Road edge detection method and device based on non-repetition scanning laser radar

Abstract

The present application relates to the technical field of data processing, and particularly provides a road edge detection method and device based on a non-repeating scanning laser radar, aiming at solving the problem of poor robustness and stability of the existing road edge detection method based on the non-repeating scanning laser radar. For this purpose, the road edge detection method comprises: obtaining initial point cloud data of the non-repeating scanning laser radar; first, candidate road edge line segments are screened out from the initial point cloud data of the non-repeating scanning laser radar, so as to facilitate subsequent extraction of road edge points and filtering out of the interference of non-road edge obstacles; then, first corner point cloud data and second corner point cloud data are determined through the elevation feature and gradient feature of each candidate point cloud data in the candidate road edge line segments; and further, road edge points are extracted from the candidate point cloud data based on the height difference of the first corner point cloud data and the second corner point cloud data, which is conducive to improving the effectiveness of the determined road edge points and improving the robustness and stability of the road edge detection method.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and specifically provides a roadside detection method and device based on a non-repetitive scanning laser radar. Background Art

[0002] Curb detection is a key environmental perception function for autonomous vehicles. It provides road boundary information, enabling them to constrain their driving area based on this information. LiDAR (Lidar), a key sensor for autonomous vehicles, is commonly used for curb detection in related technologies. However, existing curb detection methods are primarily based on mechanically rotating LiDARs. With technological advancements, non-repeating scanning LiDARs have also gained traction. However, there are relatively few curb detection methods for this type of LiDAR. Furthermore, non-repeating scanning LiDARs lack a fixed number and angle of scanning beams, and their scanning paths never repeat. Therefore, existing curb detection algorithms based on mechanically rotating LiDARs cannot be easily adapted to curb detection methods based on non-repeating scanning LiDARs. Furthermore, the uneven distribution of beams and large gaps in non-repeating scanning LiDARs complicates the robustness and stability of curb detection methods based on non-repeating scanning LiDARs.

[0003] Therefore, it is necessary to provide a roadside detection method suitable for non-repetitive scanning laser radar. Summary of the Invention

[0004] The present invention aims to solve the above technical problems, that is, to solve the problem that the existing road edge detection method based on non-repetitive scanning laser radar has poor robustness and stability.

[0005] In a first aspect, the present invention provides a roadside detection method based on a non-repetitive scanning laser radar, comprising:

[0006] Obtaining initial point cloud data of a non-repeated scanning laser radar in a three-dimensional coordinate system with the driving device as the origin;

[0007] Based on a dynamic sliding window algorithm, a plurality of candidate roadside line segments are screened from the non-repetitive scanning laser radar initial point cloud data according to the maximum height value, the minimum height value in the height direction of the non-repetitive scanning laser radar initial point cloud data in each sliding window and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data, where the height direction is the direction perpendicular to the ground and upward in the three-dimensional coordinate system;

[0008] Determining the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment;

[0009] Determine the final output feature of each candidate point cloud data in each candidate roadside segment based on the elevation feature and the gradient feature;

[0010] Filter the first corner point cloud data and the second corner point cloud data of each candidate road segment according to the final output feature;

[0011] Determine a curb point based on a difference in height between the first corner point cloud data and the second corner point cloud data respectively corresponding to the height direction;

[0012] A wayside line is determined based on the wayside points.

[0013] In some embodiments, determining the curb point based on the difference in heights corresponding to the first corner point cloud data and the second corner point cloud data in the height direction includes:

[0014] Obtaining a difference in height between the first corner point cloud data and the second corner point cloud data respectively corresponding to the height direction;

[0015] If the difference is within the first height difference threshold range, the first corner point cloud data, the second corner point cloud data and the candidate point cloud data corresponding to the candidate roadside segment located between the first corner point cloud data and the second corner point cloud data are determined as the roadside points.

[0016] In some embodiments, based on a dynamic sliding window algorithm, multiple candidate roadside segments are screened from the non-repeating scanning laser radar initial point cloud data according to the maximum height value, the minimum height value in the height direction of the non-repeating scanning laser radar initial point cloud data in each sliding window and the maximum distance between two adjacent non-repeating scanning laser radar initial point cloud data, including:

[0017] Respectively determine the maximum height and minimum height of the non-repetitive scanning laser radar initial point cloud data in the height direction within each sliding window, the height difference between the maximum height and the minimum height of the non-repetitive scanning laser radar initial point cloud data in the height direction, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data;

[0018] When the maximum height of the non-repetitive scanning laser radar initial point cloud data in the sliding window in the height direction is within a height threshold range, the height difference between the maximum height and the minimum height in the height direction of the initial point cloud data is within a second height difference threshold range, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data is within a distance threshold range, the non-repetitive scanning laser radar initial point cloud data in the sliding window is determined as the candidate point cloud data;

[0019] A plurality of candidate roadside line segments are determined based on the candidate point cloud data.

[0020] In some embodiments, determining the final output feature of each candidate point cloud data in each candidate roadside segment based on the elevation feature and the gradient feature includes:

[0021] Performing weighted summation on the elevation feature and the gradient feature to determine the final output feature of the candidate point cloud data;

[0022] or,

[0023] Performing sliding window filtering on the elevation feature and the gradient feature respectively;

[0024] Determining the final output feature of the candidate point cloud data based on a weighted summation of the elevation feature and the gradient feature after sliding window filtering;

[0025] or,

[0026] Calculating the first derivative of the gradient feature relative to the sliding window length;

[0027] Performing a weighted summation of the first-order derivative of the gradient feature relative to the sliding window length and the elevation feature to determine the final output feature of the candidate point cloud data;

[0028] or,

[0029] Performing a first sliding window filter on the elevation feature and the gradient feature respectively;

[0030] Calculating the first-order derivative of the gradient feature after the first sliding window filtering relative to the sliding window length and performing a second sliding window filtering on the derivative result;

[0031] The first-order derivative of the gradient feature after the second sliding window filtering relative to the sliding window length and the elevation feature after the first sliding window filtering are weighted summed to determine the final output feature of the candidate point cloud data.

[0032] In some embodiments, the three-dimensional coordinate system has a y-axis direction parallel to the ground and pointing from the origin to the direction of the traveling device, a z-axis direction passing through the origin and perpendicular to the ground and upward, and an x-axis direction perpendicular to the plane containing the y-axis and the z-axis and satisfying the right-hand rule. Determining the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment includes:

[0033] The dynamic sliding window algorithm is used to determine the elevation feature of each candidate point cloud data in each candidate road segment using the following expression:

[0034]

[0035] Among them, H i is the elevation feature, z j+1 The z-axis coordinate corresponding to the j+1th point, W is the sliding window length;

[0036] and,

[0037] The dynamic sliding window algorithm is used to determine the gradient feature of each candidate point cloud data in each candidate road segment by the following expression:

[0038]

[0039] Among them G i is the gradient feature, z j+1 is the z-axis coordinate of the j+1th point, x j+1 is the x-axis coordinate of the j+1th point, y j+1 is the y-axis coordinate of the j+1th point, and W is the length of the sliding window.

[0040] In some embodiments, after selecting multiple candidate roadside segments from the non-repetitive scanning laser radar initial point cloud data based on the dynamic sliding window algorithm and the maximum height value, the minimum height value, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data in the height direction within each sliding window, the method further includes:

[0041] Determine whether the candidate roadside line segment belongs to the left roadside or the right roadside according to the centroid distribution of the candidate roadside point set of each candidate roadside line segment;

[0042] The candidate roadside line segment is divided into a left candidate roadside line segment and a right candidate roadside line segment according to the judgment result, so as to respectively determine a left roadside point based on the left candidate roadside line segment and a right roadside point based on the right candidate roadside line segment; and the determining of the roadside based on the roadside points includes:

[0043] The left curb point and the right curb point are fitted respectively using a random sampling consistency algorithm to determine the curb line.

[0044] In some embodiments, before determining a wayside line based on the wayside point, the method further includes:

[0045] Performing grid division on the roadside points in each candidate roadside line segment belonging to the same side roadside;

[0046] Determining the number of valid grids corresponding to the roadside point in each candidate roadside segment after grid division and the total valid grid range of the valid grids determined by all the candidate roadside segments belonging to the same side roadside, wherein when the projection of the roadside point falls into the area of ​​the grid, the grid is a valid grid;

[0047] When the number of valid grids meets a preset number condition and the total valid grid range meets a preset grid range threshold, the roadside points in each candidate roadside line segment belonging to the same side roadside are determined to be valid.

[0048] In a second aspect, the present invention provides a roadside detection device based on a non-repetitive scanning laser radar, comprising:

[0049] An acquisition module, which is used to acquire initial point cloud data of a non-repeated scanning laser radar in a three-dimensional coordinate system with the driving device as the origin;

[0050] a screening module for screening a plurality of candidate roadside line segments from the non-repetitive scanning laser radar initial point cloud data based on a dynamic sliding window algorithm and a maximum height value, a minimum height value, and a maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data in a height direction within each sliding window, wherein the height direction is a direction perpendicular to the ground and upward in the three-dimensional coordinate system;

[0051] An analysis module is configured to determine the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment; and determine the final output features of each candidate point cloud data in each candidate roadside segment based on the elevation features and the gradient features;

[0052] An extraction module, configured to filter the first corner point cloud data and the second corner point cloud data of each candidate roadside segment according to the final output feature; and determine a roadside point based on a difference in heights corresponding to the first corner point cloud data and the second corner point cloud data in the height direction;

[0053] A detection module is used to determine a wayside line based on the wayside point.

[0054] In a third aspect, the present application provides a computer readable storage medium, having stored therein a computer program, which, when executed by a processor, implements the road edge detection method based on the non-repeating scanning laser radar according to any one of the preceding aspects.

[0055] In a fourth aspect, the present application provides an electronic device comprising a memory and a processor, the memory having stored therein a computer program, which, when executed by the processor, implements the road edge detection method based on the non-repeating scanning laser radar according to any one of the preceding aspects.

[0056] In the above technical solution, the present application can obtain the initial point cloud data of the non-repeating scanning laser radar, filter out a plurality of candidate road edge line segments from the initial point cloud data of the non-repeating scanning laser radar, determine the final output features of the candidate point cloud data from the elevation features and gradient features of each candidate point cloud data in each candidate road edge line segment, and determine the first corner point cloud data and the second corner point cloud data from the candidate point cloud data based on the final output features of the candidate point cloud data, and then determine the road edge point according to the difference in height corresponding to the first corner point cloud data and the second corner point cloud data in the direction perpendicular to the ground upward in the three-dimensional coordinate system. Finally, the road edge line is determined based on the road edge point. This method first filters out the candidate road edge line segments from the initial point cloud image of the non-repeating scanning laser radar, which facilitates the subsequent extraction of the road edge point and can filter out the interference of non-road edge obstacles. Then, the final output features of the candidate point cloud data are determined from the elevation features and gradient features of each candidate point cloud data, the first corner point cloud data and the second corner point cloud data are determined based on the final output features, and the road edge point is extracted from the candidate point cloud data of the candidate road edge line segment based on the difference in height of the first corner point cloud data and the second corner point cloud data, which is conducive to improving the effectiveness of the determined road edge point and the robustness and stability of the road edge detection method.

[0057] In another aspect, the present application can further perform grid division on the road edge points in each candidate road edge line segment belonging to the same side road edge before determining the road edge line based on the road edge points, so as to determine the effectiveness of the road edge points based on whether the number of effective grids after grid division and the total effective grid range meet a preset condition, thereby further improving the accuracy of road edge point detection and the robustness and stability of road edge detection. BRIEF DESCRIPTION OF DRAWINGS

[0058] The preferred embodiments of the present application will be described below with reference to the accompanying drawings, in which:

[0059] Figure 1 is a flowchart of a road edge detection method based on a non-repeating scanning laser radar provided by an embodiment of the present application;

[0060] Figure 2 is a partial schematic diagram of the initial point cloud image provided by the present invention;

[0061] Figure 3 1 is a flow chart of a method for screening candidate roadside segments provided by an embodiment of the present invention;

[0062] Figure 4 The point cloud image after the candidate roadside segments are screened out provided by the present invention;

[0063] Figure 5 1 is a flow chart of a method for determining a roadside point in a candidate roadside segment provided by an embodiment of the present invention;

[0064] Figure 6 It is a point cloud image after the roadside points are determined provided by the present invention;

[0065] Figure 7 It is a schematic diagram of a roadside provided by the present invention;

[0066] Figure 8 1 is a flow chart of a roadside detection method based on a non-repetitive scanning laser radar provided in another embodiment of the present invention;

[0067] Figure 9 1 is a schematic structural diagram of a roadside detection device based on a non-repetitive scanning laser radar provided in an embodiment of the present invention;

[0068] Figure 10 It is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0069] Some embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0070] Curb detection is a key environmental perception function for autonomous vehicles. It provides road boundary information, enabling them to constrain their driving area based on this information. LiDAR (Lidar), a key sensor for autonomous vehicles, is commonly used for curb detection in related technologies. However, existing curb detection methods are primarily based on mechanically rotating LiDARs. With technological advancements, non-repeating scanning LiDARs have also gained traction. However, there are relatively few curb detection methods for this type of LiDAR. Furthermore, non-repeating scanning LiDARs lack a fixed number and angle of scanning beams, and their scanning paths never repeat. Therefore, existing curb detection algorithms based on mechanically rotating LiDARs cannot be easily adapted to curb detection methods based on non-repeating scanning LiDARs. Furthermore, the uneven distribution of beams and large gaps in non-repeating scanning LiDARs complicates the robustness and stability of curb detection methods based on non-repeating scanning LiDARs.

[0071] In view of this, the roadside detection method based on non-repetitive scanning laser radar provided by the present invention can obtain the initial point cloud data of the non-repetitive scanning laser radar; screen out multiple candidate roadside segments from the initial point cloud data of the non-repetitive scanning laser radar; determine the final output features of the candidate point cloud data by the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment, and determine the first corner point cloud data and the second corner point cloud data from the candidate point cloud data based on the final output features of the candidate point cloud data, and then determine the roadside point according to the difference in height corresponding to the first corner point cloud data and the second corner point cloud data in the direction perpendicular to the ground in the three-dimensional coordinate system; finally, determine the roadside line based on the roadside point. The method first screens out candidate roadside segments from the initial point cloud image of a non-repetitive scanning laser radar, which facilitates the subsequent extraction of roadside points and can filter out the interference of non-roadside obstacles. The method then determines the final output features of the candidate point cloud data through the elevation features and gradient features of each candidate point cloud data, determines the first corner point cloud data and the second corner point cloud data based on the final output features, and further extracts the roadside points from the candidate point cloud data of the candidate roadside segment based on the difference in height between the first corner point cloud data and the second corner point cloud data. This method is beneficial to improving the effectiveness of the determined roadside points and improving the robustness and stability of the roadside detection method.

[0072] See also Figure 1 As shown, Figure 1 : This is a flow chart of a roadside detection method based on a non-repetitive scanning laser radar provided by an embodiment of the present invention, which may include:

[0073] Step S11: Acquire initial point cloud data of a non-repeated scanning laser radar in a three-dimensional coordinate system with the driving device as the origin;

[0074] Step S12: Based on a dynamic sliding window algorithm, multiple candidate roadside segments are screened from the non-repeating scanning laser radar initial point cloud data according to the maximum height value, the minimum height value in the height direction of the non-repeating scanning laser radar initial point cloud data within each sliding window, and the maximum distance between two adjacent non-repeating scanning laser radar initial point cloud data, where the height direction is the direction perpendicular to the ground and upward in the three-dimensional coordinate system;

[0075] Step S13: determining the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment;

[0076] Step S14: determining the final output features of each candidate point cloud data in each candidate roadside segment based on the elevation features and the gradient features;

[0077] Step S15: filtering the first corner point cloud data and the second corner point cloud data of each candidate roadside segment according to the final output features;

[0078] Step S16: determining a curb point based on a height difference between the first corner point cloud data and the second corner point cloud data in a height direction;

[0079] Step S17: Determine the roadside line based on the roadside points.

[0080] In some embodiments, step S11 can be specifically to use a non-repetitive scanning laser radar to scan the road to obtain non-repetitive scanning laser radar initial point cloud data. The collected non-repetitive scanning laser radar initial point cloud data can be projected into a three-dimensional coordinate system established with the driving device as the origin.

[0081] In some embodiments, the driving device may be equipped with multiple non-repetitive scanning laser radars. Step S11 may specifically be to obtain the initial non-repetitive scanning laser radar point cloud data collected by the multiple non-repetitive scanning laser radars in a three-dimensional coordinate system with the driving device as the origin. In addition, the initial non-repetitive scanning laser radar point cloud data collected by the multiple non-repetitive scanning laser radars may be spliced ​​to obtain an initial point cloud image. Each non-repetitive scanning laser radar may correspond to a continuous scan line in the shape of a petal. For details, see Figure 2 As shown, Figure 2 A partial schematic diagram shows an initial point cloud image obtained by stitching together initial point cloud data collected by multiple non-repeating scanning laser radars.

[0082] As an example, a three-dimensional coordinate system can be established in the following manner: with the travel device as the origin, the y-axis direction parallel to the ground and pointing from the origin toward the travel device is the y-axis direction, the z-axis direction passing through the origin and perpendicular to the ground is the z-axis direction, and the x-axis direction is the direction perpendicular to the plane containing the y-axis and z-axis and satisfying the right-hand rule is the x-axis direction. In other embodiments, other methods can also be used to establish the three-dimensional coordinate system.

[0083] In some embodiments, see Figure 3 As shown, step S12 may be specifically as follows:

[0084] Step S121: determining the maximum height and minimum height of the non-repetitive scanning laser radar initial point cloud data in the height direction within each sliding window, the height difference between the maximum height and the minimum height of the non-repetitive scanning laser radar initial point cloud data in the height direction, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data;

[0085] Step S122: When the maximum height of the non-repetitive scanning laser radar initial point cloud data in the sliding window in the height direction is within the height threshold range, the height difference between the maximum height and the minimum height in the height direction of the initial point cloud data is within the second height difference threshold range, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data is within the distance threshold range, the non-repetitive scanning laser radar initial point cloud data in the sliding window is determined as candidate point cloud data;

[0086] Step S123: Determine multiple candidate roadside segments based on the candidate point cloud data.

[0087] Taking the three-dimensional coordinate system constructed in step S11 above as an example, the height direction is the z-axis direction in the three-dimensional coordinate system. In step S121, the maximum height of the non-repeating scanning laser radar initial point cloud data within the sliding window can be represented as z_max, and the minimum height can be represented as z_min. The height difference between the maximum and minimum heights of the non-repeating scanning laser radar initial point cloud data in the z-axis direction is z_diff = |z_max - z_min|. The maximum distance between two adjacent non-repeating scanning laser radar initial point cloud data can be represented as dis_max. Whether the non-repeating scanning laser radar initial point cloud data within the sliding window is candidate point cloud data is determined by determining whether z_max is within a height threshold range, whether z_diff is within a second height difference threshold range, and whether dis_max is within a distance threshold range. If z_max is within the height threshold range, z_diff is within the second height difference threshold range, and dis_max is within the distance threshold range, the non-repeating scanning laser radar initial point cloud data within the sliding window can be determined as candidate point cloud data. By constructing a candidate roadside point set based on the candidate point cloud data determined by each sliding window, multiple candidate roadside segments can be determined. Figure 4 As shown in the figure, the white segments within the rectangular box represent the multiple candidate roadside segments identified. Since the roadside has a certain height, candidate roadside segments are determined by considering whether z_max is within a height threshold and whether z_diff is within a second height difference threshold, eliminating interference from obstacle points. Furthermore, due to the scanning characteristics of non-repeating scanning LiDAR, the acquired initial point cloud data is orderly. Based on this, candidate roadside segments are determined by considering whether dis_max is within a spacing threshold, eliminating some non-contiguous point sets.

[0088] In an embodiment of the present invention, a petal-shaped continuous scanning line collected by a non-repetitive scanning laser radar can be divided into multiple candidate roadside line segments through step S12 to filter out interference from non-roadside obstacles, thereby facilitating the determination of subsequent roadside points.

[0089] In some embodiments, based on the three-dimensional coordinate system established in step S11, step S13 may specifically be:

[0090] The dynamic sliding window algorithm is used to determine the elevation features of each candidate point cloud data in each candidate road segment using the following expression:

[0091]

[0092] Among them, H i is the elevation feature, z j+1 The z-axis coordinate corresponding to the j+1th point, W is the length of the sliding window;

[0093] and,

[0094] Using the dynamic sliding window algorithm, the gradient characteristics of each candidate point cloud data in each candidate road segment are determined by the following expression:

[0095]

[0096] Among them G i is the gradient feature, z j+1 is the z-axis coordinate of the j+1th point, x j+1 is the x-axis coordinate of the j+1th point, y j+1 is the y-axis coordinate of the j+1th point, and W is the length of the sliding window.

[0097] In some embodiments, step S14 may be specifically as follows: performing weighted summation of the elevation feature and the gradient feature to determine the final output feature of the candidate point cloud data. The final output feature can be expressed as T i =αH i +βG i , where α is the weight corresponding to the elevation feature, and β is the weight corresponding to the gradient feature.

[0098] In some embodiments, in order to remove noise in the measurement, step S14 may further include: performing sliding window filtering on the elevation features and gradient features respectively; performing weighted summation based on the elevation features and gradient features after sliding window filtering to determine the final output features of the candidate point cloud data.

[0099] In some other embodiments, step S14 may further include:

[0100] Calculate the first derivative of the gradient feature relative to the sliding window length;

[0101] The first-order derivative of the gradient feature relative to the sliding window length and the elevation feature are weighted summed to determine the final output feature of the candidate point cloud data.

[0102] In some other embodiments, step S14 may further include:

[0103] Perform the first sliding window filtering on the elevation features and gradient features respectively;

[0104] Calculate the first-order derivative of the gradient feature after the first sliding window filtering relative to the sliding window length and perform a second sliding window filtering on the derivative result;

[0105] The final output features of the candidate point cloud data are determined by weighted summing of the first-order derivative of the gradient feature after the second sliding window filtering relative to the sliding window length and the elevation feature after the first sliding window filtering.

[0106] In some embodiments, step S15 can be specifically: traversing from the head of the candidate road edge line segment and the tail of the candidate road edge line segment respectively for each candidate road edge line segment, comparing the final output feature corresponding to each candidate point cloud data with the feature threshold value, and determining the candidate point cloud data with the final output feature greater than the feature threshold value as the first corner point cloud data when traversing from the head of the candidate road edge line segment; comparing the final output feature corresponding to each candidate point cloud data with the feature threshold value, and determining the candidate point cloud data with the final output feature greater than the feature threshold value as the second corner point cloud data when traversing from the tail of the candidate road edge line segment.

[0107] In some embodiments, referring to FIG. 7, step S16 can be specifically: Figure 5

[0108] Step S161: obtaining the difference of the heights corresponding to the first corner point cloud data and the second corner point cloud data in the height direction respectively.

[0109] Step S162: if the difference is within the first height difference threshold range, determining the first corner point cloud data, the second corner point cloud data, and the candidate point cloud data corresponding to the candidate road edge line segment and located between the first corner point cloud data and the second corner point cloud data as the road edge points.

[0110] Based on the three-dimensional coordinate system established in step S11 above, the height direction can be the z-axis direction, the height corresponding to the first corner point cloud data in the z-axis direction can be represented as z head , and the height corresponding to the second corner point cloud data in the z-axis direction can be represented as z tail .

[0111] In step S161, the difference of the heights corresponding to the first corner point cloud data and the second corner point cloud data in the z-axis direction can be obtained by the following expression: z diff = |z tail -z head |.

[0112] In some embodiments, step S162 can be specifically: when z diff is within the first height difference threshold range, determining the first corner point cloud data, the second corner point cloud data, and the candidate point cloud data corresponding to the candidate road edge line segment and located between the first corner point cloud data and the second corner point cloud data as the road edge points, referring to FIG. 7, the dark gray points in the oval frame are the determined multiple road edge points. By collecting the determined multiple road edge points, a road edge point set can be obtained. Figure 6

[0113] In some embodiments, step S17 can be specifically:

[0114] ​​According to the centroid distribution of the roadside point set, the roadside points are divided into left roadside points and right roadside points;

[0115] The random sampling consensus algorithm is used to fit the left and right curb points respectively to determine the roadside line. Figure 7 shown.

[0116] In the embodiment of the present invention, by adopting the random sampling consistency algorithm, the point cloud density can be balanced, the fitting accuracy can be improved, and the problem of the fitting curve being biased towards the densely distributed area due to the uneven distribution of the determined roadside points can be avoided.

[0117] In some embodiments, after selecting multiple candidate roadside segments from the initial point cloud data of the non-repeated scanning laser radar, the following method may be further included:

[0118] According to the centroid distribution of the candidate roadside point set of each candidate roadside segment, determine whether the candidate roadside segment belongs to the left roadside or the right roadside;

[0119] According to the judgment result, the candidate roadside line segments are divided into left candidate roadside line segments and right candidate roadside line segments, so as to determine the left roadside point based on the left candidate roadside line segments and the right roadside point based on the right candidate roadside line segments respectively.

[0120] Based on this, in step S17 , the left side roadside point and the right side roadside point can be directly fitted to determine the roadside line.

[0121] It should be noted that in other embodiments, after obtaining the non-repetitive scanning laser radar initial point cloud image in step S11, the initial point cloud data can be divided into non-repetitive scanning laser radar initial point cloud data corresponding to the left roadside and non-repetitive scanning laser radar initial point cloud data corresponding to the right roadside according to the coordinates of the non-repetitive scanning laser radar initial point cloud data in the three-dimensional coordinate system, and steps S12-S16 are executed respectively based on the data on each side after division to obtain the left roadside points and the right roadside points. In step S17, there is no need to divide the roadside points, and fitting can be performed directly based on the obtained left roadside points and right roadside points.

[0122] The above is the roadside detection method based on non-repetitive scanning laser radar provided by the present invention, which can obtain the initial point cloud data of the non-repetitive scanning laser radar; screen out multiple candidate roadside segments from the initial point cloud data of the non-repetitive scanning laser radar; determine the final output features of the candidate point cloud data by the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment, and determine the first corner point cloud data and the second corner point cloud data from the candidate point cloud data based on the final output features of the candidate point cloud data, and then determine the roadside point according to the difference in height corresponding to the first corner point cloud data and the second corner point cloud data in the direction perpendicular to the ground in the three-dimensional coordinate system; finally, determine the roadside line based on the roadside point. The method first screens out candidate roadside segments from the initial point cloud image of a non-repetitive scanning laser radar, which facilitates the subsequent extraction of roadside points and can filter out the interference of non-roadside obstacles. The method then determines the final output features of the candidate point cloud data through the elevation features and gradient features of each candidate point cloud data, determines the first corner point cloud data and the second corner point cloud data based on the final output features, and further extracts the roadside points from the candidate point cloud data of the candidate roadside segment based on the difference in height between the first corner point cloud data and the second corner point cloud data. This method is beneficial to improving the effectiveness of the determined roadside points and improving the robustness and stability of the roadside detection method.

[0123] See also Figure 8 As shown, Figure 8 FIG. 5 is a flow chart of a roadside detection method based on a non-repeating scanning laser radar provided in another embodiment of the present invention, which may include:

[0124] Step S21: Acquire non-repeated scanning initial point cloud data in a three-dimensional coordinate system with the driving device as the origin;

[0125] Step S22: Based on a dynamic sliding window algorithm, a plurality of candidate roadside segments are screened from the non-repeating scanning laser radar initial point cloud data according to the maximum height value, the minimum height value in the height direction of the non-repeating scanning laser radar initial point cloud data within each sliding window, and the maximum distance between two adjacent non-repeating scanning laser radar initial point cloud data, where the height direction is the direction perpendicular to the ground and upward in the three-dimensional coordinate system;

[0126] Step S23: determining the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment;

[0127] Step S24: determining the final output features of each candidate point cloud data in each candidate roadside segment based on the elevation features and the gradient features;

[0128] Step S25: filtering the first corner point cloud data and the second corner point cloud data of each candidate roadside segment according to the final output features;

[0129] Step S26: determining a curb point based on a height difference between the first corner point cloud data and the second corner point cloud data in a height direction;

[0130] Step S27: performing grid division on the roadside points in each candidate roadside line segment belonging to the same side roadside;

[0131] Step S28: determining the number of valid grids corresponding to the roadside points in each candidate roadside segment after grid division and the total valid grid range of the valid grids determined by the candidate roadside segments belonging to the same side of the roadside, wherein a grid is considered valid when the projection of the roadside point falls within the grid area;

[0132] Step S29: when the number of valid grids meets the preset number condition and the total valid grid range meets the preset grid range threshold, determining that the roadside points in each candidate roadside segment belonging to the same side roadside are valid;

[0133] Step S30: Determine the roadside line based on the roadside points.

[0134] In some embodiments, when the number of valid grids does not meet a preset number condition and / or the total valid grid range does not meet a preset grid range threshold, the roadside point is determined to be invalid and subsequent roadside line fitting is no longer performed based on the roadside point.

[0135] Among them, steps S21-S26 and step S30 can be implemented in the same manner as steps S11-S17 in the above embodiment. For the sake of brevity, they are not repeated here. For details, please refer to the above description.

[0136] In some embodiments, after the candidate roadside segments are screened out in step S22, they can be divided into: candidate roadside segments corresponding to the left roadside and candidate roadside segments corresponding to the right roadside according to the coordinates of the candidate roadside point cloud data in the candidate roadside segments.

[0137] In some embodiments, for each candidate roadside line belonging to the same side of the roadside, step S27 may specifically set the length of each grid area and divide the space of interest, such as the space corresponding to the roadside points in the candidate roadside line segment, into a grid. For each roadside point, the corresponding coordinates after the roadside point is projected onto the grid are calculated using the following formula:

[0138] bin i =[(y i -y min ) / L], where bin i Represents the corresponding coordinates of the i-th roadside point after projection to the grid, y iis the coordinate of the i-th roadside point in the y-axis direction in the three-dimensional coordinate system established based on the above step S21, and L represents the length of the grid area.

[0139] In some embodiments, it is possible to determine whether there is a projection point of the roadside point in the corresponding grid based on the corresponding coordinates of the roadside point after being projected onto the grid. If so, the grid is determined to be a valid grid.

[0140] In some embodiments, it can be determined whether the number of valid grids is greater than a preset number and whether the total valid grid range is within a preset grid range threshold. If so, the roadside points in each candidate roadside line segment belonging to the same side roadside are determined to be valid, and the roadside line can be determined based on the valid roadside points.

[0141] This method combines the curb length and curb distribution characteristics to judge the effectiveness of curb points, which can improve the accuracy of curb point detection as well as the robustness and stability of curb detection.

[0142] Another aspect of the present invention provides a roadside detection device based on a non-repeating scanning laser radar, see Figure 9 As shown, Figure 9 : is a schematic structural diagram of a roadside detection device based on a non-repetitive scanning laser radar provided by an embodiment of the present invention, which may include:

[0143] An acquisition module 41 is used to acquire non-repeated scanning initial point cloud data in a three-dimensional coordinate system with the driving device as the origin;

[0144] a screening module 42 for screening a plurality of candidate roadside segments from the non-repeating scanning laser radar initial point cloud data based on a dynamic sliding window algorithm and a maximum height value, a minimum height value, and a maximum distance between two adjacent non-repeating scanning laser radar initial point cloud data in a height direction within each sliding window, wherein the height direction is a direction perpendicular to the ground and upward in a three-dimensional coordinate system;

[0145] An analysis module 43 is configured to determine the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment; and to determine the final output features of each candidate point cloud data in each candidate roadside segment based on the elevation features and gradient features;

[0146] An extraction module 44 is configured to filter the first corner point cloud data and the second corner point cloud data of each candidate roadside segment according to the final output feature; and determine the roadside point based on the difference in height between the first corner point cloud data and the second corner point cloud data in the height direction;

[0147] The detection module 45 is configured to determine a roadside line based on the roadside points.

[0148] The roadside detection device based on non-repetitive scanning laser radar provided by the present invention can be used to execute the above-mentioned roadside detection method based on non-repetitive scanning laser radar, and achieve the same beneficial effects as the roadside detection method based on non-repetitive scanning laser radar in the above-mentioned embodiment. Furthermore, it should be understood that since the setting of each module is only for the purpose of illustrating the functional units of the device of the present invention, the physical devices corresponding to these modules can be the processor itself, or a part of the software in the processor, a part of the hardware, or a part of the combination of software and hardware. Therefore, Figure 9 The number of modules in the device is merely illustrative. Those skilled in the art will appreciate that the modules in the device can be adaptively split or merged. Such splitting or merging of specific modules does not cause the technical solution to deviate from the principles of the present invention. Therefore, the resulting technical solutions will fall within the scope of protection of the present invention.

[0149] It will be understood by those skilled in the art that all or part of the processes in the method for implementing the above embodiment of the present invention may also be completed by instructing the relevant hardware through a computer program. The computer program may be stored in a computer-readable storage medium. When the computer program is executed by a processor, it may implement the steps of each of the above method embodiments. The computer program includes computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable storage medium may include: any entity or device, medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electric carrier signal, telecommunication signal, and software distribution medium, etc., which can carry the computer program code.

[0150] Another aspect of the present invention provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program can implement the roadside detection method based on a non-repeating scanning laser radar described in any of the above embodiments. The computer-readable storage medium can be a storage device formed by various electronic devices. Optionally, in embodiments of the present invention, the computer-readable storage medium is non-transitory.

[0151] See also Figure 10 As shown, Figure 10 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The device may include a memory 51 and a processor 52. Memory 51 stores a computer program, including but not limited to a program for executing the methods of the aforementioned method embodiments. When executed by processor 52, the computer program can implement the roadside detection method based on a non-repeating scanning laser radar in any of the aforementioned embodiments.

[0152] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. A roadside detection method based on non-repetitive scanning laser radar, characterized in that: include: Obtaining initial point cloud data of a non-repeated scanning laser radar in a three-dimensional coordinate system with the driving device as the origin; Based on a dynamic sliding window algorithm, a plurality of candidate roadside line segments are screened from the non-repetitive scanning laser radar initial point cloud data according to the maximum height value, the minimum height value in the height direction of the non-repetitive scanning laser radar initial point cloud data in each sliding window and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data, where the height direction is the direction perpendicular to the ground and upward in the three-dimensional coordinate system; Determining the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment; Determine the final output feature of each candidate point cloud data in each candidate roadside segment based on the elevation feature and the gradient feature; Filter the first corner point cloud data and the second corner point cloud data of each candidate road segment according to the final output feature; Determine a curb point based on a difference in heights corresponding to the first corner point cloud data and the second corner point cloud data in the height direction and a first height difference threshold; A wayside line is determined based on the wayside points.

2. The method according to claim 1, characterized in that The determining of the curb point based on the height difference between the first corner point cloud data and the second corner point cloud data respectively corresponding to each other in the height direction and a first height difference threshold comprises: Obtaining a difference in height between the first corner point cloud data and the second corner point cloud data respectively corresponding to the height direction; If the difference is within the first height difference threshold range, the first corner point cloud data, the second corner point cloud data and the candidate point cloud data corresponding to the candidate roadside segment located between the first corner point cloud data and the second corner point cloud data are determined as the roadside point.

3. The method according to claim 1, characterized in that Based on a dynamic sliding window algorithm, multiple candidate roadside segments are screened from the non-repeating scanning laser radar initial point cloud data according to the maximum height value, the minimum height value in the height direction of the non-repeating scanning laser radar initial point cloud data in each sliding window and the maximum distance between two adjacent non-repeating scanning laser radar initial point cloud data, including: Respectively determine the maximum height and minimum height of the non-repetitive scanning laser radar initial point cloud data in the height direction, the height difference between the maximum height and the minimum height of the non-repetitive scanning laser radar initial point cloud data in the height direction, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data in each sliding window; When the maximum height of the non-repetitive scanning laser radar initial point cloud data in the sliding window in the height direction is within a height threshold range, the height difference between the maximum height and the minimum height in the height direction of the initial point cloud data is within a second height difference threshold range, and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data is within a distance threshold range, the non-repetitive scanning laser radar initial point cloud data in the sliding window is determined as the candidate point cloud data; A plurality of candidate roadside line segments are determined based on the candidate point cloud data.

4. The method according to any one of claims 1 to 3, characterized in that The determining, based on the elevation feature and the gradient feature, the final output feature of each candidate point cloud data in each candidate roadside segment comprises: Performing weighted summation on the elevation feature and the gradient feature to determine the final output feature of the candidate point cloud data; or, Performing sliding window filtering on the elevation feature and the gradient feature respectively; Determining the final output feature of the candidate point cloud data based on a weighted summation of the elevation feature and the gradient feature after sliding window filtering; or, Calculating the first derivative of the gradient feature relative to the sliding window length; Performing a weighted summation of the first-order derivative of the gradient feature relative to the sliding window length and the elevation feature to determine the final output feature of the candidate point cloud data; or, Performing a first sliding window filter on the elevation feature and the gradient feature respectively; Calculating the first-order derivative of the gradient feature after the first sliding window filtering relative to the sliding window length and performing a second sliding window filtering on the derivative result; The first-order derivative of the gradient feature after the second sliding window filtering relative to the sliding window length and the elevation feature after the first sliding window filtering are weighted summed to determine the final output feature of the candidate point cloud data.

5. The method according to claim 1, wherein The three-dimensional coordinate system has a y-axis direction parallel to the ground and pointing from the origin to the direction of the travel device, a z-axis direction passing through the origin and perpendicular to the ground and pointing upward, and an x-axis direction perpendicular to the plane containing the y-axis and z-axis and satisfying the right-hand rule. Determining the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment includes: The dynamic sliding window algorithm is used to determine the elevation feature of each candidate point cloud data in each candidate road segment using the following expression: Among them, H i is the elevation feature, z j+1 Corresponding to the z-axis coordinate of the j+1th point, W is the length of the sliding window; and, The dynamic sliding window algorithm is used to determine the gradient feature of each candidate point cloud data in each candidate road segment by the following expression: Among them G i is the gradient feature, z j+1 is the z-axis coordinate of the j+1th point, x j+1 is the x-axis coordinate of the j+1th point, y j+1 is the y-axis coordinate of the j+1th point, and W is the length of the sliding window.

6. The method according to claim 1, characterized in that After selecting a plurality of candidate roadside segments from the non-repetitive scanning laser radar initial point cloud data based on the dynamic sliding window algorithm and according to the maximum height value, the minimum height value in the height direction of the non-repetitive scanning laser radar initial point cloud data in each sliding window and the maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data, the method further includes: Determine whether the candidate roadside line segment belongs to the left roadside or the right roadside according to the centroid distribution of the candidate roadside point set of each candidate roadside line segment; Dividing the candidate roadside line segment into a left candidate roadside line segment and a right candidate roadside line segment according to the judgment result, so as to respectively determine a left roadside point based on the left candidate roadside line segment and a right roadside point based on the right candidate roadside line segment; The determining of a roadside line based on the roadside point comprises: The left curb point and the right curb point are fitted respectively using a random sampling consistency algorithm to determine the curb line.

7. The method according to claim 1 or 6, characterized in that Before determining the roadside line based on the roadside point, the method further includes: Performing grid division on the roadside points in each candidate roadside line segment belonging to the same side roadside; Determining the number of valid grids corresponding to the roadside point in each candidate roadside segment after grid division and the total valid grid range of the valid grids determined by all the candidate roadside segments belonging to the same side roadside, wherein when the projection of the roadside point falls into the area of ​​the grid, the grid is a valid grid; When the number of valid grids meets a preset number condition and the total valid grid range meets a preset grid range threshold, the roadside points in each candidate roadside line segment belonging to the same side roadside are determined to be valid.

8. A roadside detection device based on a non-repetitive scanning laser radar, characterized in that: include: An acquisition module, which is used to acquire initial point cloud data of a non-repeated scanning laser radar in a three-dimensional coordinate system with the driving device as the origin; a screening module for screening a plurality of candidate roadside line segments from the non-repetitive scanning laser radar initial point cloud data based on a dynamic sliding window algorithm and a maximum height value, a minimum height value, and a maximum distance between two adjacent non-repetitive scanning laser radar initial point cloud data in a height direction within each sliding window, wherein the height direction is a direction perpendicular to the ground and upward in the three-dimensional coordinate system; An analysis module, which is used to determine the elevation features and gradient features of each candidate point cloud data in each candidate roadside segment; and, determining a final output feature of each of the candidate point cloud data in each of the candidate roadside segments based on the elevation feature and the gradient feature; An extraction module is configured to filter the first corner point cloud data and the second corner point cloud data of each candidate roadside segment according to the final output feature; and determine a roadside point based on a difference in heights corresponding to the first corner point cloud data and the second corner point cloud data in the height direction and a first height difference threshold; A detection module is used to determine a wayside line based on the wayside point.

9. 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 the processor, it implements the roadside detection method based on non-repetitive scanning laser radar as described in any one of claims 1 to 7.

10. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the method for detecting a roadside based on a non-repetitive scanning laser radar according to any one of claims 1 to 7 is implemented.

Images