road defects.

[0003] Traditional road defect detection methods, while capable of examining and assessing anomalies on the road surface and internal structure through visual inspection, machine vision algorithms, or ground-penetrating radar, often suffer from several drawbacks. When both linear (crack-like) and regional road defects exist simultaneously during road construction, the same identification process can be applied to different defect types, leading to a mixture of linear extension features and regional grayscale distribution characteristics. Furthermore, directly identifying defects from original color images of the road surface is susceptible to variations in image color channels, local road surface grayscale changes, and inconsistent defect morphologies, resulting in a lack of continuous correspondence between suspected defect coordinate extraction, defect category differentiation, and defect coordinate location marking. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and system for detecting road defects.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a road defect detection method for detecting road surface defects during road construction, comprising the following steps: Acquire the original color image of the road surface, generate a two-dimensional grayscale matrix of the road surface from the original color image of the road surface, and divide the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices. Based on each candidate region matrix, extract the coordinates of suspected defects from the original color image of the road surface; Based on the coordinates of the temporarily stored pixels of suspected diseases, determine the set of coordinates of linear diseases or the remaining two-dimensional grayscale matrix area. Construct an initial circular closed contour of the disease at the center of the matrix region of the remaining two-dimensional gray matrix region, and calculate the total distribution difference that characterizes the difference in gray distribution of the road surface inside and outside the initial circular closed contour of the disease. The disease boundary coordinates of each pixel on the initial circular closed contour of the disease are obtained and adjusted. The total amount of different distribution differences during the adjustment process is collected. Based on the changes in the total amount of different distribution differences, a set of regional disease coordinates is selected. The set of linear disease coordinates and the set of regional disease coordinates are integrated as the road disease detection result.

[0006] As a further aspect of the present invention, the step of acquiring the original color image of the road surface, generating a two-dimensional grayscale matrix of the road surface from the original color image of the road surface, and dividing the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices includes: During road inspection, the original color images of the road surface are continuously captured by digital cameras through the data transmission bus. The color channel values ​​of the red, green and blue color channels at the location of each pixel in the original color image of the road surface are calculated according to the preset fixed weight ratio coefficient of the color channels. The weighted sum of the color channels is used as the gray value of the image pixel in the original color image of the road surface. The horizontal and vertical two-dimensional coordinates of the image pixel in the original color image of the road surface are obtained and integrated with the corresponding image pixel gray value to construct a two-dimensional gray matrix of the road surface. The two-dimensional grayscale matrix of the road surface is divided into grids according to the preset fixed pixel length and width of the region to generate multiple local candidate region matrices.

[0007] As a further aspect of the present invention, the step of extracting the coordinates of suspected defects temporarily stored pixels on the original color image of the road surface based on each candidate region matrix includes: Obtain a preset Radon transform angle traversal sequence starting from 0 degrees and ending at 180 degrees, as well as a fixed Radon transform angle increment interval step size and projection angle direction. Obtain the parallel straight line trajectory perpendicular to the projection angle direction within each candidate region matrix. Calculate the linear gray value accumulation result of all image pixel points along the parallel straight line trajectory in each projection angle direction, and use it as the parallel straight line projection value when the projection angle direction intersects with the parallel straight line trajectory. Select the first item from all the parallel line projection values ​​corresponding to each projection angle direction as the peak value of the angle projection corresponding to the projection angle direction. Obtain the horizontal and vertical two-dimensional coordinates of all image pixels on the parallel straight line trajectory corresponding to the peak value of the angle projection that is greater than the preset peak value threshold. Based on the horizontal and vertical two-dimensional coordinates of the image pixels, divide the coordinates of suspected defects into temporary pixel points on the original color image of the road surface.

[0008] As a further aspect of the present invention, determining the set of linear disease coordinates or the remaining two-dimensional grayscale matrix region based on the coordinates of temporarily stored pixels of suspected diseases includes: Obtain the connected region of the suspected disease formed by the coordinates of the temporarily stored pixels of the suspected disease, and calculate the length and width of the connected region of the suspected disease, and at the same time calculate the aspect ratio of the length and width of the connected region. The image pixel points with horizontal and vertical two-dimensional coordinates corresponding to the aspect ratio values ​​of the connected regions that are greater than the preset aspect ratio threshold are filtered out and classified as linear road defects with directional stability and spatial continuity. The coordinate positions corresponding to the linear road defects are stored as a linear defect coordinate set. Extract the gray values ​​of the remaining image pixels corresponding to the horizontal and vertical two-dimensional coordinate positions of the remaining image pixels that were not identified as having linear road defects, and obtain the remaining two-dimensional gray matrix region.

[0009] As a further aspect of the present invention, the step of constructing an initial circular closed contour of the disease at the center of the matrix region of the remaining two-dimensional gray-level matrix region, and calculating the total distribution difference characterizing the difference in gray-level distribution of the road surface inside and outside the initial circular closed contour of the disease, includes: Based on the Chan-Vese level set algorithm, an initial circular closed contour of the disease is constructed at the center of the matrix region of the remaining two-dimensional gray matrix region. All image pixels in the initial circular closed contour of the disease are obtained, and the average gray value of all image pixels is calculated as the average value inside the contour. At the same time, the sum of squares of the internal gray value deviation between the gray value of each image pixel inside the initial circular closed contour of the disease and the average value inside the contour is calculated as the internal difference. Calculate the average gray value of all image pixels outside the initial circular closed contour of the disease as the outer average value of the contour. At the same time, calculate the sum of squares of the outer gray value deviation between each image pixel outside the initial circular closed contour of the disease and the outer average value of the contour as the outer difference. Calculate the sum of the internal difference and the external difference to obtain the total distribution difference that characterizes the difference in grayscale distribution of the road surface inside and outside the initial circular closed contour of the defect.

[0010] As a further aspect of the present invention, the steps of acquiring and adjusting the disease boundary coordinates of each pixel on the initial circular closed contour of the disease, collecting the total amount of different distribution differences during the adjustment process, selecting a set of regional disease coordinates based on the changes in the total amount of different distribution differences, and integrating the set of linear disease coordinates and the set of regional disease coordinates as the road disease detection result include: Record the disease boundary coordinates of all pixels on the initial circular closed contour of the disease as the initial evolution state of the contour. Adjust the disease boundary coordinates of each pixel on the initial circular closed contour of the disease to perform contour evolution on the initial evolution state of the contour, and obtain the evolved disease boundary coordinates. Based on the evolved disease boundary coordinates, redivide the initial circular closed contour of the disease and repeat the calculation process of the total distribution difference. Calculate the total absolute difference of the total distribution difference obtained in two adjacent contour evolution states. Obtain the set of pixel coordinates of the closed geometric contour when the absolute difference of the total amount is less than the preset tolerance threshold for the difference change, and generate a set of coordinates of regional diseases; The location of road defects is marked on the original color image of the road surface according to the set of linear defect coordinates and the set of regional defect coordinates, and the road defect detection results are generated.

[0011] A road defect detection system is used for detecting road surface defects during road construction. The system includes: The image preprocessing module acquires the original color image of the road surface, generates a two-dimensional grayscale matrix of the road surface from the original color image, and divides the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices. The suspected disease extraction module extracts the coordinates of temporary pixels for suspected diseases from the original color image of the road surface based on each candidate region matrix. The linear disease screening module determines the set of linear disease coordinates or the remaining two-dimensional grayscale matrix area based on the coordinates of the temporarily stored pixels of suspected diseases. The total difference calculation module constructs an initial circular closed contour of the disease at the center of the matrix region of the remaining two-dimensional gray matrix region, and calculates the total distribution difference that represents the difference in gray distribution of the road surface inside and outside the initial circular closed contour of the disease. The disease result integration module acquires and adjusts the disease boundary coordinates of each pixel on the initial circular closed contour of the disease, collects the total amount of different distribution differences during the adjustment process, selects the regional disease coordinate set based on the changes in the total amount of different distribution differences, and integrates the linear disease coordinate set and the regional disease coordinate set as the road disease detection result.

[0012] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, the original color image of the road surface is converted into a two-dimensional grayscale matrix, and candidate region matrices are divided according to fixed pixel dimensions, so that the road surface image first forms a unified grayscale coordinate basis. Then, the coordinates of suspected defects are temporarily stored by extracting the angular traversal sequence of Radon transform and the projection values ​​of parallel lines. The coordinate set of linear defects is determined by combining the aspect ratio of the connected regions of the suspected defects, so that linear road defects with directional stability and spatial continuity are separated first. Subsequently, the initial circular closed contour of the defects is constructed in the remaining two-dimensional grayscale matrix region based on the Chan-Vese level set algorithm, and the coordinate set of regional defects is selected according to the change of the total distribution difference, so that regional road defects can be further determined in the grayscale region after excluding the interference of linear road defects. The Radon transform and the Chan-Vese level set algorithm form a sequential processing relationship. The former first locates linear road defects, while the latter processes regional road defects in the remaining two-dimensional grayscale matrix area. Finally, the coordinate sets of linear defects and regional defects are integrated to generate road defect detection results. This achieves the function of identifying different forms of road defects and uniformly labeling them under the same coordinate system of the original color image of the road surface. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the method steps of the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0015] Please see Figure 1 This invention provides a technical solution, a method for detecting road defects, used for detecting road surface defects during road construction, comprising the following steps: S1: Obtain the original color image of the road surface, generate a two-dimensional grayscale matrix of the road surface from the original color image, and divide the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices, including: During road inspection, the original color images of the road surface are continuously captured by digital cameras through the data transmission bus. The color channel values ​​of the red, green and blue color channels at the location of each pixel in the original color image of the road surface are calculated according to the preset fixed weight ratio coefficient of the color channels. The data transmission bus receives raw color images of the road surface continuously captured by a digital camera. This bus serves as the data channel for transmitting these raw color images from the digital camera to the image processing computer. The digital camera is the device used to acquire raw color images of the road surface during road inspections. The raw color image consists of color pixel data formed by the red, green, and blue color channel values. The horizontal and vertical coordinates of each pixel represent its position within the raw color image. Next, the raw color image is input into an image processing computer equipped with Python, NumPy, and scikit-image. Python is the software environment for matrix reading and image computation. NumPy is a software library for storing, reading, and sorting pixel data according to their horizontal and vertical coordinates. scikit-image is the software library for subsequently performing the Radon transform and Chan-Vese level set algorithm. In this step, NumPy is used to read the red, green, and blue color channel values ​​of each pixel in the raw color image. The preset fixed weight ratios for color channels are the fixed ratios used when converting red, green, and blue color channel values ​​to single-channel grayscale values. The process of setting the preset fixed weight ratios for color channels is as follows: first, determine the value ranges of red, green, and blue color channel values ​​in the original color image of the road surface; then, set the fixed weight ratios for red, green, and blue color channels according to the contribution relationship of the red, green, and blue color channel values ​​of the same image pixel to the grayscale value of the image pixel. The three fixed weight ratios for color channels satisfy the formula: fixed weight ratio for red color channel + fixed weight ratio for green color channel + fixed weight ratio for blue color channel = 1. The fixed weight ratios for red, green, and blue color channels remain fixed during the same road inspection process. Subsequently, the weighted sum of color channels is calculated as follows: red color channel value × red color channel fixed weight ratio coefficient + green color channel value × green color channel fixed weight ratio coefficient + blue color channel value × blue color channel fixed weight ratio coefficient, generating the weighted sum of color channels for each pixel in the original color image of the road surface.

[0016] The weighted sum of the color channels is used as the gray value of the image pixel in the original color image of the road surface. The horizontal and vertical two-dimensional coordinates of the image pixel in the original color image of the road surface are obtained and integrated with the corresponding image pixel gray value to construct a two-dimensional gray matrix of the road surface. The weighted sum of the color channels corresponding to each pixel in the original color image of the road surface is recorded as the grayscale value of the image pixel in the original color image of the road surface. The grayscale value of the image pixel is a single-channel value formed by processing the red, green, and blue color channel values ​​with a preset fixed weight ratio coefficient. Then, a two-dimensional grayscale matrix of the road surface is created in NumPy with the same number of horizontal and vertical image pixels as the original color image of the road surface. The horizontal position of the two-dimensional grayscale matrix of the road surface corresponds to the horizontal two-dimensional coordinate position of the image pixel in the original color image of the road surface, and the vertical position of the two-dimensional grayscale matrix of the road surface corresponds to the vertical two-dimensional coordinate position of the image pixel in the original color image of the road surface. Next, the grayscale value of the image pixel is written one by one according to the horizontal and vertical two-dimensional coordinate positions of the image pixels. When the horizontal and vertical two-dimensional coordinate positions of the image pixels correspond to the grayscale value of the image pixels, the grayscale value of the image pixels is written to the corresponding position in the two-dimensional grayscale matrix of the road surface. When there is no corresponding image pixel grayscale value, no valid grayscale value is written to the corresponding position in the two-dimensional grayscale matrix of the road surface, thus generating the two-dimensional grayscale matrix of the road surface.

[0017] The two-dimensional grayscale matrix of the road surface is divided into grids according to the preset fixed pixel length and width of the region to generate multiple local candidate region matrices. The process of setting the fixed pixel length and width of the region is as follows: obtain the physical mapping length and width of the single pixel road surface of the digital camera, and the physical coverage length and width of the road defect reference; calculate the ratio of the physical coverage length and width of the road defect reference to the physical mapping length and width of the single pixel road surface, and configure it as the preset fixed pixel length and width of the region. Read the two-dimensional grayscale matrix of the road surface and obtain the length and width dimensions of the physical mapping of the road surface per pixel and the length and width dimensions of the physical coverage of road defects from the digital camera. The length and width dimensions of the physical mapping of the road surface per pixel are the actual length and width dimensions of the road surface corresponding to a single image pixel in the original color image of the road surface. The length and width dimensions of the physical coverage of road defects are the reference coverage length and reference coverage width dimensions of the road defects on the road surface. The process of setting the length and width dimensions of the single-pixel road surface physical mapping is as follows: First, read the length and width dimensions of the road surface coverage corresponding to the original color image of the road surface captured by the digital camera. The length and width dimensions of the road surface coverage are obtained from the coverage length and coverage width dimensions of the digital camera's field of view on the road surface. Then, read the number of horizontal and vertical image pixels of the original color image of the road surface. The number of horizontal image pixels of the original color image of the road surface is obtained by counting each horizontally arranged image pixel in the original color image of the road surface, and the number of vertical image pixels of the original color image of the road surface is obtained by counting each vertically arranged image pixel in the original color image of the road surface. The length dimension of the single-pixel road surface physical mapping = the coverage length dimension of the road surface captured / the number of horizontal image pixels of the original color image of the road surface, and the width dimension of the single-pixel road surface physical mapping = the coverage width dimension of the road surface captured / the number of vertical image pixels of the original color image of the road surface. The reference physical coverage dimensions of road defects are registered before the road surface defect detection task begins during road construction. The reference physical coverage length corresponds to the reference length of the road defect on the road surface, and the reference physical coverage width corresponds to the reference width of the road defect on the road surface. Next, the region fixed pixel length is calculated as: region fixed pixel length = reference physical coverage length / single pixel road surface physical mapping length; region fixed pixel width is calculated as: region fixed coverage width / single pixel road surface physical mapping width. The region fixed pixel dimensions are composed of the region fixed pixel length and region fixed pixel width. Subsequently, in NumPy, the two-dimensional grayscale matrix of the road surface is meshed according to the region fixed pixel dimensions, generating multiple local candidate region matrices.

[0018] S2: Based on each candidate region matrix, extract the coordinates of suspected defects temporary pixels from the original color image of the road surface, including: Obtain a preset Radon transform angle traversal sequence starting from 0 degrees and ending at 180 degrees, as well as a fixed Radon transform angle increment interval step size and projection angle direction. Obtain the parallel straight line trajectory perpendicular to the projection angle direction within each candidate region matrix. Calculate the linear gray-level accumulation result of all image pixel gray values ​​along the parallel straight line trajectory in each projection angle direction, and use it as the parallel straight line projection value when the projection angle direction intersects with the parallel straight line trajectory. The candidate region matrix is ​​input into scikit-image, and projection processing is performed according to a preset Radon transform angle traversal sequence starting from 0 degrees and ending at 180 degrees, and a fixed Radon transform angle increment interval. The Radon transform is used to linearly accumulate the gray values ​​of image pixels in the candidate region matrix along the projection angle direction. The projection angle direction is the angle direction used when performing Radon transform on the candidate region matrix, and the parallel straight line trajectory is the path of image pixel values ​​in the candidate region matrix that is perpendicular to the projection angle direction. The preset Radon transform angle traversal sequence, starting from 0 degrees and ending at 180 degrees, is set up as follows: 0 degrees is used as the starting position of the angle traversal, and 180 degrees is used as the ending position. Projection angle directions are generated item by item between 0 degrees and 180 degrees according to a fixed Radon transform angle increment step size. The fixed Radon transform angle increment step size is set up by first determining the number of angle intervals between 0 degrees and 180 degrees. The number of angle intervals is the total number of intervals between adjacent projection angle directions that need to be divided between 0 degrees and 180 degrees. Then, the fixed Radon transform angle increment step size is determined according to the formula: Fixed Radon transform angle increment step size = 180 degrees / number of angle intervals. Subsequently, for each projection angle direction, multiple parallel straight line trajectories perpendicular to the projection angle direction are extracted from the candidate region matrix, and the gray values ​​of all image pixels are read along each parallel straight line trajectory; the parallel straight line projection value is the sum of the gray values ​​of all image pixels on the same parallel straight line trajectory; when there are image pixel gray values ​​on the parallel straight line trajectory, the parallel straight line projection value is generated; when there are no image pixel gray values ​​on the parallel straight line trajectory, no parallel straight line projection value is generated, and the parallel straight line projection value corresponding to each projection angle direction is generated.

[0019] Select the first item from all the parallel line projection values ​​corresponding to each projection angle direction as the peak value of the angle projection corresponding to the projection angle direction. Read all parallel line projection values ​​corresponding to each projection angle direction. The parallel line projection values ​​are the accumulated grayscale values ​​of the parallel line trajectories corresponding to the candidate region matrix under the same projection angle direction. Next, in NumPy, sort all parallel line projection values ​​corresponding to each projection angle direction according to their numerical values. The first value in the sorted list is the parallel line projection value with the largest numerical value under the projection angle direction. If there is only one parallel line projection value in the sorted list, register it as the angular projection peak value corresponding to the projection angle direction. If there are multiple values, select one according to the order of the parallel line trajectories in the candidate region matrix and register it as the angular projection peak value corresponding to the projection angle direction. If there is no corresponding parallel line projection value for a projection angle direction, do not register the angular projection peak value corresponding to the projection angle direction. This process generates the angular projection peak value for each projection angle direction.

[0020] Obtain the horizontal and vertical two-dimensional coordinates of all image pixels on the parallel straight line trajectory corresponding to the peak angle projection value that is greater than the preset peak projection target threshold. Based on the horizontal and vertical two-dimensional coordinates of the image pixels, divide the coordinates of suspected defects into temporary storage pixels on the original color image of the road surface. The system reads the angular projection peak value and the preset projection target peak threshold corresponding to each projection angle direction. The angular projection peak value is the projection value of the first parallel line in the same projection angle direction. The preset projection target peak threshold is the comparison boundary for the angular projection peak value to enter the extraction of the coordinates of suspected defect temporary storage pixels. The process of setting the preset projection target peak threshold is as follows: First, obtain the angular projection peak values ​​in all candidate region matrices corresponding to the original color image of the road surface, and count the total number of angular projection peak values. The total number of angular projection peak values ​​is obtained by counting the angular projection peak values ​​in all candidate region matrices one by one. The average value of all angular projection peak values ​​= the sum of all angular projection peak values ​​ / the total number of angular projection peak values. The average square deviation of all angular projection peak values ​​= the sum of the squares of the deviations between all angular projection peak values ​​and the average value of all angular projection peak values ​​ / the total number of angular projection peak values. The dispersion of all angular projection peak values ​​is... The preset projection target peak threshold is calculated as: average of all angle projection peak values ​​+ dispersion of all angle projection peak values. Then, the angle projection peak values ​​are compared with the preset projection target peak threshold. If the angle projection peak value is greater than the preset projection target peak threshold, the horizontal and vertical two-dimensional coordinates of all image pixels on the parallel straight line trajectory corresponding to the angle projection peak value are read; if it is equal to the preset projection target peak threshold, the horizontal and vertical two-dimensional coordinates of image pixels on the parallel straight line trajectory are not read; if it is less than the preset projection target peak threshold, the horizontal and vertical two-dimensional coordinates of image pixels on the parallel straight line trajectory are not read. Next, the read horizontal and vertical two-dimensional coordinates of the image pixels are mapped onto the original color image of the road surface to generate temporary coordinates of suspected defects.

[0021] S3: Based on the coordinates of the temporarily stored pixels of suspected defects, determine the set of coordinates of linear defects or the remaining two-dimensional grayscale matrix region, including: Obtain the suspected disease connected regions formed by the coordinates of the temporarily stored pixels of suspected diseases, and calculate the length and width of the connected regions of suspected diseases. At the same time, calculate the aspect ratio of the length and width of the connected regions. The process of obtaining the suspected disease connected regions is as follows: calculate the Euclidean distance between any two temporarily stored pixels of suspected diseases; extract the coordinates of multiple temporarily stored pixels of suspected diseases whose Euclidean distance is less than or equal to the preset connected search distance threshold, and establish the suspected disease connected regions. The coordinates of suspected road surface defects are read from the temporary storage pixels. These coordinates are the horizontal and vertical two-dimensional coordinates of the image pixels registered after angular projection peak comparison on the original color image of the road surface. The connected region of suspected defects is a continuous coordinate region established by multiple temporary storage pixels based on Euclidean distance, which is the straight-line interval between two temporary storage pixels in the original color image of the road surface. Next, the Euclidean distance between any two temporary storage pixels is calculated in NumPy: the horizontal coordinate difference = the horizontal coordinate position of the first temporary storage pixel - the horizontal coordinate position of the second temporary storage pixel; the vertical coordinate difference = the vertical coordinate position of the first temporary storage pixel - the vertical coordinate position of the second temporary storage pixel. The preset connectivity search distance threshold is the maximum allowed Euclidean distance when establishing connected regions for suspected disease nodes based on the coordinates of temporarily stored pixels. The process for setting this threshold involves first reading the fixed pixel length and width dimensions of the region, then selecting the smaller of these as the candidate connectivity scale. The preset connectivity search distance threshold is less than or equal to the candidate connectivity scale, ensuring that the connectivity relationships between temporarily stored pixels correspond to the local scale of the candidate region matrix. Subsequently, when the Euclidean distance is less than or equal to the preset connectivity search distance threshold, multiple temporarily stored pixels are grouped into the same suspected disease connected region; when it is greater than the threshold, multiple temporarily stored pixels are not grouped into the same suspected disease connected region. Next, the horizontal and vertical spans of the image pixels in the suspected disease connected region are read. The horizontal span = the maximum horizontal coordinate position in the suspected disease connected region - the minimum horizontal coordinate position in the suspected disease connected region; the vertical span = the maximum vertical coordinate position in the suspected disease connected region - the minimum vertical coordinate position in the suspected disease connected region. The length of the connected region is the larger of the horizontal and vertical spans, and the width of the connected region is the smaller of the horizontal and vertical spans. The aspect ratio of the connected region = the length of the connected region / the width of the connected region. The suspected disease connected region, the length of the connected region, the width of the connected region, and the aspect ratio of the connected region are generated.

[0022] The image pixel positions corresponding to the aspect ratio values ​​of connected regions that are greater than the preset aspect ratio threshold are filtered and classified as linear road defects with directional stability and spatial continuity. The coordinate positions corresponding to the linear road defects are stored as a linear defect coordinate set. The process reads the aspect ratio of the connected regions of suspected road defects and a preset aspect ratio threshold. The aspect ratio is the ratio of the length to the width of the connected region of a suspected road defect. The preset aspect ratio threshold is the comparison boundary for classifying suspected road defects as linear road defects with directional stability and spatial continuity. Directional stability means that suspected road defects are distributed along the same main extension direction, and spatial continuity means that the coordinates of the temporary pixels of suspected road defects within the connected regions meet the comparison condition of the preset connected search distance threshold. The process of setting the preset aspect ratio threshold is as follows: first, read the reference physical coverage length and width of the road defect, and then determine the preset aspect ratio threshold according to the formula: Preset aspect ratio threshold = Reference physical coverage length / Reference physical coverage width. The ratio between the reference physical coverage length and width corresponds to the degree of extension of the linear road defect in the length direction relative to the width direction. Subsequently, the aspect ratio of the connected region is compared with a preset aspect ratio threshold. When the aspect ratio is greater than the preset aspect ratio threshold, the corresponding image pixel's horizontal and vertical two-dimensional coordinate positions are classified as linear road defects with directional stability and spatial continuity. When the aspect ratio is equal to the preset aspect ratio threshold, it is not classified as a linear road defect with directional stability and spatial continuity. When the aspect ratio is less than the preset aspect ratio threshold, it is not classified as a linear road defect with directional stability and spatial continuity, and a set of linear defect coordinates is generated.

[0023] Extract the gray values ​​of the remaining image pixels corresponding to the horizontal and vertical two-dimensional coordinate positions that were not identified as having linear road defects, and obtain the remaining two-dimensional gray matrix region. The process begins by retrieving the horizontal and vertical coordinates of all pixels in the original color image of the road surface, along with the road surface's two-dimensional grayscale matrix and the coordinate set of linear defects. The linear defect coordinate set represents the set of coordinate positions corresponding to linear road defects that exhibit directional stability and spatial continuity. The remaining two-dimensional grayscale matrix region consists of the grayscale values ​​of image pixels not identified as having linear road defects. Next, in NumPy, the horizontal and vertical coordinates of each pixel in the original color image of the road surface are compared with the coordinate set of linear defects. If a pixel's horizontal and vertical coordinates are present in the linear defect coordinate set, its corresponding grayscale value is not retrieved. If not, the corresponding grayscale value is retrieved from the road surface's two-dimensional grayscale matrix. Finally, the horizontal and vertical coordinates of the image pixels not identified as having linear road defects and their corresponding grayscale values ​​are arranged in a matrix to generate the remaining two-dimensional grayscale matrix region.

[0024] S4: Construct an initial circular closed contour of the disease at the center of the remaining two-dimensional gray-level matrix region, and calculate the total distribution difference that characterizes the difference in gray-level distribution of the road surface inside and outside the initial circular closed contour of the disease, including: Based on the Chan-Vese level set algorithm, an initial circular closed contour of the disease is constructed at the center of the matrix region of the remaining two-dimensional gray matrix region. All image pixels in the initial circular closed contour of the disease are obtained, and the average gray value of all image pixels is calculated as the average value inside the contour. At the same time, the sum of squares of the internal gray value deviation between the gray value of each image pixel inside the initial circular closed contour of the disease and the average value inside the contour is calculated as the internal difference. Based on the Chan-Vese level set algorithm, an initial circular closed contour of the lesion is constructed at the center of the matrix region within the remaining two-dimensional grayscale matrix area. All image pixels within the initial circular closed contour are acquired, and the average grayscale value of all image pixels is calculated as the internal average value of the contour. Simultaneously, the sum of squared internal grayscale deviations between the grayscale value of each image pixel within the initial circular closed contour and the internal average value of the contour is calculated as the internal difference. The remaining two-dimensional grayscale matrix region is input into scikit-image. The Chan-Vese level set algorithm uses the level set function φ(x, y) to represent the initial circular closed contour of the lesion. φ(x, y) is the level set function value corresponding to the horizontal and vertical two-dimensional coordinate positions (x, y) of the image pixel. The zero horizontal line of the level set function φ(x, y) is used to represent the boundary of the initial circular closed contour of the lesion. First, the horizontal and vertical coordinate ranges of the remaining two-dimensional grayscale matrix region are read. The center position of the matrix region is determined by the midpoint of the horizontal and vertical coordinate ranges of the remaining two-dimensional grayscale matrix region. Then, the initial circular closed contour of the lesion is constructed with the center position of the matrix region as the center. The initial circular closed contour of the lesion is a circular closed boundary set in the remaining two-dimensional gray-level matrix region. When the level set function φ(x, y) = 0, the horizontal and vertical two-dimensional coordinate positions (x, y) of the image pixel are located on the boundary of the initial circular closed contour of the lesion and are recorded as the lesion boundary coordinates; when the level set function φ(x, y) < 0, the horizontal and vertical two-dimensional coordinate positions (x, y) of the image pixel are located inside the initial circular closed contour of the lesion; when the level set function φ(x, y) > 0, the horizontal and vertical two-dimensional coordinate positions (x, y) of the image pixel are located outside the initial circular closed contour of the lesion. Subsequently, according to the sign of the level set function φ(x, y), the horizontal and vertical two-dimensional coordinate positions of the image pixels in the remaining two-dimensional gray-level matrix region are classified as inside or outside, and the gray-level values ​​of the image pixels corresponding to the level set function φ(x, y) < 0 are classified as the gray-level values ​​of all image pixels inside the initial circular closed contour of the lesion. Next, the number of image pixels inside the initial circular closed contour of the disease is counted. The number of image pixels inside the initial circular closed contour of the disease is obtained by counting the gray values ​​of all image pixels inside the initial circular closed contour of the disease one by one. The average value inside the contour = the sum of the gray values ​​of all image pixels inside the initial circular closed contour of the disease / the number of image pixels inside the initial circular closed contour of the disease; the internal gray value deviation = the gray value of each image pixel inside the initial circular closed contour of the disease - the average value inside the contour; the square of the internal gray value deviation = the internal gray value deviation × the internal gray value deviation; the sum of the squares of the internal gray value deviations corresponding to all image pixels inside the initial circular closed contour of the disease is used to obtain the internal difference.

[0025] Calculate the average gray value of all image pixels outside the initial circular closed contour of the disease as the outer average value of the contour. At the same time, calculate the sum of squares of the outer gray value deviation between each image pixel outside the initial circular closed contour of the disease and the outer average value of the contour as the outer difference. Under the same initial circular closed contour of the disease, according to the sign of the level set function φ(x, y), read the gray values ​​of all image pixels outside the initial circular closed contour of the disease corresponding to φ(x, y) > 0. The area outside the initial circular closed contour of the disease is the set of image pixels outside the enclosed area of ​​the initial circular closed contour of the disease, and count the number of image pixels outside the initial circular closed contour of the disease. The number of image pixels outside the initial circular closed contour of the disease is obtained by counting the gray values ​​of all image pixels outside the initial circular closed contour of the disease one by one. Next, the average value of the contour outside = the sum of the gray values ​​of all image pixels outside the initial circular closed contour of the disease / the number of image pixels outside the initial circular closed contour of the disease; the external gray value deviation = the gray value of each image pixel outside the initial circular closed contour of the disease - the average value of the contour outside; the square of the external gray value deviation = the external gray value deviation × the external gray value deviation; sum the squares of the external gray value deviations corresponding to all image pixels outside the initial circular closed contour of the disease to obtain the external difference.

[0026] Calculate the sum of internal and external differences to obtain the total distribution difference that characterizes the difference in grayscale distribution of the road surface inside and outside the initial circular closed contour of the disease. The internal and external difference quantities are read. The internal difference quantity corresponds to the sum of the squared deviations of the gray values ​​of the pixels inside the initial circular closed contour of the hazard from the average value inside the contour. The external difference quantity corresponds to the sum of the squared deviations of the gray values ​​of the pixels outside the initial circular closed contour of the hazard from the average value outside the contour. Then, the internal and external difference quantities are summed to obtain the total distribution difference. The total distribution difference is used to characterize the degree of difference between the gray value distribution of the road surface inside and outside the initial circular closed contour of the hazard. The smaller the total distribution difference, the higher the concentration of the gray values ​​of the pixels inside the initial circular closed contour around the average value inside the contour, and the higher the concentration of the gray values ​​of the pixels outside the contour around the average value outside the contour. Conversely, the larger the total distribution difference, the higher the dispersion of the gray values ​​of the pixels inside and outside the initial circular closed contour.

[0027] S5: Obtain and adjust the boundary coordinates of each pixel on the initial circular closed contour of the disease, collect the total amount of different distribution differences during the adjustment process, select the regional disease coordinate set based on the changes in the total amount of different distribution differences, integrate the linear disease coordinate set and the regional disease coordinate set, and use them as the road disease detection results, including: The disease boundary coordinates of all pixels on the initial circular closed contour of the disease are recorded as the initial evolution state of the contour. The disease boundary coordinates of each pixel on the initial circular closed contour of the disease are adjusted to perform contour evolution on the initial evolution state, resulting in evolved disease boundary coordinates. Based on the evolved disease boundary coordinates, the initial circular closed contour of the disease is re-divided, and the calculation process of the total distribution difference is repeated. The absolute difference of the total distribution difference obtained in two adjacent contour evolution states is calculated. The contour evolution includes: extracting the geometric normal evolution direction of the disease boundary coordinates corresponding to each pixel on the initial circular closed contour of the disease; calculating the position translation step of the disease boundary coordinates in the geometric normal evolution direction according to the numerical descent gradient of the total distribution difference; and translating the disease boundary coordinates along the geometric normal evolution direction by the translation step to obtain the evolved disease boundary coordinates. The disease boundary coordinates of all pixels on the initial circular closed contour of the lesion are read. The disease boundary coordinates are the horizontal and vertical two-dimensional coordinate positions of the image pixels corresponding to the pixels on the initial circular closed contour of the lesion, and all disease boundary coordinates are recorded as the initial evolution state of the contour. Next, in scikit-image, the geometric normal evolution direction of the disease boundary coordinates corresponding to each pixel is extracted along the initial circular closed contour of the lesion. The geometric normal evolution direction is the normal movement direction of the disease boundary coordinates relative to the initial circular closed contour of the lesion. Then, the evolved disease boundary coordinates are formed along the geometric normal evolution direction, and the gray values ​​of all image pixels inside the initial circular closed contour of the lesion and the gray values ​​of all image pixels outside the initial circular closed contour of the lesion are re-divided based on the evolved disease boundary coordinates. The average value inside the contour, the amount of internal difference, the average value outside the contour, the amount of external difference, and the total distribution difference are repeatedly obtained. Subsequently, the numerical descent gradient of the total distribution difference is calculated based on the total distribution difference in the current contour evolution state and the total distribution difference corresponding to the evolved disease boundary coordinates. The numerical descent gradient of the total distribution difference = the total distribution difference in the current contour evolution state - the total distribution difference corresponding to the evolved disease boundary coordinates. When the numerical descent gradient of the total distribution difference is greater than 0, the total distribution difference decreases, and the disease boundary coordinates are translated along the geometric normal evolution direction by the translation step size. When it is equal to 0, the total distribution difference remains unchanged, and the disease boundary coordinates are not translated. When it is less than 0, the total distribution difference increases, and the disease boundary coordinates are not translated. The translation step size is jointly determined by the numerical descent gradient of the total distribution difference and the pixel distance in the geometric normal evolution direction where the disease boundary coordinates are located. The translation step size = the pixel distance between the evolved disease boundary coordinates corresponding to the numerical descent gradient of the total distribution difference and the disease boundary coordinates in the current contour evolution state. After the disease boundary coordinates are translated, the evolved disease boundary coordinates are obtained. Next, based on the evolved disease boundary coordinates, the total distribution difference is repeatedly calculated, and the total distribution difference obtained in two adjacent contour evolution states is read. The total absolute difference is the absolute value of the difference between the total distribution difference in the latter two adjacent contour evolution states and the total distribution difference in the former two adjacent contour evolution states. The evolved disease boundary coordinates and the total absolute difference are then generated.

[0028] Obtain the set of pixel coordinates of the closed geometric contour when the total absolute difference is less than the preset tolerance threshold for difference change, and generate a set of coordinates of regional diseases; Read the total absolute difference and the preset difference change tolerance threshold. The preset difference change tolerance threshold is the comparison boundary for judging whether the total change in the distribution difference between two adjacent contour evolution states enters the closed geometric contour coordinate extraction. The process of setting the preset tolerance threshold for difference changes is as follows: First, read the range of gray values ​​and the number of gray levels of the image pixels corresponding to the original color image of the road surface. The range of gray values ​​of the image pixels is the minimum to maximum gray value allowed for the image pixels, and the number of gray levels is the total number of gray levels that can be distinguished within the range of gray values ​​of the image pixels. When the gray values ​​of the image pixels are represented by integer gray levels from 0 to 255, the numerical interval between adjacent gray levels is 1. When the gray values ​​of the image pixels are represented by normalized gray values ​​from 0 to 1, the numerical interval between adjacent gray levels is 1 / 255. The numerical interval between adjacent gray levels is used as the gray level interval. Then, read the number of image pixels inside the initial circular closed contour of the lesion and the number of image pixels outside the initial circular closed contour of the lesion. The preset tolerance threshold for difference changes is the square of the gray level interval. The square of the gray level interval is used to identify the smallest squared difference that can be generated by the change of adjacent gray levels in the corresponding internal gray level deviation square and external gray level deviation square. Subsequently, the absolute difference of the total amount is compared with the preset tolerance threshold for difference changes. When the absolute difference of the total amount is less than the preset tolerance threshold for difference changes, the set of pixel coordinates of the current closed geometric contour is read and a set of coordinates of regional diseases is generated. When it is equal to the preset tolerance threshold for difference changes, the set of pixel coordinates of the current closed geometric contour is not read. When it is greater than the preset tolerance threshold for difference changes, the set of pixel coordinates of the current closed geometric contour is not read and a set of coordinates of regional diseases is generated.

[0029] Based on the original color image of the road surface, the coordinate positions of the defects are marked on the linear defect coordinate set and the regional defect coordinate set to generate road defect detection results. Read the original color image of the road surface, the coordinate set of linear defects, and the coordinate set of regional defects. The coordinate set of linear defects is the set of coordinate positions corresponding to linear road defects with directional stability and spatial continuity. The coordinate set of regional defects is the set of pixel coordinates corresponding to closed geometric contours. Next, in NumPy, the horizontal and vertical coordinates of each image pixel on the original color image of the road surface are read one by one and compared with the coordinate sets of linear defects and regional defects. If the horizontal and vertical coordinates of the image pixel exist only in the coordinate set of linear defects, the defect coordinate position is marked on the original color image of the road surface according to the coordinate set of linear defects. If they exist only in the coordinate set of regional defects, the defect coordinate position is marked on the original color image of the road surface according to the coordinate set of regional defects. If they exist in both the coordinate sets of linear defects and regional defects, the defect coordinate position is marked on the original color image of the road surface according to the coordinates of the overlapping coordinate sets of linear defects and regional defects. If they exist in neither the coordinate set of linear defects nor the coordinate set of regional defects, no defect coordinate position is marked.

[0030] Overall, the digital camera first acquires the original color image of the road surface, then converts it into a two-dimensional grayscale matrix. The Radon transform first separates linear road defects from the candidate region matrix. The Chan-Vese level set algorithm then processes regional road defects within the remaining two-dimensional grayscale matrix region. The coordinate set of linear defects output by the Radon transform participates in forming the remaining two-dimensional grayscale matrix region. The Chan-Vese level set algorithm continues to extract the coordinate set of regional defects based on the remaining two-dimensional grayscale matrix region. This allows the two algorithms to be seamlessly integrated on the same image coordinate basis, reducing the interference of linear road defects on the contour evolution of regional road defects, and ensuring that the coordinate results of both linear and regional road defects are uniformly labeled. Subsequently, the road defect detection results are generated from the content of the defect coordinate location marking on the original color image of the road surface.

[0031] A road defect detection system is used for detecting road surface defects during road construction. The system includes: The image preprocessing module acquires the original color image of the road surface, generates a two-dimensional grayscale matrix of the road surface from the original color image, and divides the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices. The suspected disease extraction module extracts the coordinates of temporary pixels for suspected diseases from the original color image of the road surface based on each candidate region matrix. The linear disease screening module determines the set of linear disease coordinates or the remaining two-dimensional grayscale matrix area based on the coordinates of the temporarily stored pixels of suspected diseases. The total difference calculation module constructs an initial circular closed contour of the disease at the center of the matrix region of the remaining two-dimensional gray matrix region, and calculates the total distribution difference that represents the difference in gray distribution of the road surface inside and outside the initial circular closed contour of the disease. The disease result integration module acquires and adjusts the disease boundary coordinates of each pixel on the initial circular closed contour of the disease, collects the total amount of different distribution differences during the adjustment process, selects the regional disease coordinate set based on the changes in the total amount of different distribution differences, and integrates the linear disease coordinate set and the regional disease coordinate set as the road disease detection result.

[0032] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for detecting road defects, used for detecting road surface defects during road construction, characterized in that, Includes the following steps: Acquire the original color image of the road surface, generate a two-dimensional grayscale matrix of the road surface from the original color image of the road surface, and divide the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices. Based on each candidate region matrix, extract the coordinates of suspected defects from the original color image of the road surface; Based on the coordinates of the temporarily stored pixels of suspected diseases, determine the set of coordinates of linear diseases or the remaining two-dimensional grayscale matrix area. Construct an initial circular closed contour of the disease at the center of the matrix region of the remaining two-dimensional gray matrix region, and calculate the total distribution difference that characterizes the difference in gray distribution of the road surface inside and outside the initial circular closed contour of the disease. The disease boundary coordinates of each pixel on the initial circular closed contour of the disease are obtained and adjusted. The total amount of different distribution differences during the adjustment process is collected. Based on the changes in the total amount of different distribution differences, a set of regional disease coordinates is selected. The set of linear disease coordinates and the set of regional disease coordinates are integrated as the road disease detection result.

2. The road defect detection method according to claim 1, characterized in that, The process of acquiring the original color image of the road surface, generating a two-dimensional grayscale matrix of the road surface from the original color image, and dividing the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices includes: During road inspection, the original color images of the road surface are continuously captured by digital cameras through the data transmission bus. The color channel values ​​of the red, green and blue color channels at the location of each pixel in the original color image of the road surface are calculated according to the preset fixed weight ratio coefficient of the color channels. The weighted sum of the color channels is used as the gray value of the image pixel in the original color image of the road surface. The horizontal and vertical two-dimensional coordinates of the image pixel in the original color image of the road surface are obtained and integrated with the corresponding image pixel gray value to construct a two-dimensional gray matrix of the road surface. The two-dimensional grayscale matrix of the road surface is divided into grids according to the preset fixed pixel length and width of the region to generate multiple local candidate region matrices.

3. The road defect detection method according to claim 2, characterized in that, The step of extracting the coordinates of suspected defects temporary pixels from the original color image of the road surface based on each candidate region matrix includes: Obtain a preset Radon transform angle traversal sequence starting from 0 degrees and ending at 180 degrees, as well as a fixed Radon transform angle increment interval step size and projection angle direction. Obtain the parallel straight line trajectory perpendicular to the projection angle direction within each candidate region matrix. Calculate the linear gray value accumulation result of all image pixel points along the parallel straight line trajectory in each projection angle direction, and use it as the parallel straight line projection value when the projection angle direction intersects with the parallel straight line trajectory. Select the first item from all the parallel line projection values ​​corresponding to each projection angle direction as the peak value of the angle projection corresponding to the projection angle direction. Obtain the horizontal and vertical two-dimensional coordinates of all image pixels on the parallel straight line trajectory corresponding to the peak value of the angle projection that is greater than the preset peak value threshold. Based on the horizontal and vertical two-dimensional coordinates of the image pixels, divide the coordinates of suspected defects into temporary pixel points on the original color image of the road surface.

4. The road defect detection method according to claim 3, characterized in that, The step of determining the set of linear disease coordinates or the remaining two-dimensional grayscale matrix region based on the coordinates of temporarily stored pixels of suspected diseases includes: Obtain the connected region of the suspected disease formed by the coordinates of the temporarily stored pixels of the suspected disease, and calculate the length and width of the connected region of the suspected disease, and at the same time calculate the aspect ratio of the length and width of the connected region. The image pixel points with horizontal and vertical two-dimensional coordinates corresponding to the aspect ratio values ​​of the connected regions that are greater than the preset aspect ratio threshold are filtered out and classified as linear road defects with directional stability and spatial continuity. The coordinate positions corresponding to the linear road defects are stored as a linear defect coordinate set. Extract the gray values ​​of the remaining image pixels corresponding to the horizontal and vertical two-dimensional coordinate positions of the remaining image pixels that were not identified as having linear road defects, and obtain the remaining two-dimensional gray matrix region.

5. The road defect detection method according to claim 4, characterized in that, The initial circular closed contour of the disease is constructed at the center of the matrix region in the remaining two-dimensional gray-level matrix region. The total distribution difference representing the difference in gray-level distribution inside and outside the initial circular closed contour of the disease includes: Based on the Chan-Vese level set algorithm, an initial circular closed contour of the disease is constructed at the center of the matrix region of the remaining two-dimensional gray matrix region. All image pixels in the initial circular closed contour of the disease are obtained, and the average gray value of all image pixels is calculated as the average value inside the contour. At the same time, the sum of squares of the internal gray value deviation between the gray value of each image pixel inside the initial circular closed contour of the disease and the average value inside the contour is calculated as the internal difference. Calculate the average gray value of all image pixels outside the initial circular closed contour of the disease as the outer average value of the contour. At the same time, calculate the sum of squares of the outer gray value deviation between each image pixel outside the initial circular closed contour of the disease and the outer average value of the contour as the outer difference. Calculate the sum of the internal difference and the external difference to obtain the total distribution difference that characterizes the difference in grayscale distribution of the road surface inside and outside the initial circular closed contour of the defect.

6. The road defect detection method according to claim 5, characterized in that, The process of acquiring and adjusting the boundary coordinates of each pixel on the initial circular closed contour of the disease, collecting the total amount of different distribution differences during the adjustment process, selecting a set of regional disease coordinates based on the changes in the total amount of different distribution differences, and integrating the set of linear disease coordinates and the set of regional disease coordinates as the road disease detection result includes: Record the disease boundary coordinates of all pixels on the initial circular closed contour of the disease as the initial evolution state of the contour. Adjust the disease boundary coordinates of each pixel on the initial circular closed contour of the disease to perform contour evolution on the initial evolution state of the contour, and obtain the evolved disease boundary coordinates. Based on the evolved disease boundary coordinates, redivide the initial circular closed contour of the disease and repeat the calculation process of the total distribution difference. Calculate the total absolute difference of the total distribution difference obtained in two adjacent contour evolution states. Obtain the set of pixel coordinates of the closed geometric contour when the absolute difference of the total amount is less than the preset tolerance threshold for the difference change, and generate a set of coordinates of regional diseases; The location of road defects is marked on the original color image of the road surface according to the set of linear defect coordinates and the set of regional defect coordinates, and the road defect detection results are generated.

7. The road defect detection method according to claim 2, characterized in that, The process for setting the fixed pixel dimensions of the region is as follows: Obtain the length and width dimensions of the single-pixel physical map of the road surface from the digital camera, as well as the length and width dimensions of the physical coverage of road defects. Calculate the ratio of the physical coverage dimensions of road defects to the physical mapping dimensions of a single pixel road surface, and configure it as a preset fixed pixel dimension for the region.

8. The road defect detection method according to claim 4, characterized in that, The process for obtaining the suspected diseased connected region is as follows: Calculate the Euclidean distance between the coordinates of any two temporarily stored pixels suspected of being defects; Extract the coordinates of multiple suspected disease temporary pixels whose Euclidean distance is less than or equal to a preset connectivity search distance threshold, and establish a connected region for the suspected disease.

9. The road defect detection method according to claim 6, characterized in that, The contour evolution includes: Extract the geometric normal evolution direction of the disease boundary coordinates corresponding to each pixel on the initial circular closed contour of the disease; Based on the numerical descent gradient of the total distribution difference, calculate the position translation step of the disease boundary coordinates in the geometric normal evolution direction; The disease boundary coordinates are translated along the geometric normal evolution direction by a translation step size to obtain the evolved disease boundary coordinates.

10. A road defect detection system for detecting defects in road surface during road construction, characterized in that, The system includes: The image preprocessing module acquires the original color image of the road surface, generates a two-dimensional grayscale matrix of the road surface from the original color image, and divides the two-dimensional grayscale matrix of the road surface into multiple local candidate region matrices. The suspected disease extraction module extracts the coordinates of temporary pixels for suspected diseases from the original color image of the road surface based on each candidate region matrix. The linear disease screening module determines the set of linear disease coordinates or the remaining two-dimensional grayscale matrix area based on the coordinates of the temporarily stored pixels of suspected diseases. The total difference calculation module constructs an initial circular closed contour of the disease at the center of the matrix region of the remaining two-dimensional gray matrix region, and calculates the total distribution difference that represents the difference in gray distribution of the road surface inside and outside the initial circular closed contour of the disease. The disease result integration module acquires and adjusts the disease boundary coordinates of each pixel on the initial circular closed contour of the disease, collects the total amount of different distribution differences during the adjustment process, selects the regional disease coordinate set based on the changes in the total amount of different distribution differences, and integrates the linear disease coordinate set and the regional disease coordinate set as the road disease detection result.

Images