A method and apparatus for obstacle detection for a laser radar

Abstract

The application discloses a laser radar obstacle detection method and device, and the method comprises the following steps: S1, a multi-line laser radar is used to scan a region in front of a carrier, and a single-line laser radar is used to scan a blind area range of the multi-line laser radar; S2, point cloud data scanned by the multi-line laser radar is subjected to gridding processing to obtain a multi-line radar grid point cloud map, and scanning line dense distribution features in a specified range near each grid are obtained; S3, the point cloud data scanned by the single-line laser radar is subjected to auxiliary processing by using the scanning line dense distribution features of each grid obtained by the multi-line laser radar to obtain an obstacle detection result of the single-line laser radar; and S4, the obstacle detection results of the multi-line laser radar and the single-line laser radar are fused to obtain a final detection result. The application can avoid obstacle omission and reduce a false alarm rate, and can be applied to various complex environments to realize comprehensive and accurate detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of obstacle detection, and in particular to a laser radar obstacle detection method and device. BACKGROUND

[0002] The laser radar is a key sensor configuration for robot autonomous navigation, which is used to perceive the obstacle information on the road surface to avoid collision with obstacles during the travel process. The current mainstream sensing scheme is to directly use a multi-line laser radar to realize the perception of the information around the robot, and the key perceived information includes people, motor vehicles, bicycles, motorcycles, traffic signs, and other obstacles on the road surface. Since the perception of the front area information is the focus in the navigation of the robot, and the rear area also needs to be considered, the multi-line laser radar is usually installed at the top of the robot to obtain the most effective observation field of view. The above installation method will cause a certain blind area in the front, and in order to ensure the comprehensiveness of the detection and avoid missing low obstacles, other sensing means need to be supplemented. One solution is to configure a single-line radar below the front of the robot to make up for the observation blind area of the multi-line laser radar, but in this configuration scheme, the single-line laser radar and the multi-line laser radar in the prior art independently detect obstacles, that is, the single-line laser radar point cloud is directly clustered to select the target obstacle by cooperating with the geometric characteristics, and the single-line laser radar also returns point clouds for steep areas, uphill and downhill areas, and other uneven road surfaces, so that more false alarms are generated when encountering uneven road surfaces, affecting the smoothness of the travel.

[0003] For example, Chinese patent application CN112528778A discloses an obstacle detection method, device, electronic equipment and storage medium. The scheme configures the detection range of the single-line laser radar as the blind area of the multi-line laser radar, converts the single-line laser radar and the multi-line laser radar into the vehicle body coordinate system respectively, deletes the ground point cloud of the multi-line laser radar and the point cloud of the single-line laser radar beyond the preset distance threshold and the spatial azimuth angle outside the preset value, obtains the candidate obstacle point cloud, and then uses a clustering algorithm to extract the obstacle. This scheme is that the single-line laser radar and the multi-line laser radar independently detect obstacles, and the single-line laser radar is used as a blind supplement for the multi-line laser radar, which is only applicable to the scene where the robot is on a flat road surface and does not vibrate, such as an indoor environment. However, since the point cloud processing method of the single-line laser radar is to directly set the detection range threshold to filter the non-effective point cloud, if it is applied to a non-flat ground environment, the constraint range cannot be too long to meet the needs of the non-flat ground, which will result in a short blind supplement distance. In addition, for steep areas, uphill and downhill areas, and other uneven road surfaces, the point cloud of the single-line radar will be retracted towards the front of the vehicle in this scene, which will generate a large number of false alarms and is difficult to completely eliminate. SUMMARY

[0004] The technical problem to be solved by the present application is that, in view of the technical problems existing in the prior art, the present application provides a laser radar obstacle detection method and device which has simple implementation method, high detection efficiency and precision, can avoid low obstacle missing detection, and reduce false alarm rate, and can be applied to various complex environments such as steep areas, uphill and downhill areas and the like to realize comprehensive and accurate detection.

[0005] To solve the above technical problems, the technical solution provided by the present application is:

[0006] A laser radar obstacle detection method, comprising the following steps:

[0007] S1. Scanning the area in front of the carrier using a multi-line laser radar, and scanning the blind area range of the multi-line laser radar using a single-line laser radar;

[0008] S2. Rasterizing the point cloud data scanned by the multi-line laser radar to obtain a multi-line radar grid point cloud map, and obtaining the scanning line dense distribution feature in a specified range near each grid;

[0009] S3. Obtaining the obstacle detection result of the single-line laser radar by assisting the scanning line dense distribution feature of each grid obtained by the multi-line laser radar on the point cloud data scanned by the single-line laser radar;

[0010] S4. Fusing the obstacle detection result of the multi-line laser radar and the obstacle detection result of the single-line laser radar to obtain the final detection result.

[0011] Further, the step S2 comprises:

[0012] S21. Rasterizing the point cloud data scanned by the multi-line laser radar to obtain a multi-line radar grid point cloud map;

[0013] S22. Judging the continuity of the scanning lines between the grid where each point cloud in the multi-line radar grid point cloud map is located and the neighboring grid;

[0014] S23. Counting the number of scanning lines judged as continuous in each grid, and determining the scanning line dense distribution feature of each grid according to the number of scanning lines.

[0015] Further, the step S22 comprises: traversing each point cloud in the multi-line radar grid point cloud map, taking the current grid position as the reference position each time, searching for the neighboring grid of the current grid, recording the scanning line index number when the neighboring grid has scanning points, and judging the continuity of the scanning line index number of the neighboring grid and the scanning line index number of the current grid to obtain all scanning lines judged as continuous corresponding to the current grid.

[0016] Further, the step S3 comprises:

[0017] S31. Rasterizing the point cloud data scanned by the single-line laser radar to obtain single-line radar rasterized point cloud data;

[0018] S32. Iterating the single-line radar rasterized point cloud data, using the scanning line dense distribution features of the corresponding positions of each point cloud to assist in judging whether the point cloud is an obstacle, to obtain an obstacle detection result of the single-line laser radar.

[0019] Further, the step S4 comprises:

[0020] S41. Obtaining an obstacle detection result of the multi-line laser radar according to the point cloud data scanned by the multi-line laser radar;

[0021] S42. Fusing the obstacle detection result of the multi-line laser radar with the obstacle detection result of the single-line laser radar to obtain a final detection result.

[0022] Further, the step S41 comprises: determining ground points according to the point cloud data scanned by the multi-line laser radar and a pre-constructed ground model, and deleting the ground points to obtain the obstacle detection result of the multi-line laser radar.

[0023] Further, in the step S42, when the multi-line laser radar detects an obstacle within the range of the single-line laser radar, the obstacle detection result of the multi-line laser radar is adopted; when the single-line laser radar detects an obstacle but the multi-line laser radar does not detect an obstacle, the obstacle detection result detected by the single-line laser radar is added.

[0024] A laser radar obstacle detection device, comprising:

[0025] a multi-line laser radar configured to scan a region in front of a carrier;

[0026] a single-line laser radar configured to scan a blind area range of the multi-line laser radar;

[0027] an obstacle detection module configured to rasterize point cloud data scanned by the multi-line laser radar to obtain a multi-line radar rasterized point cloud map, and to obtain scanning line dense distribution features within a specified range near each grid;

[0028] and to assist in obtaining, using the scanning line dense distribution features of each grid obtained by the multi-line laser radar, an obstacle detection result of the single-line laser radar from point cloud data scanned by the single-line laser radar;

[0029] and to fuse the obstacle detection results of the multi-line laser radar and the single-line laser radar to obtain a final detection result.

[0030] Further, the multi-line laser radar and the single-line laser radar are arranged on the carrier, and the multi-line laser radar is arranged at a vertical height higher than that of the single-line laser radar.

[0031] Further, the obstacle detection module comprises a processor and a memory, the memory is used for storing a computer program, and the processor is used for executing the computer program to perform steps S2-S4.

[0032] Compared with the prior art, the present application has the advantages that: the present application supplements the multi-line laser radar with the single-line laser radar, and on this basis, the scanning line distribution characteristics of the grid point cloud graph obtained by the multi-line laser radar are used as prior knowledge to guide the scanning of the single-line laser radar, and assist the single-line laser radar in obstacle detection, which can effectively eliminate the false alarm of single-line scanning caused by non-flat ground such as steep area, uphill and downhill area, and realize adaptive scene filtering of non-obstacle points in single-line scanning, and through the fusion of multi-line radar and single-line radar, comprehensive and accurate obstacle detection can be realized. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is a flowchart of the implementation process of the laser radar obstacle detection method of the present embodiment.

[0034] Figure 2 is a complete flowchart of the implementation of the laser radar obstacle detection of the present embodiment.

[0035] Figure 3 is a detailed flowchart of the implementation of the laser radar obstacle detection in the specific application embodiment of the present application.

[0036] Figure 4 is a structural diagram of the laser radar obstacle detection device of the present embodiment. DETAILED DESCRIPTION

[0037] The present application will be further described below in combination with the drawings of the specification and specific preferred embodiments, but the protection scope of the present application is not limited thereby.

[0038] As shown in Figure 1 , 2 , the steps of the laser radar obstacle detection method of the present embodiment include:

[0039] S1. using a multi-line laser radar to scan the area in front of the carrier, and using a single-line laser radar to scan the blind area range of the multi-line laser radar;

[0040] S2. rasterize the point cloud data of the multi-line laser radar scanning to obtain a multi-line laser radar raster point cloud map, and obtain the scanning line dense distribution characteristics in a specified range near each grid;

[0041] S3. for the point cloud data of the single-line laser radar scanning, assist using the scanning line dense distribution characteristics of each grid obtained by the multi-line laser radar to obtain the obstacle detection result of the single-line laser radar;

[0042] S4. fuse the obstacle detection result of the multi-line laser radar and the obstacle detection result of the single-line laser radar to obtain the final detection result.

[0043] On the basis of using the single-line laser radar to complement the blind area of the multi-line laser radar, the embodiment simultaneously considers that the single-line radar will cause a large number of false alarm obstacles due to local aggregation of scanning lines in non-flat ground, uphill and downhill areas and the like, obtains the scanning line dense distribution characteristics near each grid of the grid point cloud map by scanning the grid point cloud map of the multi-line laser radar, and guides the scanning of the single-line laser radar by using the scanning line distribution characteristics obtained by the multi-line laser radar as prior knowledge, to assist the single-line laser radar in obstacle detection, which can effectively eliminate the single-line scanning false alarm caused by non-flat ground such as steep areas, uphill and downhill areas, and realize adaptive scene filtering of non-obstacle points in single-line scanning, especially can effectively solve the problem of false alarm of forward obstacles caused by aggregation of single-line in the near distance in the case of uphill and downhill, and finally realize comprehensive and accurate obstacle detection by fusing the multi-line radar and the single-line radar.

[0044] In the embodiment, the multi-line laser radar and the single-line laser radar are arranged on a carrier, the vertical height of the multi-line laser radar is higher than the vertical height of the single-line laser radar, that is, the multi-line laser radar is arranged at a higher position on the carrier, which can maximize the scanning range of the multi-line laser radar, and the single-line laser radar is arranged at a lower position on the carrier, which can complement the blind scanning in the blind area of the low position. The carrier can be a vehicle, a mobile detection device or the like.

[0045] In a specific application embodiment, step S1 specifically arranges a multi-line laser radar on top of the carrier robot as the main sensing device, and hangs a single-line laser radar on the lower front side of the carrier robot as an auxiliary sensing device, and sets the effective scanning range of the device single-line laser radar as the front blind area of the multi-line laser radar, such as configuring the detection range of the single-line laser radar as the interval starting from the front of the robot and ending at the first landing line of the multi-line laser radar, to cover the blind area of the multi-line laser radar, and sets the scanning mode of the single-line laser radar as horizontal ring scanning. The single-line laser radar obtains the close-range scanning information in front of the carrier robot, and completes the scanning blind area of the main radar (multi-line laser radar) in the front close range, and by configuring the detection range of the single-line laser radar according to the blind area range of the multi-line laser radar, the multi-line front can be realized. comprehensive completion of low obstacles.

[0046] In this embodiment, step S2 specifically includes:

[0047] S21. Rasterize the point cloud data scanned by the multi-line laser radar to obtain a multi-line laser radar grid point cloud map;

[0048] S22. Determine the continuity of the scanning lines between the grid where each point cloud in the multi-line laser radar grid point cloud map is located and the adjacent grid.

[0049] S23. Count the number of scanning lines determined to be continuous in each grid, and determine the scanning line dense distribution feature of each grid according to the number of scanning lines.

[0050] The above step S22 specifically includes: traversing each point cloud in the multi-line laser radar grid point cloud map, each time taking the current grid position as the reference position, searching for the adjacent grid of the current grid, when the adjacent grid has a scanning point, recording the scanning line index number, and the scanning line index number of the adjacent grid and the scanning line index number of the current grid are judged for continuity, to obtain all scanning lines determined to be continuous corresponding to the current grid.

[0051] In a specific application embodiment, when extracting the scanning line dense distribution feature, for each grid target in the scanning map obtained by scanning the multi-line laser radar, the grid point cloud in the neighborhood is extracted, and according to the adjacent position relationship of the point cloud in the grid map and the source of the scanning line, the current grid position is taken as the main grid, the adjacent grid of the grid where the point cloud is located is searched, when the adjacent grid has a scanning point, the scanning line index number is recorded, and the scanning line index number of the neighborhood and the scanning line index number of the main grid are judged for continuity. The number of scanning lines that meet the continuity with the index of the main grid the most is taken as the calculation result of the current grid, and is output as the scanning line dense distribution feature. After the above processing is performed on each grid, the scanning line dense distribution feature of the multi-line laser radar in the single-line detection range can be obtained.

[0052] It can be understood that other features that can represent the scanning line density distribution characteristics can also be used, such as using a quantized density degree value, calculating the density degree value according to the number of continuous scanning lines near the grid, and then based on the density degree value, the single-line laser radar can directly determine whether the scanned point cloud is an obstacle. Of course, according to actual needs, a plurality of features can be used to represent the scanning line density distribution to further improve the detection accuracy.

[0053] When the single-line laser radar scans uneven road surfaces such as bumps, uphill and downhill areas, local aggregation of scanning lines will occur, and the corresponding grid of the multi-line laser radar will have dense scanning lines when scanning. Through the above steps, the scanning line dense distribution characteristics of each grid in the multi-line laser radar scanning graph can be effectively obtained. Based on the scanning line dense distribution characteristics, the single-line laser radar can subsequently determine the local aggregation point cloud of scanning lines generated due to uneven road surfaces such as bumps, uphill and downhill areas, thereby eliminating false alarms generated by the single-line laser radar due to local aggregation of scanning lines on uneven road surfaces.

[0054] Step S3 in this embodiment includes:

[0055] S31. The point cloud data scanned by the single-line laser radar is rasterized to obtain single-line radar rasterized point cloud data.

[0056] S32. The single-line radar rasterized point cloud data is traversed, and the scanning line dense distribution characteristics of the corresponding position of each point cloud are used to assist in determining whether the point cloud is an obstacle, to obtain the obstacle detection result of the single-line laser radar.

[0057] In the above step S32, if the scanning line dense distribution characteristics meet the preset conditions, such as the number of continuous scanning lines being less than the preset number threshold, or the density degree value being greater than the preset degree threshold, it is determined that it is a non-flat ground non-obstacle. Otherwise, it is determined to be an obstacle. It can be understood that other judgment methods can also be used to assist in determining whether the point cloud is an obstacle, such as pre-acquiring a scanning line distribution feature template of a non-flat ground, and then matching the real-time acquired scanning line dense distribution characteristics with the scanning line distribution feature template of the non-flat ground. If the matching is successful, it is determined to be a non-flat ground. This judgment method can be used to determine according to the characteristics of the non-flat ground, which can further improve the detection accuracy and reduce the probability of misidentifying other obstacles as non-flat ground.

[0058] In a specific application embodiment, when traversing the rasterized point cloud of the single-line laser radar, if there is a point cloud in the rasterized point cloud map of the single-line laser, it is preliminarily judged that there is an obstacle, and then the dense distribution feature of the scanning line of the multi-line laser radar is used as prior knowledge to guide the obstacle completion of the scanning map of the single-line laser. Specifically, for the point cloud obtained by scanning the single-line laser radar, it is checked whether there is a dense distribution feature of the scanning line in the continuous map description of the corresponding position of the multi-line laser. If the dense distribution feature of the scanning line meets the preset requirement, it is determined that the obstacle point cloud of the grid is false alarm point cloud, and the obstacle is not output externally; otherwise, it is the positive obstacle of the single-line laser radar, and the obstacle information is normally published, which can exclude the false alarm of the single-line laser radar due to local aggregation of point clouds of the scanning line on the uneven road surface.

[0059] The step S4 of the embodiment includes:

[0060] S41. Obtain the obstacle detection result of the multi-line laser radar according to the point cloud data scanned by the multi-line laser radar;

[0061] S42. Fuse the obstacle detection result of the multi-line laser radar with the obstacle detection result of the single-line laser radar to obtain the final detection result.

[0062] The step S41 of the embodiment specifically includes: determining the ground points according to the point cloud data scanned by the multi-line laser radar and a pre-constructed ground model, and deleting the ground points to obtain the obstacle detection result of the multi-line laser radar. The ground model is a model established for the ground, and the specific model type can be selected according to actual requirements. Based on the ground model, the ground points of the multi-line laser radar can be extracted after scanning, and the non-ground points are obtained after deleting the ground points, that is, the candidate obstacle points, and the obstacle points of the multi-line laser can be obtained after filtering.

[0063] In the step S42 of the embodiment, when fusing, the obstacle result of the single-line laser radar is fused into the obstacle result of the multi-line laser radar. If the multi-line laser radar detects the existence of an obstacle within the range of the single-line laser radar, the obstacle detection result of the multi-line laser radar is adopted; if the single-line laser radar detects an obstacle but the multi-line laser radar does not detect an obstacle, the obstacle detection result detected by the single-line laser radar is added. Since the false alarm in the obstacle detection of the single-line radar has been filtered out by means of the dense feature of the scanning line of the multi-line laser radar in the obstacle completion in the step S3 of the single-line laser radar obstacle detection process, the obstacle detection results of the multi-line laser radar and the single-line laser radar can be fused to obtain comprehensive and accurate obstacle detection results. After fusion, a filtering step (such as a sliding window filtering) is further included to filter out isolated false alarm grids, and finally a complete obstacle result is obtained and output externally.

[0064] The obstacle detection of the multi-line laser radar and the obstacle detection of the single-line radar can be performed simultaneously or sequentially, and the specific execution order can be configured according to actual needs.

[0065] The application will be further described below by taking the method for realizing obstacle detection in the specific application embodiment as an example.

[0066] As shown in Figure 3 , the detailed steps of the laser radar obstacle detection in the embodiment are as follows:

[0067] Step 1, radar configuration: configure the multi-line laser radar and the single-line laser radar, wherein the multi-line laser radar is configured as the main scanning device, and the measurement blind area range is set; the detection range of the single-line laser radar is set as an interval starting from the front of the robot and ending at the first landing line of the multi-line laser radar to cover the blind area of the multi-line laser radar.

[0068] Step 2, data extraction: extract the point cloud data scanned by the multi-line laser radar and the single-line laser radar, and then convert the measurement values from the radar coordinate system to the coordinate system with the center of the vehicle body as the origin.

[0069] Step 3, multi-line laser radar obstacle detection: based on the point cloud data of the multi-line laser radar, the ground model-based method is used to extract the obstacle area, that is, a mathematical model of the ground is established, the point cloud data of the multi-line laser is used as input, and the ground point cloud and the non-ground point cloud are obtained after processing, and after deleting the ground point cloud, the non-ground point cloud is rasterized to obtain the obstacle detection result output of the multi-line laser.

[0070] Step 4, multi-line laser radar scanning line dense distribution feature extraction: based on the multi-line laser radar data, the raster statistical method is used to extract the scanning line dense distribution feature of each target grid, specifically taking the current position of the grid as the reference position, the current grid as the main grid, searching the near neighbor grid of the grid, recording the scanning line index number when the near neighbor grid has scanning points, and continuously determining the scanning line index number of the neighbor grid and the scanning line index number of the reference grid, and the number of scanning lines with the longest continuity of the index number of the main grid is taken as the scanning line dense distribution feature output. After performing the above processing on each grid, the continuity description of the multi-line laser radar in the single-line detection range can be obtained.

[0071] Step 5, single-line laser radar obstacle detection based on multi-line laser radar guidance: the multi-line laser radar scanning line dense distribution feature is used as the prior knowledge for excluding false alarm grids in single-line scanning, and the multi-line laser radar point cloud falling within the single-line detection range is completed for obstacle.

[0072] Specifically, the rasterized point cloud of the single-line laser is traversed. If there is point cloud in the rasterized point cloud map of the single-line laser, the point cloud scanned by the single-line laser radar is searched to determine whether there is a scanning line dense distribution feature in the corresponding position in the multi-line laser continuous map description. If the scanning line dense distribution feature meets the preset requirement, it is determined that the obstacle point cloud of the grid is a false alarm point cloud, and the obstacle is not output externally. Otherwise, it is a single-line laser radar obstacle, and the obstacle information is normally published externally.

[0073] Step 6, obstacle fusion: the obstacle results of the multi-line radar and the detection results of the single-line laser radar are fused. The fusion process is as follows: when there is an obstacle in the range of the single-line laser, the detection result of the multi-line radar is used; when there is a result of the single-line laser radar and no result of the multi-line laser, if the scanning line dense distribution feature of the target grid meets the preset condition and is determined to be a non-obstacle, the single-line radar obstacle detection has been filtered in advance, otherwise, the single-line radar detection result is added to the final fusion detection result. Finally, the combined obstacles are filtered by a sliding window to isolate false alarm grids, and the complete obstacle result is obtained and output externally.

[0074] As shown in Figure 4 , the laser radar obstacle detection device of the embodiment includes:

[0075] The multi-line laser radar is used to scan the region in front of the carrier.

[0076] The single-line laser radar is used to scan the blind area range of the multi-line laser radar.

[0077] The obstacle detection module is used to rasterize the point cloud data scanned by the multi-line laser radar to obtain a multi-line radar grid point cloud map, and obtain the scanning line dense distribution feature in a specified range near each grid. The point cloud data scanned by the single-line laser radar is used to obtain the obstacle detection result of the single-line laser radar by using the scanning line dense distribution feature of each grid obtained by the multi-line laser radar. The obstacle detection results of the multi-line laser radar and the single-line laser radar are fused to obtain the final detection result.

[0078] In the embodiment, the obstacle detection module specifically includes:

[0079] The point cloud extraction unit is used to extract the point cloud data scanned by the multi-line laser radar and the single-line laser radar.

[0080] The feature extraction unit is used to rasterize the point cloud data scanned by the multi-line laser radar to obtain a multi-line radar grid point cloud map, and obtain the scanning line dense distribution feature in a specified range near each grid.

[0081] The obstacle fusion unit is used for the point cloud data of the single-line laser radar scanning, and the scanning line dense distribution characteristics of each grid obtained by using the multi-line laser radar are assisted to obtain the obstacle detection result of the single-line laser radar.

[0082] In the embodiment, the multi-line laser radar and the single-line laser radar are arranged on the carrier, the vertical height of the multi-line laser radar is higher than that of the single-line laser radar, that is, the multi-line laser radar is arranged at a higher position on the carrier, so as to expand the scanning range of the multi-line laser radar as much as possible, and the single-line laser radar is arranged at a lower position on the carrier, so as to perform blind area scanning for the blind area range at a low position. The carrier can be a vehicle, a mobile detection device or the like. The multi-line laser radar can be arranged at the top region of the carrier, so as to make the scanning range as large as possible, and the single-line laser radar is arranged at the lower side in front of the carrier to scan the blind area range of the multi-line laser radar.

[0083] The obstacle detection module in the embodiment corresponds to steps S2-S4 in the laser radar obstacle detection method, and will not be described here.

[0084] In another embodiment, the obstacle detection module can further include a processor and a memory, the memory is used for storing a computer program, and the processor is used for executing the computer program to perform steps S2-S4.

[0085] The application can be applied to the detection of obstacles in front of a mobile robot, and solves the problem of forward obstacle avoidance of the robot. The application can not only realize comprehensive detection of obstacles and avoid missing detection of low obstacles, but also greatly reduce the false alarm rate. The application can also be applied to obstacle detection in other similar occasions.

[0086] The above is only a preferred embodiment of the application, and does not limit the application in any form. Although the application has been disclosed as above, it is not intended to limit the application. Therefore, any simple modification, equivalent change and modification of the above embodiment without departing from the technical solution of the application shall fall within the protection scope of the technical solution of the application.

Claims

1. A laser radar obstacle detection method characterized by the steps of The method comprises the following steps: S1. Scanning the area in front of the carrier using a multi-line laser radar, and scanning the blind area range of the multi-line laser radar using a single-line laser radar; S2. Rasterizing the point cloud data scanned by the multi-line laser radar to obtain a multi-line radar raster point cloud map, and obtaining the scanning line dense distribution characteristics in a specified range near each grid for representing the scanning line density distribution; S3. Obtaining the obstacle detection result of the single-line laser radar by assisting the scanning line dense distribution characteristics of each grid obtained by the multi-line laser radar on the point cloud data scanned by the single-line laser radar; S4. Fusing the obstacle detection results of the multi-line laser radar and the single-line laser radar to obtain the final detection result; The step S3 comprises: S31. Rasterizing the point cloud data scanned by the single-line laser radar to obtain single-line radar rasterized point cloud data; S32. Iterating the single-line radar rasterized point cloud data, and assisting the judgment of whether the point cloud is an obstacle by using the scanning line dense distribution characteristics of the corresponding position of each point cloud to obtain the obstacle detection result of the single-line laser radar, wherein if there is point cloud in the single-line laser rasterized point cloud map, it is preliminarily judged that there is an obstacle, the scanning line dense distribution characteristics of the multi-line laser radar are used as prior knowledge to guide the obstacle completion of the scanning map of the single-line laser, and it is judged whether the obstacle point cloud of the grid is a false alarm point cloud.

2. The laser radar obstacle detection method according to claim 1, characterized by, The step S2 comprises: S21. Rasterizing the point cloud data scanned by the multi-line laser radar to obtain a multi-line radar raster point cloud map; S22. Judging the continuity of the scanning lines between the grid where each point cloud in the multi-line radar raster point cloud map is located and the neighboring grids; S23. Counting the number of scanning lines judged to be continuous in each grid, and determining the scanning line dense distribution characteristics of each grid according to the number of scanning lines.

3. The laser radar obstacle detection method according to claim 2, characterized by, The step S22 comprises: iterating each point cloud in the multi-line radar raster point cloud map, searching for the neighboring grids of the current grid each time by taking the current grid position as the reference position, recording the scanning line index number when there is a scanning point in the neighboring grids, and judging the continuity of the scanning line index number of the neighboring grids and the scanning line index number of the current grid to obtain all the scanning lines judged to be continuous corresponding to the current grid.

4. The laser radar obstacle detection method according to any one of claims 1 to 3, characterized by, The step S4 comprises: S41. Obtaining the obstacle detection result of the multi-line laser radar according to the point cloud data scanned by the multi-line laser radar; S42. Fusing the obstacle detection result of the multi-line laser radar and the obstacle detection result of the single-line laser radar to obtain the final detection result.

5. The laser radar obstacle detection method according to claim 4, characterized by, The step S41 comprises: determining the ground points according to the point cloud data scanned by the multi-line laser radar and a pre-constructed ground model, and deleting the ground points to obtain the obstacle detection result of the multi-line laser radar.

6. The laser radar obstacle detection method according to claim 4, characterized by, In the step S42, when the multi-line laser radar detects an obstacle in the range of the single-line laser radar, the obstacle detection result of the multi-line laser radar is adopted; when the single-line laser radar detects an obstacle but the multi-line laser radar does not detect an obstacle, the obstacle detection result detected by the single-line laser radar is added.

7. A ladar obstacle detection apparatus, characterized by, The method comprises the following steps: The multi-line laser radar is used for scanning a region in front of a carrier. The single-line laser radar is used for scanning a blind area of the multi-line laser radar. The obstacle detection module is configured to rasterize point cloud data scanned by the multi-line laser radar to obtain a multi-line laser radar grid point cloud map, and to obtain a scanning line dense distribution feature within a specified range near each grid; and to rasterize point cloud data scanned by the single-line laser radar, to obtain an obstacle detection result of the single-line laser radar by using the scanning line dense distribution feature of each grid obtained by the multi-line laser radar; and to fuse the obstacle detection results of the multi-line laser radar and the single-line laser radar to obtain a final detection result. The obstacle detection module comprises: A first unit is configured to rasterize point cloud data scanned by the single-line laser radar to obtain single-line laser radar rasterized point cloud data. A second unit is configured to traverse the single-line laser radar rasterized point cloud data, to use the scanning line dense distribution feature of a corresponding position of each point cloud to assist in judging whether the point cloud is an obstacle, and to obtain an obstacle detection result of the single-line laser radar; when the single-line laser radar rasterized point cloud is traversed, if there is point cloud in the single-line laser radar rasterized point cloud map, it is preliminarily judged that there is an obstacle, and then the scanning line dense distribution feature of the multi-line laser radar is used as prior knowledge to guide the obstacle completion of the scanning map of the single-line laser radar, and to determine whether the obstacle point cloud of the grid is false alarm point cloud.

8. The ladar obstacle detection apparatus of claim 7, wherein, The multi-line laser radar and the single-line laser radar are separately arranged on the carrier, and a vertical height of the multi-line laser radar is higher than a vertical height of the single-line laser radar.

9. The ladar obstacle detection apparatus of claim 7, wherein, The obstacle detection module comprises a processor and a memory, the memory is configured to store a computer program, and the processor is configured to execute the computer program to perform steps S2-S4 in the method of any one of claims 1-6.

Images