CN112633092B - Road information extraction method based on vehicle-mounted laser scanning point cloud - Google Patents
Road information extraction method based on vehicle-mounted laser scanning point cloud Download PDFInfo
- Publication number
- CN112633092B CN112633092B CN202011450170.1A CN202011450170A CN112633092B CN 112633092 B CN112633092 B CN 112633092B CN 202011450170 A CN202011450170 A CN 202011450170A CN 112633092 B CN112633092 B CN 112633092B
- Authority
- CN
- China
- Prior art keywords
- point
- points
- road
- neighborhood
- road marking
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a road information extraction method based on vehicle-mounted laser scanning point cloud. The method fully considers the characteristic that the neighborhood elevation difference of the ground point and the neighborhood elevation difference of the non-ground point have larger difference, realizes the separation of the ground points by utilizing the elevation characteristic of the neighborhood points, and better realizes the purpose of fine extraction of the road surface by utilizing the fluctuation and the elevation difference of the ground; the method has the advantages that the continuous fixed shape characteristics of the road marking are fully utilized, the road marking fine extraction and refinement based on the intensity characteristic image are realized, and meanwhile, the algorithm complexity is greatly reduced based on the point cloud intensity image statistics method; based on the clustered point cloud result, a relatively ideal result of road marking extraction is achieved by respectively using a vectorization method based on marking boundary point extraction of neighborhood distribution characteristics and a vectorization method based on marking boundary point extraction of Alpha Shape.
Description
Technical Field
The invention relates to a method for extracting road information of vehicle-mounted laser scanning point clouds, belonging to the field of remote sensing mapping.
Background
The road information refers to information of a sign line on a road, and common road information includes a road edge line, a road short and medium line, a zebra crossing, and the like. Road information is one of important components of basic geographic information, and accurate and high-precision road information plays an important role in city planning, road reconstruction and extension, traffic control, auxiliary driving, emergency response and the like. However, the road is often in a very harsh environment, and besides the load of the vehicle, the road is also subjected to actual weather conditions, so that the road is subject to serious problems such as decay and loss, and the road is required to be frequently maintained. In addition, the high-precision road information is also a core part of the driving map, and the driving assistance system can be better utilized to judge the road condition and carry out intelligent analysis and processing only based on the road information with centimeter level or higher precision. Therefore, the method has important practical significance for realizing high-precision and high-efficiency acquisition and extraction of the road information. However, the traditional road information acquisition method usually adopts a manual measurement mode, and the method has low mapping efficiency and long update period, is difficult to meet the requirement of quickly acquiring and updating road information, and has great limitation in the aspect of road information acquisition.
As a new measurement means and technology in the current surveying and mapping field, the vehicle-mounted laser scanning system can acquire large-area, high-density and high-precision three-dimensional space data in a short time, is high in surveying and mapping speed and high in automation degree, greatly reduces labor cost and reduces the risk of manual measurement, and becomes 1 important surveying and mapping technical means. Although the point cloud data obtained by scanning through the vehicle-mounted laser scanning system is rich in information quantity, the point cloud data contains non-road information such as trees and buildings, and the data quantity is large and serious redundancy is caused. In addition, the point cloud data is a collection of three-dimensional discrete points, cannot be directly used as GIS data, and requires a specific algorithm to extract specific surface feature information therefrom.
The research on how to extract road information from the scanning point cloud is a hot spot in the current international research. The vehicle-mounted laser scanning point cloud has large data volume and complex scene, and the shapes, positions and characteristics of small targets such as road markings and the like are not obvious in the mass point cloud, and are often processed and realized step by step. The method for extracting the road information of the longitudinal point cloud generally comprises the following three steps: firstly, extracting road surface information from initial point cloud; secondly, extracting the information of the road marking from the obtained road surface point cloud; and thirdly, carrying out boundary extraction and vectorization on the extracted road marking, thereby obtaining road information for practical use. In the aspect of research of road surface extraction, Manandhaar D and the like propose a method for counting point cloud height histograms on the basis of scanning lines, and Abuhaus and the like improve the method for extracting the height histograms, but the method is more suitable for flat planes and is not suitable for scenes with complex ground objects. The Weishun improves the filtering method of the point cloud, obtains the road surface through the growth of the road surface, but needs to obtain the prior information such as the driving track. In the aspect of extracting the road marking, Yuan and the like project three-dimensional point cloud into a two-dimensional image according to intensity, and then extract the road marking on the image by utilizing Hough transformation, Yang and the like provide an inverse distance weighting method to determine a gray value, and then extract the road marking by utilizing prior knowledge such as shape, arrangement and the like, but the method is limited by the size of a grid and easily causes precision loss. Yu et al calculate a plurality of intensity division thresholds using Otsu's algorithm to achieve extraction of road markings, but this method relies on setting of reflection intensity correction method or threshold estimation method parameters, and lacks universality. In the aspects of boundary extraction and vectorization of point cloud, Alrashdan and the like propose a boundary point extraction method based on a neural network, Kogyin and the like propose a method for point cloud space grid division to extract boundary points, Emelyanoy and the like extract boundary points by a point cloud triangularization method, but the methods have the problems of boundary point loss or redundancy, high complexity and the like, and the application scenes of the methods are limited to a certain extent.
At present, extracting positions and features of tiny target shapes such as road markings and the like which are not obvious in a massive vehicle-mounted point cloud scene is a hotspot and a difficulty of current research, and the prior art scheme has the defects of high complexity, precision loss, dependence on excessive prior information, low scene applicability, incomplete flow and the like, and cannot well meet the requirement of road information extraction based on vehicle-mounted laser scanning point cloud. The invention provides a vectorization full flow and a solution from extraction of road surfaces and extraction of road markings to road markings, and finally, the results are output to dxf vector files to realize warehousing of the road markings.
In conclusion, although many results aiming at vehicle-mounted point cloud road extraction are obtained at home and abroad at present, the method has own limitations, and a perfect, feasible, efficient and stable solution is lacked.
Disclosure of Invention
Aiming at the problem of extracting the road information of the vehicle-mounted laser scanning point cloud, the invention provides a set of complete flow and solution, and realizes the rapid and efficient extraction of the road information of the vehicle-mounted laser scanning point cloud. The technical problems to be solved by the invention are as follows:
(1) the extraction method of the road surface is improved, and the accuracy and the scene applicability are improved.
(2) The extraction of the road marking is improved, and the efficiency and the universality are improved.
(3) The vectorization of the road marking is improved, the vectorization precision is improved, and the complexity is reduced.
In view of the above problems, the solution provided by the present invention is as follows: the road information extraction method based on the vehicle-mounted laser scanning point cloud comprises the following steps:
the method comprises the following steps of 1, extracting ground points by using a method based on neighborhood elevation features, wherein the method comprises the steps of calculating the neighborhood point elevation difference of each point, judging according to the neighborhood point elevation difference, and extracting the ground points; then, taking the characteristics of small fluctuation of the road surface and large difference with other ground points in elevation into consideration, extracting the road surface by using a region growing method;
step 2, compared with the ground point, the road marking has a relatively regular shape in the three-dimensional space, and the reflection intensity of the road marking is often higher than that of the surrounding road surface, and aiming at the characteristic, extraction of the road marking based on an intensity characteristic image or directly on the three-dimensional space is provided;
and 3, dividing the road marking extracted in the step 2 into independent road marking point clouds by utilizing Euclidean clustering analysis, extracting boundary points of the point clouds by using a marking boundary point extraction method based on neighborhood distribution characteristics or a marking boundary point extraction method based on Alpha Shape, and outputting the extracted result in a vectorization manner.
Further, the specific implementation manner of step 1 is as follows,
(11) setting the point cloud point set as S, selecting any point p0 in S, and performing k-nearest neighbor search on p0, namely searching k points nearest to p0, wherein the point set of the k points is called S2;
(12) an arbitrary point pi is selected in S2, and the elevation squared difference dz between p0 and pi is calculated as shown in equation (1)iN is the total number of points in the point set S2;
(13) repeating the step (12) until all the points in the S2 are calculated, calculating the average neighborhood height difference h of the p0 according to the formula (2), comparing the h with a threshold value h0, if the obtained height difference is smaller than h0, then p0 is a ground point, and if the obtained height difference is larger than h0, then p0 is a non-ground point;
(14) repeating (11) (12) (13) until all the points in the S are traversed;
(dzi)2＝(zp0-zpi)2 (1)
by this point, the extraction of ground points has been completed.
Further, the extraction of the road marking based on the intensity characteristic image in the step 2 is divided into three steps: firstly, establishing a point cloud intensity characteristic image; secondly, extracting road marking based on edge detection; thirdly, refining the point cloud of the road marking; the specific implementation mode is as follows:
(21a) establishment of point cloud intensity characteristic image
(211) Traversing the road surface point cloud point set S' to obtain the range of the point cloud in the direction X, Y, namely Xmax, Xmin, Ymax and Ymin;
(212) selecting the size W of a proper projection grid, and calculating the width and height of the generated projection image according to the formula (3) and the formula (4);
(213) traversing all grids, counting the number n of points falling into the grids and the reflection intensity t of the points for the point cloud of each grid, calculating the average reflection intensity t0 of the grid according to the formula (5), and setting the average reflection intensity of the grid to be 0 when no point falls into the grid;
(214) traversing all the grids to obtain the maximum values tmax and tmin of the average reflection intensity, normalizing the average reflection intensity of all the grids according to a formula (6), and reducing the average reflection intensity to be between 0 and 255 to obtain a normalized reflection intensity value tg which is used as the gray value of the grid;
width＝(Xmax-Xmin)/W (3)
height＝(Ymax-Ymin)/W (4)
tg＝(255-0)*(t0-tmin)/(tmax-tmin) (6)
so far, a point cloud intensity feature image is established;
(22b) road marking extraction based on edge detection
Performing Sobel edge detection on the point cloud intensity characteristic image, performing plane convolution on the image by using two matrixes of formula (7) to obtain horizontal and longitudinal gray difference approximate values, combining the horizontal and longitudinal gradient approximate values of each pixel by using formula (8) to obtain the gradient magnitude G of each pixel, and finally obtaining an edge extraction image;
performing gray level histogram analysis on the obtained edge extraction image, setting threshold values k1 and k2, judging a pixel grid with gray level between k1 and k2 as an edge grid of a road marking, and recording an index of the edge grid;
traversing all edge grids, and if the difference between two continuous grids in the X direction is only one to two grids, recording the grids in the middle of the two grids, realizing the separation of road marking lines on the point cloud characteristic image, and obtaining a separated point set S1;
(23c) refinement of road markings
(231) Selecting a proper neighborhood size R, and selecting any point p0 from a point set S1, wherein the point set of the R neighborhood of p0 is S2;
(232) traversing all points in S2, calculating the average neighborhood intensity t0 of p0, simultaneously comparing the intensity of p0 and all points (p1, p2..., pn) in the neighborhood, and calculating the proportion c0 of the points with the intensity smaller than p0 in the R neighborhood of p 0;
(233) setting a proper average neighborhood intensity threshold k3 and a proper proportion threshold k4, judging i0 and c0, and if the average neighborhood intensity threshold k3 and the proportion threshold k4 are within the threshold, counting p0 as a reticle point;
(234) and traversing all the point sets, and repeating the operations (231) (232) (233) to complete the extraction of the reticle points.
Further, the step 2 of directly extracting the road marking in the three-dimensional space is divided into two steps: firstly, extracting road marking based on reflection intensity characteristics and neighborhood density characteristics of points; outlier filtering based on statistical analysis; the specific implementation mode is as follows:
(21e) road marking extraction based on reflection intensity characteristics and neighborhood density characteristics of points
(21e1) Setting a point cloud point set of the road surface as S ', selecting any point p0 in S', performing r range search by taking r as a radius, setting a set of points with the distance from p0 within r as S2, and setting the reflection intensity of p0 as t;
(21e2) traversing the point set S2, and counting the number n of points in S2;
(21e3) setting adjacent point density threshold values n1, n2 and reflection intensity threshold values t1, t2, and if n1< n < n2 and t > t1, determining that the road marking point is far away from the scanning center; if n > n2 and t > t2, determining that the road marking point is close to the scanning center;
(21e4) traversing the point set S, repeating the operation, finishing the judgment of all the points, and finally separating out the road marking point set;
(21f) outlier filtering based on statistical analysis
Calculating the average distance from each point cloud in the set of road marking points to all its adjacent points, assuming that the result is a gaussian distribution whose shape is determined by the mean and standard deviation, points whose average distance is outside the standard range can be defined as outliers and removed from the data set; then, a proper variance threshold value is set, and the road marking can be separated from the noise.
Furthermore, the specific implementation manner of the reticle boundary point extraction method based on the neighborhood distribution characteristics in the step 3 is as follows,
and setting the road marking point set as S3 and the search radius as r, wherein the specific algorithm idea is as follows:
(31a) selecting any point p0 from the point set S3, wherein the set of all points in r neighborhood of p0 is called S2;
(32a) selecting any point p1 in S2, and calculating the difference dx, dy of x and y coordinates of p0 and p1 according to the formula (9) and the formula (10);
(33a) repeating (32a) until all points in S2 are calculated, and calculating the sum sdx, sdy of all dx, dy according to the formula (11) and the formula (12);
(34a) calculating a square sum and a power value k of sdx and sdy according to formula (13), comparing the square sum and the power value k with a threshold value k0, and judging p0 as a boundary point if the square sum and the power value k are less than the threshold value;
(35a) selecting a next point in the point set S3, and repeating the operations (31a) (32a) (33a) (34a) until all the point clouds are traversed;
dx＝xp0-xp1 (9)
dy＝yp0-yp1 (10)
furthermore, the specific implementation manner of the reticle boundary point extraction method based on Alpha Shape in step 3 is as follows,
(31b) setting the road marking point set as S3, selecting any point p1 from the point set S3, drawing a circle with the radius of 2 alpha and the p1 as the center, and setting the point set in the radius as S2;
(32b) selecting any point p2 in S2;
(33b) drawing a circle with the radius of alpha through p1 and p2 to obtain a circle center p0, and judging whether the rest points in S2 are in the circle;
(34b) if the rest points are not in the circle, the p1 and the p2 are boundary points, the indexes are stored, and the operation is carried out (35 b);
(35b) if the points are in the circle, repeating the steps (32b) and (33b) until all the points in S2 are judged;
(36b) and selecting the next point from the S3, and repeating the operation until all the points are traversed.
Compared with the prior art, the invention has the advantages and beneficial effects that: 1) the characteristics of vehicle-mounted laser scanning point cloud are fully considered, the characteristics of a vectorization target are combined, road surface extraction is used as an entry point, the whole vectorization extraction process is constructed, and the aim of extracting road markings from a complex vehicle-mounted point cloud scene is fulfilled;
2) the method fully considers the characteristic that the neighborhood elevation difference of the ground point and the neighborhood elevation difference of the non-ground point have larger difference, realizes the separation of the ground points by utilizing the elevation characteristic of the neighborhood points, and better realizes the purpose of fine extraction of the road surface by utilizing the fluctuation and the elevation difference of the ground;
3) the method comprises the steps of establishing a point cloud intensity image, fully utilizing the shape characteristics of road markings which are continuously fixed by point cloud intensity statistical information, realizing the fine extraction and refinement of the road markings based on the intensity characteristic image, and greatly reducing the complexity of an algorithm by the point cloud intensity image statistical method;
4) and the separation of the individual road marked lines is realized by using the Euclidean distance clustering algorithm. On the basis, namely based on the clustered point cloud result, a relatively ideal result of the road marking extraction is achieved by respectively using the vectorization method of the marking boundary point extraction based on the neighborhood distribution characteristics and the vectorization method of the marking boundary point extraction based on the Alpha Shape.
Drawings
FIG. 1 is a flow chart of road growth.
Fig. 2 is a point cloud intensity feature image (a) and a Sobel edge extraction image (b).
Fig. 3 is a gray level histogram of an edge extraction image.
Fig. 4 is a schematic diagram of a refined road marking.
FIG. 5 is an exemplary graph of statistical analysis filtering.
Fig. 6 is schematic diagrams before and after filtering of the road marking, wherein (a) is the schematic diagram before filtering of the road marking, and (b) is the schematic diagram after filtering of the road marking.
Fig. 7 is a neighborhood point distribution of a road marking.
FIG. 8 is a schematic diagram of the Alpha Shape algorithm.
Detailed Description
The technical scheme of the invention is further explained by combining the drawings and the embodiment.
The embodiment of the invention provides a road information extraction method based on vehicle-mounted laser scanning point cloud, which mainly comprises three steps, and the detailed flow and algorithm of each step are described in detail below.
Step one, extracting road surface
Although the point cloud is irregular and discrete in spatial distribution, points with the same characteristics have strong spatial consistency, and ground points are characterized by small elevation difference of neighboring points compared with points of other ground structures (buildings and trees). Based on the characteristics, the invention extracts ground points by using a method based on neighborhood elevation characteristics, calculates the neighborhood point height difference of each point, judges through the neighborhood point height difference, realizes the extraction of the ground points, and extracts the road surface by using a region growing method in consideration of the characteristics that the road surface has small fluctuation and large elevation difference with other ground points.
Considering that the height difference of the neighborhood of the ground point is different from the height difference of the neighborhood of the non-ground point, the neighborhood points of the ground point often have similar elevations, but the distribution of the neighborhood points of the non-ground point is more dispersed, the elevation difference is larger, and the separation of the ground point can be realized through the elevation characteristics of the neighborhood points of the points. And (3) setting a point cloud point set as S, wherein the specific algorithm idea is as follows:
(11) an arbitrary point p0 is selected in S, k neighbor search is performed on p0, that is, k points closest to p0 are searched, and a point set of the k points is referred to as S2.
(12) An arbitrary point pi is selected in S2, and the elevation squared difference dz between p0 and pi is calculated as shown in equation (1)iZp0, zpi represent the elevation of the corresponding points, respectively, and n is the total number of points in the point set S2.
(13) Repeating the step (12) until all points in the S2 are calculated, calculating the average neighborhood height difference h of the p0 according to the formula (2), comparing the h with a threshold value h0 (the h0 can be obtained by comparing different height differences of the neighborhood of the ground feature), if the obtained height difference is smaller than h0, the p0 is a ground point, and if the obtained height difference is larger than h0, the p0 is a non-ground point.
(14) And (11), 12 and 13 are repeated until all points in the S are traversed.
By this point, the extraction of ground points has been completed.
(dzi)2＝(zp0-zpi)2 (1)
And (II) the ground points obtained in the last step also consist of a plurality of components, including road surfaces and ground points outside roads. According to the actual road condition, the fluctuation of the road surface is small, so that the elevation difference between a point on the road surface and a point adjacent to the point is small, and the elevation difference between a ground point outside the road and the point on the road is large. The road surface can be extracted from the ground point cloud by a region growing method.
Selecting a plurality of road points as seed points, taking the seed points as an initial area, taking the small elevation difference between the adjacent points and the initial point as a growth condition, taking the overlarge elevation difference between the adjacent points and the seed points as a termination condition of the growth of the road surface, and carrying out the growth of the road surface until no new seed points are added; wherein the road surface growth process is shown in figure 1.
Step two, extracting the road marking
Compared with the ground point, the road marking has a relatively regular shape in a three-dimensional space, and the reflection intensity of the road marking is often higher than that of the surrounding road surface.
Extraction of road marking based on point cloud intensity characteristic image
The reflection intensity characteristic based on vehicle-mounted laser scanning point cloud data is considered, the reflection intensity of marking points is higher, in addition, the marking also has the shape characteristic of continuous fixation, and the extraction of the road marking can be realized in a projection mode.
The extraction of the road marking based on the intensity characteristic image is mainly divided into three steps: firstly, establishing a point cloud intensity characteristic image; secondly, extracting road marking based on edge detection; and thirdly, refining the road marking point cloud.
The pavement point is obtained in the first step and is marked as S road
1. Establishment of point cloud intensity characteristic image
(1) And traversing the point cloud point set S (namely [ S way ]) to obtain the range of the point cloud in the direction X, Y, namely Xmax, Xmin, Ymax and Ymin.
(2) The width and height of the generated projection image are calculated by equation (3) and equation (4) by selecting a suitable projection grid size W, the size of which is determined by the scan line interval.
(3) And traversing all the grids, counting the number n of points falling into the grids and the reflection intensity t of the points for the point cloud of each grid, and calculating the average reflection intensity t0 of the grid according to the formula (5). When no point falls in the grid, the average reflection intensity of the grid is set to 0.
(4) Traversing all the grids to obtain the maximum values tmax and tmin of the average reflection intensity, normalizing the average reflection intensity of all the grids according to the formula (6), and reducing the average reflection intensity to be between 0 and 255 to obtain a normalized reflection intensity value tg which is used as the gray value of the grid.
width＝(Xmax-Xmin)/W (3)
height＝(Ymax-Ymin)/W (4)
tg＝(255-0)*(t0-tmin)/(tmax-tmin) (6)
To this end, a point cloud intensity feature image is created, as shown in fig. 2(a), with a mesh size W of 0.25 mm.
2. Road marking extraction based on edge detection
Because the gray value of the pixel point where the road marking is located is higher in the intensity characteristic image, and the gray difference between the pixel point and the surrounding pixel points is obvious, the edge detection can be used for obtaining the grid where the edge point of the road marking is located, and then the grid where the whole marking is located is obtained by detecting the two edges of the road marking. The specific algorithm steps are as follows:
sobel edge detection is carried out on the point cloud intensity characteristic image, and the two matrixes of the formula (7) are utilized to carry out plane convolution on the image so as to obtain horizontal and longitudinal gray difference approximate values. The horizontal and vertical gradient approximations of each pixel are combined by using the formula (8) to obtain the gradient magnitude G of each pixel, and finally the obtained edge extraction image is shown in fig. 2 (b).
The obtained edge extracted image is subjected to gray level histogram analysis (see fig. 3), threshold values k1 and k2 are set, a pixel grid having a gray level between k1 and k2 (determined by the two peaks of the histogram) is determined as an edge grid of a road marking, and an index thereof is recorded.
All edge meshes are traversed and if two consecutive meshes differ in the X direction by only one to two meshes, the mesh in the middle of the two meshes is also recorded. Therefore, the grid points where the complete road marking is located are recorded, and the separation of the road marking on the point cloud characteristic image is realized.
3. Refinement of road markings
Because the point cloud characteristic image is obtained by projecting the point cloud in the elevation direction, each grid and the three-dimensional point cloud space have unique corresponding position relation, under the condition that the grid where the road marking is located is known, the original point cloud of the road marking in the three-dimensional space can be obtained through indexing.
However, due to the limitation of the mesh size, the road marking point cloud obtained from the feature image often has the existence of adjacent road surface points, which means that the obtained road marking point cloud is not accurate enough, and needs to be extracted precisely, so as to obtain accurate road markings. The invention can separate the marked line point from the adjacent ground point by selecting the scale with proper size. The specific algorithm is as follows:
in the foregoing, separation of road markings on a point cloud feature image is realized, and since a separation result is rough, the part of point cloud needs to be extracted and subjected to refinement processing, and the part of point cloud is recorded as [ S separation ].
(1) Selecting a proper neighborhood size R, and selecting any point p0 from a point set S (the point cloud result of the rough extraction, namely the S separation), wherein the point set of the R neighborhood of p0 is S2.
(2) Traversing all points in S2, calculating the average neighborhood intensity t0 of p0 according to equation (5), meanwhile, comparing the intensity of p0 and all points (p1, p2... pn) in the neighborhood, and calculating the proportion c0 of the points with smaller intensity than p0 in the R neighborhood of p 0.
(3) Setting a proper average neighborhood intensity threshold k1 and a proper proportion threshold k2 (obtained by comparing the average neighborhood intensity and the intensity proportion of the reticle point and the road point), judging i0 and c0, and if the average neighborhood intensity and the intensity proportion of the reticle point and the road point are within the threshold, counting p0 as the reticle point.
(4) And (5) traversing all the point sets, repeating the operations in the steps (1), (2) and (3), and finishing the extraction of the marking points.
The extraction effect is shown in fig. 4, and it can be seen from the figure that the adjacent road points are filtered out, and the marking points are extracted.
(II) direct extraction of three-dimensional road marking
Based on the reflection intensity characteristics of the vehicle-mounted laser scanning point cloud data, the road marking points have higher reflection intensity compared with the surrounding points on the road surface, but due to the distance relationship, the reflection intensity of the road marking far away from the scanning center is often lower than that of the road marking near the scanning center, and the road marking points need to be extracted separately. In addition, based on the density characteristics of the vehicle-mounted laser scanning point cloud data, the closer the distance from the scanning center, the lower the density of the adjacent points, and therefore the density of the adjacent points at the road marking position far from the scanning center is lower than the density of the adjacent points at the road marking position near the scanning center.
By combining the reflection intensity and the neighborhood density characteristics, the road marking far away from the scanning center and the road marking close to the scanning center are separately extracted in a mode of setting a plurality of thresholds, so that the direct extraction of the road marking point cloud is realized.
The steps of directly extracting the road marking in the three-dimensional space are mainly divided into two steps: firstly, extracting road marking based on reflection intensity characteristics and neighborhood density characteristics of points; second, outlier filtering based on statistical analysis.
1. Road marking extraction based on reflection intensity characteristics and neighborhood density characteristics of points
(1) Let the point cloud point set be S (road surface point cloud point set, i.e., [ S road ]), select an arbitrary point p0 in S, perform r range search with r as a radius, and the set of points within r from p0 is called S2. The reflection intensity of p0 is t.
(2) Traversing the point set S2, and counting the number n of the points in S2;
(3) setting adjacent point density threshold values n1, n2 and reflection intensity threshold values t1, t2 (the threshold values are set according to the distribution conditions of the intensity and the density of the point cloud at different positions). If n1< n < n2 and t > t1, determining a road marking point far away from the scanning center; if n > n2 and t > t2, it is determined that the road marking point is close to the scan center.
(4) And traversing the point cloud S, repeating the operation, finishing the judgment of all the points, and finally separating a road marking point set.
2. Outlier filtering based on statistical analysis
Due to the characteristics of back scattering and the material characteristics of asphalt roads, some pavement points show reflection intensity similar to that of a road marking, so that the road marking obtained in the first step often contains some noise, and compared with the points of the road marking, the points often show dispersion and irregularity, and the points can be filtered by using filtering of statistical analysis.
The filtering principle of statistical analysis is as follows:
for each point (i.e., the set of isolated road marking points), its average distance to all its proximate points is calculated. Assuming that the result is a gaussian distribution whose shape is determined by the mean and standard deviation, points whose mean distance is outside the standard range (defined by the global distance mean and variance) can be defined as outliers and removed from the data set. An example of statistical analysis filtering is shown in figure 5.
Setting a proper variance threshold (obtained by comparing the influence of different variances on the filtered result), the road marking can be separated from the noise. As shown in fig. 6(a), the pre-filtering and post-filtering effects are road markings before filtering, i.e., outliers in the rectangular frame, and (b) are road markings after filtering.
Step three, vectorizing road marking
The invention firstly utilizes European cluster analysis to segment the marked line point cloud into independent marked line point cloud, then utilizes two boundary extraction methods of marked line boundary point and internal point neighborhood point distribution difference or Alpha Shape extraction algorithm to realize the boundary point extraction of the point cloud, compares, analyzes and extracts the effect, and outputs the extracted result vectorization.
The vectorization of the road marking is a key part of the extraction of the road information, and the vectorization aims to convert the road marking point cloud data which are difficult to be directly utilized into vector data which can be utilized. The vectorization of the road marking is mainly characterized in that marking boundary points are extracted, the boundary outline of the road marking can be obtained through the extraction of the boundary points, the topological relation among the boundary points is further calculated, and the generation and vectorization output of the boundary lines can be realized.
Step 3.1, Euclidean distance clustering
Since the extracted road marking points are independent and unorganized points, they need to be clustered before vectorization, so that each class represents a separate road marking.
Because each independent road marking is discontinuous, the distance interval between the independent road markings is usually far, and the separation of the independent road markings can be realized by a Euclidean distance clustering method. And (3) judging by taking the Euclidean distance between the point and the point as a measure function, traversing all the points in the point cloud (the Euclidean distance between the point cloud of the road marking extracted in the step two needs to eliminate the influence of other points during clustering), if the Euclidean distance between the point and the adjacent point is less than a threshold value, classifying the point cloud into the same class, otherwise, generating a new class, assigning different colors to the different classes, and finally obtaining a clustered road marking point set.
Step 3.2a, marking boundary point extraction and vectorization based on neighborhood distribution characteristics
Considering that the neighborhood distribution of the boundary points and the interior points of the road marking are different (as shown in fig. 7), the neighborhood points of the boundary points are often distributed on the same side, and the neighborhood points of the interior points of the marking are often distributed uniformly, so that the judgment can be performed through the neighborhood characteristics of the road marking, and the extraction of the boundary points is realized.
And setting the road marking point set as S3 and the search radius as r, wherein the specific algorithm idea is as follows:
(1) any point p0 is selected from the point set S3, and the set of all points in the r neighborhood of p0 is called S2.
(2) At S2, an arbitrary point p1 is selected, and the x, y coordinate difference dx, dy between p0 and p1 is calculated by the following equations (9) and (10).
(3) Repeating the step (2) until all points in S2 are calculated, and calculating the sum sdx, sdy of all dx, dy according to the formula (11) and the formula (12).
(4) The sum of squares k of sdx and sdy is calculated according to equation (13), and when the sum of squares k is smaller than a threshold value k0(k0 is obtained by comparing the difference between the sum of squares of the inner point and the boundary point), p0 is determined as the boundary point.
(5) And (4) selecting the next point in the point set S3, and repeating the operations (1), (2), (3) and (4) until all the points are traversed.
dx＝xp0-xp1 (9)
dy＝yp0-yp1 (10)
Step 3.2b, marking boundary point extraction and vectorization based on Alpha Shape
The Alpha Shape algorithm is an algorithm capable of effectively extracting edges of a discrete point set. The principle of the method is as shown in fig. 8, a set of point set S is set, the Shape of Alpha Shape of the set depends on a radius parameter α, a circle with the radius α can be imagined to roll outside the point set S, when α is large enough, the circle cannot roll into the point set, and the rolling trace of the circle is the boundary line of the point set S. Therefore, the extraction effect of the Alpha Shape algorithm depends on the size of Alpha, and better boundary extraction can be realized by selecting proper parameters. The specific algorithm idea is as follows:
(1) an arbitrary point p1 is selected from the point set S3 (i.e., the road-marking point set), and a circle is drawn with a radius of 2 α around p1, and the point set within this radius is referred to as S2.
(2) An arbitrary point p2 is selected in S2.
(3) Drawing a circle with the radius of alpha through p1 and p2 to obtain a circle center p0, and judging whether the rest points in S2 are in the circle.
(4) If all the remaining points are not within the circle, p1 and p2 are boundary points, and the index is stored and (5) is executed.
(5) If the points are in the circle, repeating the steps (2) and (3) until all the points in the S2 are judged to be completed.
(6) And selecting the next point from the S3, and repeating the operation until all the points are traversed.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments, or alternatives may be employed, by those skilled in the art, without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (5)
1. The road information extraction method based on the vehicle-mounted laser scanning point cloud is characterized by comprising the following steps of:
the method comprises the following steps of 1, extracting ground points by using a method based on neighborhood elevation features, wherein the method comprises the steps of calculating the neighborhood point elevation difference of each point, judging according to the neighborhood point elevation difference, and extracting the ground points; then extracting the road surface by using a region growing method;
step 2, regarding to a ground point, the road marking has a relatively regular shape in a three-dimensional space, and the reflection intensity of the road marking is often higher than that of the surrounding road surface, and for the characteristic, extraction of the road marking based on an intensity characteristic image is provided;
in the step 2, the extraction of the road marking based on the intensity characteristic image is divided into three steps: firstly, establishing a point cloud intensity characteristic image; secondly, extracting road marking based on edge detection; thirdly, refining the point cloud of the road marking; the specific implementation mode is as follows:
(21a) point cloud intensity feature image establishment
(211) Traversing the road surface point cloud point set S' to obtain the range of the point cloud in the direction X, Y, namely Xmax, Xmin, Ymax and Ymin;
(212) selecting the size W of the projection grid, and calculating the width and height of the generated projection image according to the formula (3) and the formula (4);
(213) traversing all grids, counting the number n of points falling into the grids and the reflection intensity t of the points for the point cloud of each grid, calculating the average reflection intensity t0 of the grid according to the formula (5), and setting the average reflection intensity of the grid to be 0 when no point falls into the grid;
(214) traversing all grids to obtain maximum values tmax and tmin of the average reflection intensity, normalizing the average reflection intensity of all the grids according to a formula (6), and reducing the average reflection intensity to be between 0 and 255 to obtain a normalized reflection intensity value tg which is used as a gray value of the grid;
width＝(Xmax-Xmin)/W (3)
height＝(Ymax-Ymin)/W (4)
tg＝(255-0)*(t0-tmin)/(tmax-tmin) (6)
so far, a point cloud intensity feature image is established;
(22b) road marking extraction based on edge detection
Performing Sobel edge detection on the point cloud intensity characteristic image, performing plane convolution on the image by using two matrixes of formula (7) to obtain horizontal and longitudinal gray difference approximate values, combining the horizontal and longitudinal gradient approximate values of each pixel by using formula (8) to obtain the gradient magnitude G of each pixel, and finally obtaining an edge extraction image;
performing gray level histogram analysis on the obtained edge extraction image, setting threshold values k1 and k2, judging a pixel grid with gray level between k1 and k2 as an edge grid of a road marking, and recording an index of the edge grid;
traversing all edge grids, and if the difference between two continuous grids in the X direction is only one to two grids, recording the grids in the middle of the two grids, realizing the separation of road marking lines on the point cloud characteristic image, and obtaining a separated point set S1;
(23c) refinement of road markings
(231) Selecting neighborhood size R, selecting any point p0 from the point set S1, wherein the point set of the R neighborhood of p0 is S2;
(232) traversing all points in S2, calculating the average neighborhood intensity t0 of p0, simultaneously comparing the intensity of p0 and all points (p1, p2..., pn) in the neighborhood, and calculating the proportion c0 of the points with the intensity smaller than p0 in the R neighborhood of p 0;
(233) setting an average neighborhood intensity threshold k3 and a proportion threshold k4, judging i0 and c0, and if the average neighborhood intensity threshold k3 and the proportion threshold k4 are within the threshold, counting p0 as a reticle point;
(234) traversing all the point sets, repeating the operations (231) (232) (233) and finishing the extraction of the marking points;
and 3, dividing the road marking extracted in the step 2 into independent road marking point clouds by utilizing Euclidean clustering analysis, extracting boundary points of the point clouds by using a marking boundary point extraction method based on neighborhood distribution characteristics or a marking boundary point extraction method based on Alpha Shape, and outputting the extracted result in a vectorization manner.
2. The road information extraction method based on vehicle-mounted laser scanning point cloud as claimed in claim 1, characterized in that: the specific implementation of step 1 is as follows,
(11) setting the point cloud point set as S, selecting any point p0 in S, and performing k-nearest neighbor search on p0, namely searching k points nearest to p0, wherein the point set of the k points is called S2;
(12) an arbitrary point pi is selected in S2, and the elevation squared difference dz between p0 and pi is calculated as shown in equation (1)iZp0 and zpi respectively represent the elevations of corresponding points, and n is the total number of points in the point set S2;
(13) repeating the step (12) until all the points in the S2 are calculated, calculating the average neighborhood height difference h of the p0 according to the formula (2), comparing the h with a threshold value h0, if the obtained height difference is smaller than h0, then p0 is a ground point, and if the obtained height difference is larger than h0, then p0 is a non-ground point;
(14) repeating (11) (12) (13) until all the points in the S are traversed;
(dzi)2＝(zp0-zpi)2 (1)
by this point, the extraction of ground points has been completed.
3. The road information extraction method based on vehicle-mounted laser scanning point cloud as claimed in claim 1, characterized in that: the step 2 of extracting the road marking also comprises the step of directly extracting the road marking in a three-dimensional space, and the method comprises the following two steps: firstly, extracting road marking based on reflection intensity characteristics and neighborhood density characteristics of points; outlier filtering based on statistical analysis; the specific implementation mode is as follows:
(21e) road marking extraction based on reflection intensity characteristics and neighborhood density characteristics of points
(21e1) Setting a point cloud point set of the road surface as S ', selecting any point p0 in S', performing r range search by taking r as a radius, setting a set of points with the distance from p0 within r as S2, and setting the reflection intensity of p0 as t;
(21e2) traversing the point set S2, and counting the number n of the points in S2;
(21e3) setting adjacent point density thresholds n1 and n2 and reflection intensity thresholds t1 and t2, and if n1 is more than n and less than n2 and t is more than t1, judging that the road marking point is far away from the scanning center; if n > n2 and t > t2, determining that the road marking point is close to the scanning center;
(21e4) traversing the point set S', repeating the operation, finishing the judgment of all the points, and finally separating out a road marking line point set;
(21f) outlier filtering based on statistical analysis
For each point cloud in the set of isolated road marking points, calculating its average distance to all its neighbors, assuming that the result is a gaussian distribution whose shape is determined by the mean and standard deviation, points whose average distance is outside the standard range can be defined as outliers and removed from the data set; then, setting a variance threshold value, namely separating the road marking from the noise.
4. The road information extraction method based on vehicle-mounted laser scanning point cloud as claimed in claim 1, characterized in that: the specific implementation manner of the reticle boundary point extraction method based on the neighborhood distribution characteristics in the step 3 is as follows,
and setting the road marking point set as S3 and the search radius as r, wherein the specific algorithm idea is as follows:
(31a) selecting any point p0 from the point set S3, wherein the set of all points in r neighborhood of p0 is called S2;
(32a) selecting any point p1 in S2, and calculating the difference dx, dy of x and y coordinates of p0 and p1 according to the formula (9) and the formula (10);
(33a) repeating (32a) until all points in S2 are calculated, and calculating the sum sdx, sdy of all dx, dy according to the formula (11) and the formula (12);
(34a) calculating a square sum and a power value k of sdx and sdy according to formula (13), comparing the square sum and the power value k with a threshold value k0, and judging p0 as a boundary point if the square sum and the power value k are less than the threshold value;
(35a) selecting a next point in the point set S3, and repeating the operations (31a) (32a) (33a) (34a) until all point clouds are traversed;
dx＝xp0-xp1 (9)
dy＝yp0-yp1 (10)
5. the road information extraction method based on vehicle-mounted laser scanning point cloud as claimed in claim 1, characterized in that: the specific implementation manner of the marking boundary point extraction method based on Alpha Shape in the step 3 is as follows,
(31b) setting the road marking point set as S3, selecting any point p1 from the point set S3, drawing a circle with the radius of 2 alpha and the p1 as the center, and setting the point set in the radius as S2;
(32b) selecting any point p2 in S2;
(33b) drawing a circle with the radius of alpha through p1 and p2 to obtain a circle center p0, and judging whether the rest points in S2 are in the circle;
(34b) if the rest points are not in the circle, the p1 and the p2 are boundary points, the indexes are stored, and the operation is carried out (35 b);
(35b) if the points are in the circle, repeating the steps (32b) (33b) until all the points in the S2 are judged;
(36b) and selecting the next point from the S3, and repeating the operation until all the points are traversed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011450170.1A CN112633092B (en) | 2020-12-09 | 2020-12-09 | Road information extraction method based on vehicle-mounted laser scanning point cloud |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011450170.1A CN112633092B (en) | 2020-12-09 | 2020-12-09 | Road information extraction method based on vehicle-mounted laser scanning point cloud |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112633092A CN112633092A (en) | 2021-04-09 |
| CN112633092B true CN112633092B (en) | 2022-06-14 |
Family
ID=75309988
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011450170.1A Active CN112633092B (en) | 2020-12-09 | 2020-12-09 | Road information extraction method based on vehicle-mounted laser scanning point cloud |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112633092B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113280798B (en) * | 2021-07-20 | 2021-10-01 | 四川省公路规划勘察设计研究院有限公司 | Geometric correction method for vehicle-mounted scanning point cloud under tunnel GNSS rejection environment |
| CN113759947B (en) * | 2021-09-10 | 2023-08-08 | 中航空管系统装备有限公司 | Airborne flight obstacle avoidance assisting method, device and system based on laser radar |
| CN114119998B (en) * | 2021-12-01 | 2023-04-18 | 成都理工大学 | Vehicle-mounted point cloud ground point extraction method and storage medium |
| CN114549879B (en) * | 2022-04-25 | 2022-07-08 | 四川省公路规划勘察设计研究院有限公司 | Target identification and central point extraction method for tunnel vehicle-mounted scanning point cloud |
| CN115649069B (en) * | 2022-12-14 | 2023-03-03 | 四川省公路规划勘察设计研究院有限公司 | Vehicle-mounted laser point cloud scanner |
| CN116363159B (en) * | 2023-04-19 | 2024-05-03 | 济南大学 | Point cloud road boundary extraction method based on multi-window differential threshold filtering method |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013092516A (en) * | 2011-10-06 | 2013-05-16 | Asia Air Survey Co Ltd | Laser ortho-image generation device, laser ortho-image generation program, and laser ortho-image generation method |
| CN106199557A (en) * | 2016-06-24 | 2016-12-07 | 南京林业大学 | A kind of airborne laser radar data vegetation extracting method |
| CN107679458A (en) * | 2017-09-07 | 2018-02-09 | 中国地质大学（武汉） | The extracting method of roadmarking in a kind of road color laser point cloud based on K Means |
| CN108062517A (en) * | 2017-12-04 | 2018-05-22 | 武汉大学 | Unstructured road boundary line extraction method based on vehicle-mounted laser point cloud |
| CN109272458A (en) * | 2018-08-10 | 2019-01-25 | 河海大学 | A kind of point cloud filtering method based on prior information |
| CN208432477U (en) * | 2018-07-27 | 2019-01-25 | 西南交通大学 | The Experimental mimic system of Foundation Base in Railway Tunnel structure |
| CN110502973A (en) * | 2019-07-05 | 2019-11-26 | 同济大学 | A kind of roadmarking automation extraction and recognition methods based on vehicle-mounted laser point cloud |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106525000B (en) * | 2016-10-31 | 2017-07-11 | 武汉大学 | Roadmarking automation extracting method based on laser scanning discrete point intensity gradient |
| JP6871782B2 (en) * | 2017-03-31 | 2021-05-12 | 株式会社パスコ | Road marking detector, road marking detection method, program, and road surface detector |
| CN107092020B (en) * | 2017-04-19 | 2019-09-13 | 北京大学 | Merge the surface evenness monitoring method of unmanned plane LiDAR and high score image |
-
2020
- 2020-12-09 CN CN202011450170.1A patent/CN112633092B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013092516A (en) * | 2011-10-06 | 2013-05-16 | Asia Air Survey Co Ltd | Laser ortho-image generation device, laser ortho-image generation program, and laser ortho-image generation method |
| CN106199557A (en) * | 2016-06-24 | 2016-12-07 | 南京林业大学 | A kind of airborne laser radar data vegetation extracting method |
| CN107679458A (en) * | 2017-09-07 | 2018-02-09 | 中国地质大学（武汉） | The extracting method of roadmarking in a kind of road color laser point cloud based on K Means |
| CN108062517A (en) * | 2017-12-04 | 2018-05-22 | 武汉大学 | Unstructured road boundary line extraction method based on vehicle-mounted laser point cloud |
| CN208432477U (en) * | 2018-07-27 | 2019-01-25 | 西南交通大学 | The Experimental mimic system of Foundation Base in Railway Tunnel structure |
| CN109272458A (en) * | 2018-08-10 | 2019-01-25 | 河海大学 | A kind of point cloud filtering method based on prior information |
| CN110502973A (en) * | 2019-07-05 | 2019-11-26 | 同济大学 | A kind of roadmarking automation extraction and recognition methods based on vehicle-mounted laser point cloud |
Non-Patent Citations (3)
| Title |
|---|
| An Adaptive Multifeature Method for Semiautomatic Road Extraction From High-Resolution Stereo Mapping Satellite Images;H. Pan;《in IEEE Geoscience and Remote Sensing Letters》;20181015;201-205 * |
| 城市交通动态研判应用技术研究;杨洪;《中国优秀博硕士学位论文全文数据库(硕士)工程科技Ⅱ辑》;20131115(第11期);C034-292 * |
| 面向对象的移动激光扫描点云数据提取城市道路信息方法研究;房华乐;《中国优秀博硕士学位论文全文数据库(硕士)基础科学辑》;20150215(第02期);A008-162 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112633092A (en) | 2021-04-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112633092B (en) | Road information extraction method based on vehicle-mounted laser scanning point cloud | |
| CN106204547B (en) | The method that rod-shaped atural object spatial position is automatically extracted from Vehicle-borne Laser Scanning point cloud | |
| CN108062517B (en) | Vehicle-mounted laser point cloud-based method for automatically extracting boundary line of unstructured road | |
| CN111192284B (en) | Vehicle-mounted laser point cloud segmentation method and system | |
| CN110084272B (en) | Cluster map creation method and repositioning method based on cluster map and position descriptor matching | |
| CN107516077B (en) | Traffic sign information extraction method based on fusion of laser point cloud and image data | |
| CN112070769B (en) | Layered point cloud segmentation method based on DBSCAN | |
| CN106022259B (en) | A kind of mountain road extracting method based on laser point cloud three-dimensional feature descriptive model | |
| CN111079611B (en) | Automatic extraction method for road surface and marking line thereof | |
| CN111325138B (en) | Road boundary real-time detection method based on point cloud local concave-convex characteristics | |
| CN108845569A (en) | Generate semi-automatic cloud method of the horizontal bend lane of three-dimensional high-definition mileage chart | |
| CN112131963B (en) | Road identification line extraction method based on driving direction structural feature constraint | |
| CN110532963B (en) | Vehicle-mounted laser radar point cloud driven road marking accurate extraction method | |
| CN105718872A (en) | Auxiliary method and system for rapid positioning of two-side lanes and detection of deflection angle of vehicle | |
| CN114119863A (en) | Method for automatically extracting street tree target and forest attribute thereof based on vehicle-mounted laser radar data | |
| CN108665466A (en) | Pavement disease extraction method in a kind of road laser point cloud | |
| CN112561944A (en) | Lane line extraction method based on vehicle-mounted laser point cloud | |
| WO2023060632A1 (en) | Street view ground object multi-dimensional extraction method and system based on point cloud data | |
| CN112184725B (en) | Method for extracting center of structured light bar of asphalt pavement image | |
| CN105069395A (en) | Road marking automatic identification method based on terrestrial three-dimensional laser scanning technology | |
| CN114722944A (en) | Point cloud precision determination method, electronic device and computer storage medium | |
| CN117115012A (en) | Road surface point cloud marking segmentation denoising method | |
| CN114170149A (en) | Road geometric information extraction method based on laser point cloud | |
| CN111323026B (en) | Ground filtering method based on high-precision point cloud map | |
| CN116977970A (en) | Road drivable area detection method based on fusion of laser radar and millimeter wave radar |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |