Pothole extraction method from vehicle-borne road point cloud based on light intensity and statistical characteristics

Abstract

The present invention discloses a method for extracting potholes from a vehicle-mounted road surface point cloud based on illumination intensity and statistical features, comprising the following steps: step 1, using a time index attribute to segment the vehicle-mounted road surface point cloud into equal time intervals; step 2, constructing a hemispherical illumination model, and calculating the illumination intensity of the point cloud, retaining the point cloud with illumination intensity within a preset illumination intensity threshold range as the initial pothole point cloud; step 3, clustering and partitioning the initial pothole point cloud through DBSCAN to obtain a pothole cluster point cloud; step 4, performing height histogram statistics on each pothole cluster point cloud, and screening and retaining the pothole point cloud according to the height histogram statistical features. The method of the present invention is simple to operate, highly feasible, and greatly improves the efficiency and accuracy of automatically extracting pothole points.

Description

Technical Field

[0001] The invention relates to the technical field of vehicle-mounted laser scanning point cloud data processing, and in particular to a method for extracting potholes in a vehicle-mounted road surface point cloud based on light intensity and statistical characteristics. Background Art

[0002] Road traffic occupies an important position in the national economy and people's lives, and has greatly promoted the development of the national economy. Road information extraction and mining has always been one of the key technologies in the fields of vehicle intelligent driving, traffic management and road 3D modeling. Potholes in asphalt pavements not only directly affect driving comfort and safety, but also seriously affect the service life of the pavement. Fast and accurate pothole detection results are the basis for scientific evaluation of pavement damage conditions and an important basis for determining maintenance measures and timing. At present, pothole detection is generally performed by technicians who manually depict the pothole range in the pavement image based on experience or disease manuals. This method has low recognition efficiency and can only obtain the area index of the pothole, but cannot accurately grasp multi-dimensional information such as depth or volume.

[0003] As an advanced measurement method, the vehicle-mounted laser mobile measurement system has the characteristics of fast, non-contact with the measured object, real-time, dynamic, active, high density and high precision. It can quickly obtain road point clouds in complex environments without affecting traffic, providing a reliable data source for extracting high-precision and rich three-dimensional road information. At present, the processing of vehicle-mounted laser scanning point cloud data mainly focuses on point cloud data classification, line extraction and building facade information extraction; there is less research on the automatic extraction of road pothole point clouds, which mainly focuses on indoor simulation of potholes, or pothole identification by height, height difference, relative distance, etc.

[0004] Although the above methods have achieved relatively good results to a certain extent, they are limited by ideal or model conditions and do not fully utilize the spatial morphological characteristics of the pit point cloud and the mathematical and statistical characteristics of the surrounding neighborhood point cloud, and have not yet formed a mature automatic extraction method for pit point clouds. It can be seen that the existing technology needs to be further improved. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a method for extracting potholes in a vehicle-mounted road surface point cloud based on light intensity and statistical characteristics. The method of the present invention is simple to operate, has high feasibility, and greatly improves the efficiency and accuracy of automatically extracting pothole points.

[0006] The present invention adopts the following technical solutions to solve the above technical problems:

[0007] A method for extracting potholes from a vehicle-mounted road surface point cloud based on light intensity and statistical characteristics is proposed in the present invention, comprising the following steps:

[0008] Step 1: Use the time index attribute to segment the vehicle-borne road surface point cloud into equal time intervals;

[0009] Step 2: Construct a hemispherical illumination model and calculate the illumination intensity of the point cloud. The point cloud with illumination intensity within the preset illumination intensity threshold range is retained as the initial pit point cloud. The specific steps are as follows:

[0010] (2.1) The hemispherical illumination model is: assume that a hemispherical light source is placed at the pth point in the vehicle-mounted road point cloud. The hemispherical light source emits N rays uniformly upward and outward. These N rays constitute the illumination vector set V of the pth point. N ; Where P≥p≥1, P is the total number of all points in the vehicle-mounted road point cloud;

[0011] (2.2) Traverse all points in the point cloud until the illumination intensity of all points is calculated, where the illumination intensity I of the pth point is calculated. P ; The details are as follows:

[0012] Take the pth point as the starting point and the other points in the vehicle-mounted road surface point cloud except the pth point as the end point to construct vectors. These vectors constitute the point cloud vector set V of the pth point. P ;

[0013] Judge V N Middle and V P The number of vectors that do not intersect and overlap is normalized to [0, 1], and the normalized value is used as the light intensity I of the pth point P ;

[0014] (2.3) Keep the light intensity range in [0, I T ] is the initial pit point cloud, I T is the preset light intensity threshold;

[0015] Step 3: Cluster and divide the initial pit point cloud through DBSCAN to obtain the pit cluster point cloud;

[0016] Step 4: Perform height histogram statistics on each pit cluster point cloud, and filter and retain the pit point cloud based on the height histogram statistical features.

[0017] As a further optimization scheme of the pothole extraction method in the vehicle-mounted road surface point cloud based on light intensity and statistical characteristics described in the present invention, the specific process of constructing the hemispherical illumination model in (2.1) is as follows: assume that a hemispherical light source O is placed at the pth point, the arc surface of O is on top, and N light rays are emitted outward; take the pth point as the origin, and construct N vectors with the Z direction always positive within the 360° range of the XOY plane of the spatial rectangular coordinate system and the hemisphere with the Z direction positive. These vectors constitute the illumination vector set V of the pth point N .

[0018] As a further optimization scheme of the pothole extraction method in the vehicle-mounted road surface point cloud based on light intensity and statistical characteristics described in the present invention, in (2.2), the specific process of calculating the light intensity of the pth point is:

[0019] ① Detect the pth point in the vehicle-mounted road surface point cloud. If the pth point has not been processed, initialize the illumination intensity of the pth point in the vehicle-mounted road surface point cloud to Ip = 0, take the pth point as the starting point, and other points in the vehicle-mounted road surface point cloud as the end point, and construct vectors. These vectors constitute the point cloud vector set V of point p P ;

[0020] ②Traverse the illumination vector set V of the pth point N If V N There are M vectors and V P The vectors in do not intersect and overlap, so the light intensity I at the pth point P =M / N.

[0021] As a further optimization scheme of the method for extracting potholes from a vehicle-mounted road surface point cloud based on light intensity and statistical characteristics described in the present invention, the specific steps of step 3 are as follows:

[0022] (3.1) Use DBSCAN to cluster and block the initial pit point cloud to obtain multiple cluster point clouds;

[0023] (3.2) Calculate the boundary range of each cluster point cloud projected onto the XOY plane of the spatial rectangular coordinate system. The minimum value of the cluster point cloud along the X-axis is minX, the maximum value of the cluster point cloud along the X-axis is maxX, the minimum value of the cluster point cloud along the Y-axis is minY, and the maximum value of the cluster point cloud along the Y-axis is maxY. Expand the boundary and retain the point cloud within the expanded boundary as the pit cluster point cloud.

[0024] As a further optimization scheme of the method for extracting potholes from a vehicle-mounted road surface point cloud based on light intensity and statistical characteristics described in the present invention, the specific steps of step 4 are as follows:

[0025] (4.1) Performing height histogram statistics on each pit cluster point cloud to obtain the statistical characteristics of the height histogram of the pit cluster point cloud;

[0026] (4.2) Determine the separation height H between the pothole and the road surface based on the statistical characteristics of the height histogram T ;

[0027] (4.3) Keep the height range of the pit cluster point cloud in [minZ, H T The points in ] are the pit point cloud, and minZ is the minimum height of the pit cluster point cloud.

[0028] As a further optimization scheme of the pothole extraction method in the vehicle-mounted road surface point cloud based on light intensity and statistical characteristics described in the present invention, in (4.2), after the boundary expansion of step (3.2), the pothole cluster point cloud not only contains the pothole point cloud, but also contains the road surface point cloud. The elevation histogram of the pothole cluster point cloud presents a double peak shape, and the elevation value at the trough is selected as the segmentation height of the pothole and the road surface of the cluster.

[0029] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0030] (1) First, the original road surface point cloud data is divided into blocks according to the time index; based on the spatial geometric characteristics of the pothole points, a hemispherical illumination model is constructed, and the initial pothole points are obtained by screening according to the feature threshold; then, the pothole point cloud is extracted quickly and accurately based on the spatial geometric characteristics and mathematical statistical characteristics of the pothole point cloud and the road surface point cloud;

[0031] (2) The method of the present invention is simple to operate, highly feasible, and greatly improves the efficiency and accuracy of automatic extraction of pit points. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the pothole extraction method in the vehicle-mounted road surface point cloud based on light intensity and statistical characteristics;

[0033] Figure 2 This is a schematic diagram of the hemispherical lighting model;

[0034] Figure 3 A schematic diagram of a road point cloud rendered using light intensity (darker areas are potholes);

[0035] Figure 4 It is the elevation histogram of the point cloud of the pit cluster. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0037] like Figure 1 As shown, the present invention provides a method for extracting potholes in a vehicle-mounted road surface point cloud based on light intensity and statistical characteristics, and the specific implementation method is as follows:

[0038] Step 1. Use the time index attribute of the point cloud scan line to segment the vehicle-mounted road surface point cloud, and the time interval of each point cloud segment is T0. The best time interval parameter is selected through experiments as T0 = 5 seconds.

[0039] Step 2. Construct a hemispherical illumination model and calculate the illumination intensity of the point cloud. The point cloud with illumination intensity within the preset illumination intensity threshold range is retained as the initial pit point cloud. The specific steps are as follows:

[0040] ① The specific process of constructing the hemispherical illumination model is as follows: Assume that a hemispherical light source O is placed at the pth point, with the arc surface of O facing upward and emitting N rays outward; take the pth point as the origin, and construct N vectors with the Z direction always positive within the 360° range of the XOY plane of the spatial rectangular coordinate system and the hemisphere with the Z direction positive. These vectors constitute the illumination set V of the pth point N ; The range of parameter N is 300-1000. Figure 2 It is a schematic diagram of the hemispherical lighting model of the present invention.

[0041] ② Take the pth point as the starting point and the other points in the vehicle-mounted road surface point cloud except the pth point as the end point to construct vectors. These vectors constitute the point cloud vector set V of the pth point P ;

[0042] ③Judge V N Middle and V P The number of vectors that do not intersect and overlap is normalized to [0, 1], and the normalized value is used as the light intensity I of the pth point P ;

[0043] ④ Determine the light intensity I at the pth point p The preset light intensity threshold I T If I p ≥I T , then the pth point is added to the initial pit point set; otherwise, it is added to the non-pit point set. The optimal illumination threshold parameter is selected through experiments as I T =0.7.

[0044] ⑤Traverse all points in the point cloud and repeat steps ①②③④ until all points are included in the initial pit point set or non-pit point set. Figure 3 This is a schematic diagram of the road point cloud rendered using light intensity (the darker areas are potholes).

[0045] Step 3. Use DBSCAN to cluster and block the initial pit point cloud, and project the initial pit point cloud to the boundary of the XOY plane of the spatial rectangular coordinate system to obtain the pit cluster point cloud. The specific steps are as follows:

[0046] ① Detect the pth point in the initial pit points that have not been checked. If the pth point has not been processed (classified as a cluster or marked as noise), check its neighborhood. If the number of objects contained is not less than the minimum number of contained points minPts, establish a new cluster C and add all the points in it to the candidate set X;

[0047] ② For all the qth points in the candidate set X that have not been processed, check their neighborhoods. If they contain at least minPts points, add these points to X; if the qth point does not belong to any cluster, add the qth point to C;

[0048] ③ Repeat step ② and continue to check the unprocessed points in X. The current candidate set X is empty;

[0049] ④ Repeat steps ①②③ until all points are classified into a cluster or marked as noise.

[0050] ⑤ Calculate the boundary range of each cluster point cloud projected onto the XOY plane of the spatial rectangular coordinate system: the minimum value of the cluster point cloud along the X axis is minX, the maximum value of the cluster point cloud along the X axis is maxX, the minimum value of the cluster point cloud along the Y axis is minY, and the maximum value of the cluster point cloud along the Y axis is maxY. Expand the boundary by GSD, where GSD is the expansion distance. The new boundary range is: minX+GSD, maxX+GSD, minY+GSD, maxY+GSD. Keep the point cloud within the new boundary range as the pit cluster point cloud. The best expansion distance parameter is selected through experiments as GSD=5 cm.

[0051] Step 4. Perform height histogram statistics on each pit cluster point cloud. According to the height histogram statistical features, select and retain the accurate pit point cloud. The specific steps are as follows:

[0052] ① Perform height histogram statistics on each pit cluster point cloud to obtain the statistical characteristics of the height histogram of the pit cluster point cloud.

[0053] ② The pothole cluster point cloud contains not only the pothole point cloud but also the road surface point cloud. Its elevation histogram is in the shape of double peaks. The elevation H at the trough is selected. T is the segmentation height between the pothole and the road surface of the cluster; Figure 4 It is the elevation histogram of a single pit cluster point cloud.

[0054] ③ Keep the height range in [minZ, H T] is the pit point cloud, and minZ is the minimum height of the pit cluster point cloud.

[0055] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a technician familiar with the technical field within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A method for extracting potholes from vehicle-mounted road surface point clouds based on light intensity and statistical characteristics, characterized in that: The following steps are involved: Step 1: Use the time index attribute to segment the vehicle-borne road surface point cloud into equal time intervals; Step 2: Construct a hemispherical illumination model and calculate the illumination intensity of the point cloud. The point cloud with illumination intensity within the preset illumination intensity threshold range is retained as the initial pit point cloud. The specific steps are as follows: (2.1) The hemispherical illumination model is: assume that a hemispherical light source is placed at the pth point in the vehicle-mounted road point cloud. The hemispherical light source emits N rays uniformly upward and outward. These N rays constitute the illumination vector set V of the pth point. N ; Where P≥p≥1, P is the total number of all points in the vehicle-mounted road point cloud; (2.2) Traverse all points in the point cloud until the illumination intensity of all points is calculated, where the illumination intensity I of the pth point is calculated. P ; The details are as follows: Take the pth point as the starting point and the other points in the vehicle-mounted road surface point cloud except the pth point as the end point to construct vectors. These vectors constitute the point cloud vector set V of the pth point. P ; Judge V N Middle and V P The number of vectors that do not intersect and overlap is normalized to [0, 1], and the normalized value is used as the light intensity I of the pth point P ; (2.3) Keep the light intensity range in [0, I T ] is the initial pit point cloud, I T is the preset light intensity threshold; Step 3: Cluster and divide the initial pit point cloud through DBSCAN to obtain the pit cluster point cloud; Step 4: Perform height histogram statistics on each pit cluster point cloud, and filter and retain the pit point cloud based on the height histogram statistical features.

2. The method for extracting potholes from vehicle-mounted road surface point clouds based on light intensity and statistical characteristics according to claim 1, characterized in that: In (2.1), the specific process of constructing the hemispherical illumination model is as follows: Assume that a hemispherical light source O is placed at the pth point, with the arc surface of O facing upward and emitting N light rays outward; take the pth point as the origin, and construct N vectors with positive Z direction in the hemisphere within the 360° range of the XOY plane of the spatial rectangular coordinate system and with positive Z direction. These vectors constitute the illumination vector set V of the pth point N .

3. The method for extracting potholes from vehicle-mounted road surface point clouds based on light intensity and statistical characteristics according to claim 1, characterized in that: In (2.2), the specific process of calculating the light intensity at the pth point is: ① Detect the pth point in the vehicle-mounted road surface point cloud. If the pth point has not been processed, initialize the illumination intensity of the pth point in the vehicle-mounted road surface point cloud to Ip = 0, take the pth point as the starting point, and other points in the vehicle-mounted road surface point cloud as the end point, and construct vectors. These vectors constitute the point cloud vector set V of point p P ; ②Traverse the illumination vector set V of the pth point N If V N There are M vectors and V P The vectors in do not intersect and overlap, so the light intensity I at the pth point P =M / N.

4. The method for extracting potholes from vehicle-mounted road surface point clouds based on light intensity and statistical characteristics according to claim 1, characterized in that: Step 3 The specific steps are as follows: (3.1) Use DBSCAN to cluster and block the initial pit point cloud to obtain multiple cluster point clouds; (3.2) Calculate the boundary range of each cluster point cloud projected onto the XOY plane of the spatial rectangular coordinate system. The minimum value of the cluster point cloud along the X-axis is minX, the maximum value of the cluster point cloud along the X-axis is maxX, the minimum value of the cluster point cloud along the Y-axis is minY, and the maximum value of the cluster point cloud along the Y-axis is maxY. Expand the boundary and retain the point cloud within the expanded boundary as the pit cluster point cloud.

5. The method for extracting potholes from vehicle-mounted road surface point clouds based on light intensity and statistical characteristics according to claim 4, characterized in that: Step 4 The specific steps are as follows: (4.1) Performing height histogram statistics on each pit cluster point cloud to obtain the statistical characteristics of the height histogram of the pit cluster point cloud; (4.2) Determine the separation height H between the pothole and the road surface based on the statistical characteristics of the height histogram T ; (4.3) Keep the height range of the pit cluster point cloud in [minZ, H T The points in ] are the pit point cloud, and minZ is the minimum height of the pit cluster point cloud.

6. The method for extracting potholes from vehicle-mounted road surface point clouds based on light intensity and statistical characteristics according to claim 5, characterized in that: In (4.2), after the boundary expansion in step (3.2), the pothole cluster point cloud contains not only the pothole point cloud but also the road surface point cloud. The elevation histogram of the pothole cluster point cloud presents a double peak shape, and the elevation value at the trough is selected as the segmentation height of the pothole and the road surface of the cluster.

Images