A method and system for rapid extraction of pavement disease point clouds

Abstract

The present invention relates to the technical field of rapid extraction of pavement defect point clouds, and specifically to a system and method for rapid extraction of pavement defect point clouds. By converting three-dimensional pavement point cloud data into a two-dimensional projection image represented by grayscale values, the processing difficulty of the three-dimensional pavement point cloud data is reduced. The plane formed by each pixel in the two-dimensional projection image selects the space that meets the disease-free pavement state as the disease-free pavement space, thereby intuitively screening the pixel points that do not meet the disease-free pavement state. The recognition method is simple and has high accuracy. At the same time, in order to further improve the recognition efficiency of the diseased points and avoid repeated judgment of the recognized pixels, the two-dimensional projection image is divided into multiple windows and the diseased points in the windows are judged and collected in turn. This can not only judge the area of the diseased points, but also ensure the accuracy and efficiency of recognition, thereby meeting actual use needs.

Description

Technical Field

[0001] The present invention relates to the technical field of rapid extraction of pavement defect point clouds, and in particular to a system and method for rapid extraction of pavement defect point clouds. Background Art

[0002] As one of the most important and common forms of transportation for integrating surrounding exchanges and promoting regional integration, the scale of road transportation network construction has been continuously expanded and improved in recent years.

[0003] Asphalt roads are currently a widely used form of road. As their service life increases, different forms of diseases will inevitably appear. Among the many types of diseases, pavement diseases are more common and more harmful. They not only seriously affect the appearance of the road and reduce driving comfort and safety, but may also develop further and affect the overall structural stability of the road.

[0004] Traditional manual inspection methods for pavement defects consume a lot of manpower, material and financial resources. Obtaining pavement defect data is difficult, inefficient, limited in accuracy, highly dangerous, and has poor reproducibility. Road closures may also be required, hindering normal road operations.

[0005] In recent years, a large number of studies have used digital image processing technology to identify asphalt pavement defects using two-dimensional road surface images. However, the height characteristics of pavement defects cannot be accurately extracted from two-dimensional digital images. The use of three-dimensional point cloud equipment can obtain millimeter-level elevation information of the road surface and reconstruct the three-dimensional shape of the road surface. Although this can improve the accuracy of pavement defect identification, the current common method of extracting pavement defect information from three-dimensional point clouds has the following main shortcomings:

[0006] (1) The defect extraction method based on two-dimensional image conversion converts point cloud data into two-dimensional images, and extracts defects by segmenting the two-dimensional images. Compared with the traditional method of using only two-dimensional road surface images, this method improves the accuracy of defect extraction, but it cannot eliminate the interference of texture and noise.

[0007] (2) The defect information extraction method based on mathematical morphology uses data such as height difference, normal vector, gradient, and second-order derivative to extract possible defect points that meet the mathematical characteristics of the defect, and generates the defect area through morphological methods. However, this method is mainly suitable for identifying pavement cracks and has a weak ability to extract fine features for defects with a certain area, such as potholes.

[0008] (3) A pavement disease information extraction method based on high-level learning and neural networks utilizes relevant neural network technologies and builds a network based on the characteristics of pavement diseases and point clouds. However, this method requires manual labeling at the pixel level for training, which requires a certain amount of time and labor costs.

[0009] Under the existing methods, problems such as low road surface disease identification accuracy, low intelligence level, and low versatility still exist. Summary of the Invention

[0010] The purpose of the present invention is to provide a system and method for quickly extracting pavement defect point clouds to solve the current problem of low efficiency in obtaining pavement defect information.

[0011] The present invention solves the above-mentioned technical problems by:

[0012] A method for quickly extracting pavement defect point clouds, characterized by comprising the following steps:

[0013] S1. Obtaining the 3D point cloud data of the road surface to be inspected and cropping it to obtain a single point cloud data;

[0014] S2. linearly mapping the acquired single point cloud data to obtain a two-dimensional projection image of the point cloud represented by grayscale values;

[0015] S3, using the pixels of the two-dimensional projection image of the point cloud to perform plane fitting to obtain a disease-free road surface space that meets the disease-free road surface state;

[0016] S4. Divide the point cloud two-dimensional projection image into multiple windows, use the non-defective road surface space as a reference, and define the pixel points in the window that are outside the non-defective road surface space as defect points. Collect the defect points in all windows to obtain a defect point set.

[0017] S5. Calculate the point cloud height values corresponding to all the defect points in the defect point set according to the grayscale values of the defect points, and combine all the point cloud height values with the pavement point cloud three-dimensional data corresponding to all the defect points to obtain the pavement defect three-dimensional point cloud data.

[0018] It is further defined that step S2 is specifically as follows:

[0019] Linearly map the height value of each point cloud in a single point cloud data through the grayscale value to obtain the grayscale value corresponding to each point cloud:

[0020]

[0021] Among them, p ij is the grayscale value corresponding to the point cloud in the i-th row and j-th column of a single point cloud 3D data, a ij is the point cloud height value of the i-th row and j-th column, a max is the maximum height in a single point cloud data, a min It is the minimum height value in a single point cloud data;

[0022] The grayscale value corresponding to each point cloud is used as the grayscale value of the corresponding pixel, and the pixels obtained by mapping all point clouds form a two-dimensional projection image of the point cloud.

[0023] It is further defined that step S3 includes the following steps:

[0024] S31, select any three pixel points in the point cloud two-dimensional projection image to perform plane fitting to obtain the fitting plane P σ :

[0025] A σ x+B σ y+C σ z+D σ =0

[0026] Among them, P σ is the σth fitting plane;

[0027] S32, continue to execute step S31 until any pixel point in the point cloud two-dimensional projection image and any two pixels in the point cloud two-dimensional projection image have been plane-fitted, and then execute step S33;

[0028] S33, calculate each pixel point (x n ,y n ,z n ) to the fitting plane Pσ

[0029]

[0030] in, is the distance between the nth pixel point in the two-dimensional projection image of the point cloud and the σth fitting plane;

[0031] S34, set the fitting plane P σ The height of ε above and the fitting plane P σ The space between the heights ε below is the fitting plane P σ The spatial domain of , where ε is calculated as:

[0032]

[0033] Among them, 1≤b≤3;

[0034] S35, count each pixel point in the two-dimensional projection image of the point cloud to the fitting plane P σ The distance meets The number of pixels α σ , change α σ as the score of the σth fitting plane;

[0035] S36. Select the fitting plane with the highest score as the optimal pavement fitting plane, and use the spatial domain of the optimal pavement fitting plane as the disease-free pavement space.

[0036] It is further defined that step S4 includes the following steps:

[0037] S41, dividing the point cloud two-dimensional projection image into multiple windows of the same size;

[0038] S42, determining whether each pixel in the window is within a non-defective road surface space in turn;

[0039] S43, taking pixel points outside the non-defective road surface space as defective points;

[0040] S44. Collect the defect points in each window to obtain a defect point set.

[0041] It is further defined that step S5 includes the following steps:

[0042] S51, selecting a defect point from the defect point set, and calculating the point cloud height value of the defect point according to the gray value of the defect point;

[0043] S52, combining the obtained point cloud height value with the two-dimensional coordinates of the defect point to express it in three-dimensional coordinates;

[0044] S53, looping through steps S51 and S52 until the calculation of the point cloud height values of all the defect points in the defect point set is completed, and all the defect points are represented in the same three-dimensional coordinate to obtain the pavement defect three-dimensional point cloud data.

[0045] It is further defined that the point cloud height value of the defect point is calculated by the gray value of the defect point as follows:

[0046]

[0047] Among them, α is the αth disease point, N α is the point cloud height value of the αth defect point.

[0048] A pavement defect point cloud rapid extraction system, used to implement the method according to any one of claims 1 to 6, characterized in that the pavement defect point cloud rapid extraction system comprises:

[0049] A road surface point cloud acquisition device, used to acquire three-dimensional road surface point cloud data of the road surface to be inspected;

[0050] The point cloud data processing unit is used to linearly map the height value of each point cloud in a single point cloud data through the grayscale value to obtain the grayscale value corresponding to each point cloud, and use the grayscale value corresponding to each point cloud as the grayscale value of the corresponding pixel to obtain a two-dimensional projection image of the point cloud;

[0051] A space fitting unit is used to perform plane fitting in the two-dimensional projection image of the point cloud to obtain a disease-free road surface space;

[0052] The defect point processing unit is used to divide the two-dimensional projection image of the point cloud into multiple windows, take the defect-free road surface space as a reference, and take the pixel points in the window located outside the defect-free road surface space as defect points. It is used to calculate the point cloud height values corresponding to all defect points in the defect point set based on the grayscale value of the defect point, and combine all point cloud height values with the three-dimensional road surface point cloud data corresponding to all defect points to obtain the three-dimensional point cloud data of road surface defects.

[0053] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above method when executing the computer program.

[0054] A computer-readable storage medium is characterized in that the computer-readable storage medium stores a computer program, and the computer program implements the above method when executed by a processor.

[0055] A computer program product, characterized in that the computer program product comprises a computer program, and the computer program implements the above method when executed by a processor.

[0056] The beneficial effects of the present invention are:

[0057] 1. The present invention converts the single point cloud data obtained after cropping the three-dimensional road surface electric cloud data into a two-dimensional projection image represented by grayscale values, thereby reducing the difficulty of processing the three-dimensional road surface point cloud data. The plane formed by each pixel in the two-dimensional projection image selects the space that meets the disease-free road surface state as the disease-free road surface space, so that the pixel points that do not meet the disease-free road surface state can be intuitively screened, that is, the pixel points in the disease-free road surface space in the two-dimensional projection image are disease points, and the texture and noise interference are eliminated. The recognition method is simple and accurate. At the same time, in order to further improve the recognition efficiency of disease points and avoid repeated judgment of recognized pixels, the two-dimensional projection image is divided into multiple windows and the disease points in the windows are judged and collected in turn. It can not only judge the area of the disease point, but also ensure the accuracy and efficiency of recognition, and meet actual use needs.

[0058] 2. The present invention processes the three-dimensional point cloud data of the road surface by combining a plane fitting algorithm with window scanning, thereby enhancing the refined feature extraction of such defects, and is used for the rapid and accurate identification and extraction of point cloud data of road surface defects (potholes, cracks, etc.). This method can still effectively extract the point cloud of road surface defects when the road surface has a certain slope.

[0059] 3. This invention uses three 2D pixels containing 3D information to perform plane fitting. Using the distance between the pixels and the point cloud's 2D projection image as a scoring criterion, it screens for the optimal pavement-fitting plane that corresponds to a defect-free pavement. The space above and below the optimal pavement-fitting plane is then used as the defect-free pavement space. This allows for more accurate and efficient screening of defect points, effectively reducing noise during point cloud extraction of pavement defects and improving the accuracy of point cloud extraction. This method is simple and easy to implement, significantly improving the efficiency and accuracy of pavement defect identification and extraction algorithms based on 3D point clouds. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a flow chart of the method for rapid extraction of pavement defect point clouds according to the present invention;

[0061] Figure 2 This is a schematic diagram of a single point cloud data of the road surface to be inspected according to the present invention;

[0062] Figure 3 This is a schematic diagram of a two-dimensional projection image of a point cloud according to the present invention;

[0063] Figure 4 Schematic diagram of the window division of the point cloud two-dimensional projection image of the present invention;

[0064] Figure 5 This is a schematic diagram of the three-dimensional point cloud data of pavement defects in the present invention. DETAILED DESCRIPTION

[0065] Example 1

[0066] refer to Figure 1 The present invention provides a method for quickly extracting pavement defect point clouds, comprising the following steps:

[0067] S1. Obtaining the 3D point cloud data of the road surface to be inspected and cropping it to obtain a single point cloud data;

[0068] S2. linearly mapping the acquired single point cloud data to obtain a two-dimensional projection image of the point cloud represented by grayscale values;

[0069] S3, using the pixels of the two-dimensional projection image of the point cloud to perform plane fitting to obtain a disease-free road surface space that meets the disease-free road surface state;

[0070] S4. Divide the point cloud two-dimensional projection image into multiple windows, use the non-defective road surface space as a reference, and define the pixel points in the window that are outside the non-defective road surface space as defect points. Collect the defect points in all windows to obtain a defect point set.

[0071] S5. Calculate the point cloud height values corresponding to all the defect points in the defect point set according to the grayscale values of the defect points, and combine all the point cloud height values with the pavement point cloud three-dimensional data corresponding to all the defect points to obtain the pavement defect three-dimensional point cloud data.

[0072] Among them, reference Figure 2 In step S1, the road surface to be inspected can be scanned by existing equipment to obtain the road surface point cloud three-dimensional data of the road surface. The road surface point cloud data is cropped according to the usage requirements to obtain single point cloud data. The single point cloud data contains multiple independent point clouds in three-dimensional coordinates. The data volume is large and the processing difficulty is high, so subsequent processing is required.

[0073] Among them, reference Figure 3 , step S2 is specifically as follows:

[0074] Linearly map the height value of each point cloud in a single point cloud data through the grayscale value to obtain the grayscale value corresponding to each point cloud:

[0075]

[0076] Among them, p ij is the grayscale value corresponding to the point cloud in the i-th row and j-th column of a single point cloud data, a ij is the point cloud height value of the i-th row and j-th column, a max is the maximum height in a single point cloud data, a min It is the minimum height value in a single point cloud data;

[0077] The grayscale value corresponding to each point cloud is used as the grayscale value of the pixel corresponding to each point cloud, and the pixels obtained by mapping all point clouds form a two-dimensional projection image of the point cloud as a whole.

[0078] Specifically, each point cloud is linearly mapped using the above formula, so that the height of each electric cloud corresponds to the grayscale value of a pixel, so that the distribution of each electric cloud is used as the distribution of the corresponding pixel, and the height of each electric cloud is displayed in grayscale, so that a grayscale image containing a single point cloud data can be obtained. The grayscale image is used as the two-dimensional projection image of the point cloud. The two-dimensional coordinates of each electric cloud are the same as the corresponding pixel plane coordinates. Therefore, the pixels in the two-dimensional projection image of the point cloud are not arranged closely in sequence. In this way, complex three-dimensional data can be converted into a simple and easy-to-process two-dimensional grayscale image, reducing the difficulty of data processing.

[0079] Wherein, step S3 includes the following steps:

[0080] S31, select any three pixel points in the point cloud two-dimensional projection image to perform plane fitting to obtain the fitting plane P σ :

[0081] A σ x+Bσ y+C σ z+D σ =0

[0082] Among them, P σ is the σth fitting plane;

[0083] Since each pixel in the point cloud two-dimensional projection image corresponds to the three-dimensional coordinates of a point cloud, any three pixels in the point cloud two-dimensional projection image are selected for plane fitting, so that the first fitting plane P1 can be fitted. The grayscale value of the pixel can be used to achieve plane fitting in the vertical direction. Then, three pixels that have not been selected at the same time are reselected and plane fitting is performed to obtain the second fitting plane P2. This cycle is repeated until any three pixels in the point cloud two-dimensional projection image complete the plane fitting to obtain the corresponding fitting plane P σ , P σ is the σth fitting plane, σ is a positive integer; for example, there are 5 pixel points a to e in the point cloud 2D projection image. Since the fitting planes obtained by selecting three points abc or bca are the same, in order to improve efficiency, the three identical pixel points are selected as one. At this time, any three pixel points may be any of abc, abd, abe, acd, ace, ade, bcd, bce or cde.

[0084] S32, continue to execute step S31 until any pixel point in the point cloud two-dimensional projection image and any two pixels in the point cloud two-dimensional projection image have been plane-fitted, and then execute step S33;

[0085] According to the method of step S31, any three pixel points in the two-dimensional projection image of the point cloud are fitted to finally obtain multiple non-repeated fitting planes, and then step S33 is executed.

[0086] S33, calculate each pixel point (x n ,y n ,z n ) to the fitting plane P σ distance

[0087]

[0088] in, is the distance between the nth pixel point in the two-dimensional projection image of the point cloud and the σth fitting plane;

[0089] Specifically, after each fitting to obtain a new fitting plane, the above formula is used to calculate the distance from all pixels in the point cloud two-dimensional projection image to the fitting plane P σ distance For example, the distance from the third pixel to the first fitting plane in the point cloud 2D projection image The cycle continues in this way. After a fitting plane is obtained, the distances from all pixels to the plane are calculated. It should be noted that each pixel is regarded as the nth pixel, which does not limit the order of pixel calculation. It is mainly to illustrate that each pixel needs to participate in the distance calculation. The pixels at different positions are not fixed in the distance calculation with different fitting planes. However, in order to facilitate calculation and reduce processing difficulty, it is preferred to mark the pixels in the two-dimensional projection image of the point cloud in sequence, that is, the pixels in the two-dimensional projection image of the point cloud participate in the distance calculation in a zigzag order.

[0090] S34, set the fitting plane P σ The height of ε above and the fitting plane P σ The space between the heights ε below is the fitting plane P σ The spatial domain of , where ε is calculated as:

[0091]

[0092] Wherein, 1≤b≤3, preferably b=3;

[0093] Usually, the road surface is uneven during construction, so in order to simulate the real road surface conditions, the spatial domain of the fitting plane is set to obtain the space of the actual normal road surface.

[0094] S35, count each pixel point in the two-dimensional projection image of the point cloud to the fitting plane P σ The distance meets The number of pixels α σ , change α σ as the score of the σth fitting plane;

[0095] During the calculation, for example, the first fitting plane is selected, that is, σ = 1. At this time, the distance from each pixel in the point cloud two-dimensional projection image to the fitting plane P1 is calculated, and the distance from each pixel to the fitting plane P1 is obtained. At this time, the positive and negative values of the calculated distance are only used for direction determination, and the distance values are all positive. Then, the distance from each pixel to the fitting plane P1 and the value of ε are judged, and the number of pixels α whose distance value from the pixel to the fitting plane P1 is less than ε is selected. σ , for example, there are 10 pixels whose distance to the fitting plane P1 is less than ε, then α σ=10, the score of the fitting plane P1 is 10 points. After this is completed, the fitting plane with σ=2 is selected to calculate the score. This cycle is repeated until the score calculation of all fitting planes is completed to obtain the score of each fitting plane. If the distance between the pixel in the two-dimensional projection image of the point cloud and the fitting plane is too far, it is considered that the point cloud corresponding to the pixel is abnormal. Therefore, the pixel points within the set spatial domain are considered to belong to the normal range of normal road surface. The pixel points located in and outside the spatial domain are considered to belong to a row of abnormal points on the abnormal road surface, which are points with defects.

[0096] S36, selecting the fitting plane with the highest score as the optimal pavement fitting plane, and using the spatial domain of the optimal pavement fitting plane as the disease-free pavement space;

[0097] Since road surface defects inspection often requires fewer defect points than those that cannot be observed visually, the number of normal points in actual use is far greater than that of defect points. Therefore, in order to improve the accuracy of defect extraction and eliminate the interference of texture noise, the fitting plane with the highest score is selected as the best fitting plane, and the spatial domain of the best fitting plane is used as the defect-free road surface space, thereby obtaining a road surface closest to the actual situation. The scoring mechanism can also effectively fit road conditions with a certain slope, thereby realizing road surface defect identification when there is a slope.

[0098] Step S4 includes the following steps:

[0099] S41, dividing the point cloud two-dimensional projection image into multiple windows of the same size;

[0100] S42, determining whether each pixel in the window is within a non-defective road surface space in turn;

[0101] S43, taking pixel points outside the non-defective road surface space as defective points;

[0102] S44. Collect the defect points in each window to obtain a defect point set.

[0103] Specifically, refer to Figure 4Since there are a large number of pixels in each point cloud two-dimensional projection image, in order to avoid repeated calculation of pixel points, the point cloud two-dimensional projection image is divided into multiple windows of the same size, and then all pixel points in each window are judged to be damage points in turn. The basis for judging whether it is a damage point is whether the pixel point is located outside the damage-free road surface space. If it is located outside the damage-free road surface space, it means that the point cloud corresponding to the pixel point has a deeper or higher crack or bulge, so the pixel point is regarded as a damage point, and the point cloud two-dimensional projection image is divided into multiple windows of the same size, so that in subsequent calculations, calculations can be performed in sequence according to the order of the windows to avoid omissions; at the same time, the recognition ability of defects with a certain area such as potholes is enhanced; then all defect points are collected to obtain a defect point set.

[0104] Step S5 includes the following steps:

[0105] S51, selecting a defect point from the defect point set, and calculating the point cloud height value of the defect point according to the gray value of the defect point;

[0106] S52, combining the obtained point cloud height value with the two-dimensional coordinates of the defect point to express it in three-dimensional coordinates;

[0107] S53, looping through steps S51 and S52 until the calculation of the point cloud height values of all the defect points in the defect point set is completed, and all the defect points are represented in the same three-dimensional coordinate to obtain the pavement defect three-dimensional point cloud data.

[0108] The point cloud height value of the defect point is calculated by the gray value of the defect point as follows:

[0109]

[0110] Among them, α is the αth disease point, N α is the point cloud height value of the αth defect point.

[0111] Specifically, refer to Figure 5 , the above formula is used to convert the pixels corresponding to the diseased points into the height values of the point cloud, so that the diseased points can be quickly represented in a three-dimensional coordinate and the three-dimensional point cloud data of the road surface disease can be obtained. This method is simple to calculate and has high accuracy, which can meet the actual use needs.

[0112] Example 2

[0113] This embodiment further provides a system for rapidly extracting road surface defect point clouds, which is used to implement the method described in Example 1, including:

[0114] A road surface point cloud acquisition device, used to acquire three-dimensional road surface point cloud data of the road surface to be inspected;

[0115] The point cloud data processing unit is used to linearly map the height value of each point cloud in a single point cloud data through the grayscale value to obtain the grayscale value corresponding to each point cloud, and use the grayscale value corresponding to each point cloud as the grayscale value of the corresponding pixel to obtain a two-dimensional projection image of the point cloud;

[0116] A space fitting unit is used to perform plane fitting in the two-dimensional projection image of the point cloud to obtain a disease-free road surface space;

[0117] The defect point processing unit is used to divide the two-dimensional projection image of the point cloud into multiple windows, take the defect-free road surface space as a reference, and take the pixel points in the window located outside the defect-free road surface space as defect points. It is used to calculate the point cloud height values corresponding to all defect points in the defect point set based on the grayscale value of the defect point, and combine all point cloud height values with the three-dimensional road surface point cloud data corresponding to all defect points to obtain the three-dimensional point cloud data of road surface defects.

[0118] This embodiment further provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in Embodiment 1 when executing the computer program.

[0119] This embodiment further provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in Example 1 is implemented.

[0120] This embodiment further provides a computer program product, characterized in that the computer program product includes a computer program, and when the computer program is executed by a processor, the method described in Example 1 is implemented.

Claims

1. A method for rapid extraction of pavement defect point clouds, characterized in that: The following steps are involved: S1. Obtaining the 3D point cloud data of the road surface to be inspected and cropping it to obtain a single point cloud data; S2. linearly mapping the acquired single point cloud data to obtain a two-dimensional projection image of the point cloud represented by grayscale values; S3, using the pixels of the two-dimensional projection image of the point cloud to perform plane fitting to obtain a disease-free road surface space that meets the disease-free road surface state; The step S3 comprises the following steps: S31, select any three pixel points in the point cloud two-dimensional projection image to perform plane fitting to obtain the fitting plane P σ : A σ x+B σ y+C σ z+D σ =0 Among them, P σ is the σth fitting plane; x and y are the coordinate values of any pixel in the two-dimensional projection image of the point cloud; A σ and B σ is the fitting coefficient corresponding to the fitting straight line equation in the two-dimensional projection image of the point cloud; D σ is the constant term corresponding to the fitted straight line equation in the two-dimensional projection image of the point cloud; z is the height value of the point cloud; C σ is the fitting coefficient of the point cloud height direction; S32, continue to execute step S31 until any pixel point in the point cloud two-dimensional projection image and any two pixels in the point cloud two-dimensional projection image have been plane-fitted, and then execute step S33; S33, calculate each pixel point (x n ,y n ,z n ) to the fitting plane P σ distance in, is the distance between the nth pixel point in the two-dimensional projection image of the point cloud and the σth fitting plane; S34, set the fitting plane P σ The height of ε above and the fitting plane P σ The space between the heights ε below is the fitting plane P σ The spatial domain of , where ε is calculated as: Among them, 1≤b≤3; S35, count each pixel point in the two-dimensional projection image of the point cloud to the fitting plane P σ The distance meets The number of pixels α σ , change α σ as the score of the σth fitting plane; S36, selecting the fitting plane with the highest score as the optimal pavement fitting plane, and using the spatial domain of the optimal pavement fitting plane as the disease-free pavement space; S4. Divide the point cloud two-dimensional projection image into multiple windows, use the non-defective road surface space as a reference, and define the pixel points in the window that are outside the non-defective road surface space as defect points. Collect the defect points in all windows to obtain a defect point set. S5. Calculate the point cloud height values corresponding to all the defect points in the defect point set according to the grayscale values of the defect points, and combine all the point cloud height values with the pavement point cloud three-dimensional data corresponding to all the defect points to obtain the pavement defect three-dimensional point cloud data.

2. The method for rapid extraction of pavement defect point clouds according to claim 1, characterized in that: The step S2 is specifically as follows: Linearly map the height value of each point cloud in a single point cloud data through the grayscale value to obtain the grayscale value corresponding to each point cloud: Among them, p ij is the grayscale value corresponding to the point cloud in the i-th row and j-th column of a single point cloud 3D data, a ij is the point cloud height value of the i-th row and j-th column, a max is the maximum height in a single point cloud data, a min It is the minimum height value in a single point cloud data; The grayscale value corresponding to each point cloud is used as the grayscale value of the corresponding pixel, and the pixels obtained by mapping all point clouds form a two-dimensional projection image of the point cloud.

3. The method for rapid extraction of pavement defect point clouds according to claim 2, characterized in that: The step S4 comprises the following steps: S41, dividing the point cloud two-dimensional projection image into multiple windows of the same size; S42, determining whether each pixel in the window is within a non-defective road surface space in turn; S43, taking pixel points outside the non-defective road surface space as defective points; S44. Collect the defect points in each window to obtain a defect point set.

4. The method for rapid extraction of pavement defect point clouds according to claim 3, characterized in that: The step S5 comprises the following steps: S51, selecting a defect point from the defect point set, and calculating the point cloud height value of the defect point according to the gray value of the defect point; S52, combining the obtained point cloud height value with the two-dimensional coordinates of the defect point to express it in three-dimensional coordinates; S53, looping through steps S51 and S52 until the calculation of the point cloud height values of all the defect points in the defect point set is completed, and all the defect points are represented in the same three-dimensional coordinate to obtain the pavement defect three-dimensional point cloud data.

5. The method for rapid extraction of pavement defect point clouds according to claim 4, characterized in that: The point cloud height value of the defect point is calculated by the gray value of the defect point as follows: Among them, α is the αth disease point, N α is the point cloud height value of the αth defect point.

6. A rapid extraction system for pavement defect point clouds, used to implement the method according to any one of claims 1 to 5, characterized in that: The pavement damage point cloud rapid extraction system includes: A road surface point cloud acquisition device, used to acquire three-dimensional road surface point cloud data of the road surface to be inspected; The point cloud data processing unit is used to linearly map the height value of each point cloud in a single point cloud data through the grayscale value to obtain the grayscale value corresponding to each point cloud, and use the grayscale value corresponding to each point cloud as the grayscale value of the corresponding pixel to obtain a two-dimensional projection image of the point cloud; A space fitting unit is used to perform plane fitting in the two-dimensional projection image of the point cloud to obtain a disease-free road surface space; The spatial fitting unit is specifically: Select any three pixel points in the point cloud two-dimensional projection image to perform plane fitting to obtain the fitting plane P σ : A σx +B σ y+C σ z+D σ =0 Among them, P σ is the σth fitting plane; x and y are the coordinate values of any pixel in the two-dimensional projection image of the point cloud; A σ and B σ is the fitting coefficient corresponding to the fitting straight line equation in the two-dimensional projection image of the point cloud; D σ is the constant term corresponding to the fitted straight line equation in the two-dimensional projection image of the point cloud; z is the height value of the point cloud; C σ is the fitting coefficient of the point cloud height direction; Continue to select any three pixel points in the point cloud two-dimensional projection image to perform plane fitting to obtain the fitting plane P σ , until any pixel point in the point cloud two-dimensional projection image and any two pixels in the point cloud two-dimensional projection image are plane-fitted; Calculate each pixel (x n ,y n ,z n ) to the fitting plane P σ distance in, is the distance between the nth pixel point in the two-dimensional projection image of the point cloud and the σth fitting plane; Set the fitting plane P σ The height of ε above and the fitting plane P σ The space between the heights ε below is the fitting plane P σ The spatial domain of , where ε is calculated as: Among them, 1≤b≤3; Statistical point cloud 2D projection image of each pixel to the fitting plane P σ The distance meets The number of pixels α σ , change α σ as the score of the σth fitting plane; The fitting plane with the highest score is selected as the optimal pavement fitting plane, and the spatial domain of the optimal pavement fitting plane is used as the disease-free pavement space; The defect point processing unit is used to divide the two-dimensional projection image of the point cloud into multiple windows, take the defect-free road surface space as a reference, and take the pixel points in the window located outside the defect-free road surface space as defect points. It is used to calculate the point cloud height values corresponding to all defect points in the defect point set based on the grayscale value of the defect point, and combine all point cloud height values with the three-dimensional road surface point cloud data corresponding to all defect points to obtain the three-dimensional point cloud data of road surface defects.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

9. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images