Full-scene target sensing method based on radar point cloud data processing

Abstract

The full-scene target sensing method based on radar point cloud data processing comprises the following steps: extracting point cloud data from radar echoes, determining detection result data of a power line based on a point cloud reconstruction method, storing the detection result data in a first storage area, taking the detection result data as current data to form first fusion data, and sending the first fusion data to a second storage area; the first fusion data is stored in the fourth storage area; forming second fusion data, and covering or replacing the first fusion data with the second fusion data; and respectively updating data stored in a second storage area and a third storage area by using the second fusion data, and sending the second fusion data to an avionics display system of an aerial carrier in a screening mode of setting a confidence threshold, thereby providing powerful support for improving the low-altitude anti-collision safety perception capability.

Description

Technical Field

[0001] The present invention belongs to the technical field of aircraft collision avoidance methods, and in particular relates to a full-scene target perception method based on radar point cloud data processing. Background Art

[0002] Helicopters / UAVs need to perform missions at low altitudes. Existing anti-collision radar systems can only detect and display typical isolated obstacles, high-voltage lines, and terrain. Their disadvantages are:

[0003] 1) Failure to effectively utilize historical data, and only local storage of historical detection results to meet the needs of data fusion in a short period of time, and data from one flight cannot be integrated with data from another flight;

[0004] 2) There are potential safety hazards. For example, when a target enters a blind spot (due to a short distance or a turn), no new detection data can be generated. The target will be deleted after a period of time, which will cause the previously detected target to be unable to be effectively displayed.

[0005] In summary, the existing detection method leads to unstable high-voltage line detection, reduces pilot reaction time, endangers flight safety, and has low performance. Summary of the invention

[0006] In view of this, the full-scene target perception method based on radar point cloud data processing of the present invention solves the technical problem that the current method causes low radar performance.

[0007] A full-scene target perception method based on radar point cloud data processing is applicable to detection by an airborne radar, wherein the radar is a millimeter-wave radar and includes a processor, wherein the processor includes a first storage area, a second storage area, a third storage area, and a fourth storage area, and the method includes:

[0008] S101: extracting point cloud data from radar echoes, and determining detection result data of power lines based on a point cloud reconstruction method, and storing the data in the first storage area, with the detection result data serving as current data;

[0009] S102: The second storage area and the third storage area are initially cleared, and the data stored in the second storage area and the third storage area can be matched and fused in real time to form first fused data, and the first fused data is stored in the fourth storage area;

[0010] S103: The detection result data is fused with the first fused data to form second fused data, and the second fused data covers or replaces the first fused data; the second fused data is used to update the data stored in the second storage area and the third storage area respectively, and the second fused data is sent to the avionics display system of the carrier aircraft by setting a confidence threshold.

[0011] Beneficial Effects

[0012] In view of the problems of existing anti-collision safety perception methods in terms of insufficient scalability of perceived targets, insufficient information utilization, and poor environmental adaptability, the present invention proposes a full-scene target perception method based on radar point cloud data processing. This method utilizes all effective echo information of the radar to form dense point cloud data, and suppresses interference such as noise and side lobes through point cloud filtering processing. It realizes the perception of various types of targets through point cloud segmentation, clustering and other methods, and realizes full-scene mapping through point cloud matching, fusion and reconstruction. This method can greatly improve the comprehensiveness and accuracy of safety perception, and provide technical support for the anti-collision safety perception of drones and helicopters. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0014] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0015] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0016] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0017] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0018] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure, and only show the components related to the present disclosure rather than being drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout type may also be more complicated.

[0019] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, one skilled in the art will appreciate that aspects may be practiced without these specific details.

[0020] Traditional power line detection and identification based on millimeter-wave radar point cloud reconstruction cannot meet the safety perception needs in environments such as densely populated urban buildings, complex bridge structures, and complex terrain in mountainous areas. In order to adapt to the full-scene use needs of helicopters / UAVs, the radar system needs to have full-scene target safety perception capabilities. Vector maps are a very common map format in electronic maps. Road networks, rivers, and lakes are often represented as vector maps. They store the location and size information of the target, which is displayed in a certain form when drawing. It can be infinitely enlarged without blurring. The current anti-collision perception radar extracts typical terrain, high-voltage lines, chimneys and other target features based on echoes to achieve detection functions. The main problems are:

[0021] 1) Poor scalability. The current method needs to extract features from echoes based on target features. Since ground targets are complex and diverse with different characteristics, it is difficult to generalize the method of one type of feature to another type of target and realize the perception of targets in all scenes;

[0022] 2) Insufficient information utilization. Current methods such as CFAR or zero-crossing detection only utilize minimal information in the echo. For ground echoes, which have a large number of useful echoes, it is a better method to treat them as dense point clouds;

[0023] 3) Weak environmental adaptability. The current method is prone to missed detection when the target echo is weak and false alarm when the noise sidelobe is strong, and has the problem of insufficient environmental adaptability.

[0024] like Figure 1 The full-scene target perception method based on radar point cloud data processing is suitable for detection by airborne radar, wherein the radar is a millimeter-wave radar and includes a processor, wherein the processor includes a first storage area ( Figure 1 The current map stored in the second storage area ( Figure 1 The local map stored in the third storage area ( Figure 1 A global map stored in (a) and a fourth storage area, wherein the method comprises:

[0025] S101: extracting point cloud data from radar echoes, and determining detection result data of power lines based on a point cloud reconstruction method, and storing the data in a first storage area, wherein the detection result data is used as current data, the power lines, for example, high voltage lines;

[0026] S102: The second storage area and the third storage area are initially cleared, and the data stored in the second storage area and the third storage area can be matched and fused in real time to form first fused data (e.g., multiple detection results of the same high-voltage line are fused as historical data) and the first fused data is stored in the fourth storage area;

[0027] S103: The detection result data is fused with the first fused data to form second fused data, and the second fused data covers or replaces the first fused data; the data stored in the second storage area and the third storage area are respectively updated with the second fused data, and the second fused data is sent to the avionics display system of the carrier aircraft by setting a "confidence threshold" screening method.

[0028] Point cloud data is obtained from radar echoes, and filtering, ground segmentation and object segmentation are performed to identify high-voltage tower targets in the scene. Heuristic search is used to effectively identify power lines. For example, a large amount of dense point cloud data is obtained from radar echoes, and the perception of targets in the entire scene is achieved through point cloud denoising, segmentation, clustering and reconstruction.

[0029] Preferably, point cloud data is extracted from radar echoes, including: presetting a power threshold value, extracting echo point information above the preset power threshold value from each frame of radar echo, and recording the latitude, longitude, altitude and heading information of the carrier in each frame, echo point information, longitude, longitude, altitude and heading information to form point cloud data.

[0030] As a specific implementation method provided in this case, the method based on point cloud reconstruction in S101 determines the detection result data of the power line, including:

[0031] S301: Preprocessing of point cloud data: motion compensation and denoising of point cloud data according to the position of the aircraft in each frame; for example, outlier removal method is used to remove noise points and false alarm points.

[0032] S302: The pre-processed point cloud data is subjected to ground segmentation processing, including constructing a voxel map, designing the voxel map according to the azimuth resolution and distance resolution of the radar, and determining the position of the voxel map where the point cloud data is located, and storing it accordingly. For example, the voxel map includes a plurality of first grids, and the point cloud data is stored in the corresponding first grids.

[0033] S303: plane fitting, dividing the voxel map into multiple small areas, and using the RANSAC algorithm in the small areas to estimate the plane model and the ground features of the plane model, the ground features including the platform normal verticality, average height deviation and flatness;

[0034] S304: Ground likelihood estimation: if each small area meets the preset feature threshold of the ground feature, it is used as ground point cloud data, and non-ground point cloud data in the point cloud data is filtered out according to the ground point cloud data, for example, a building or a high-voltage tower is identified;

[0035] S305: Segmentation of non-ground point cloud data, determining the position of the non-ground point cloud data in the second grid in the voxel map (for example, selecting according to a mapping relationship), and storing the non-ground point cloud data in the corresponding second grid;

[0036] Any second grid is used as the initial seed voxel, and the initial seed voxel is used as the initial supervoxel. The supervoxel is grown by the neighbor search method. For example, if there is a voxel in the preset neighbor range, the voxel is added to the supervoxel in a clustering manner, and the neighbor search is continued until it is impossible to continue growing according to the current search rule, and the next supervoxel is grown until there is no unsearched voxel in the space. Preferably, the supervoxels with adjacent positions and high approximation are connected and grown together;

[0037] S306: Power line detection and identification based on heuristic search, including:

[0038] Get the geometry of high voltage towers and power lines;

[0039] The supervoxels that meet the geometric characteristics of high-voltage towers are selected as input for high-voltage line identification;

[0040] According to the geometric distribution characteristics of power lines, high-voltage towers are connected as power lines, and the confidence of power lines is calculated. At the same time, the reference point information of the current scanning line carrier is obtained; the power lines and their confidence and reference point information are used as the detection result data.

[0041] As a specific implementation method provided in this case, S102 forms first fusion data, including (for example, multiple detection results of the same high-voltage line are fused as historical data);

[0042] The data stored in the second storage area (local map) and the third storage area (global map) are converted to the same three-dimensional coordinates, for example, the coordinates are transformed. And retrieve the corresponding reference point information Ref l ;

[0043] Get all high voltage line data from the third storage area (global map) And the high voltage line data The absolute coordinates are transformed to the reference point Ref corresponding to the data in the second storage area. l That is, the high-voltage lines in the local map and the global map are placed under the same reference point;

[0044] The data stored in the second storage area and the third storage area can be matched and integrated in real time, including:

[0045] Power line matching, high voltage line data {L l 1.L l 2, ..., L l n And high voltage line data {L g 1,L g 2,…,L g n} traverses the process, for example, (pairing two by two), if the average distance between the two lines is less than the set average distance And the angle between the two lines is less than the preset angle θ<6°, then the two lines are matched successfully;

[0046] After the matching is successful, the high-voltage line fusion is performed. If the distance between the two power lines is less than the preset distance threshold, for example, the preset distance threshold is 50m, the coordinates of the tower points of the power lines are weighted averaged according to the confidence level to obtain the coordinates of the new tower points after fusion, thereby obtaining the new high-voltage line after fusion, which is expressed as:

[0047]

[0048] Where (x′, y′) is the coordinate of the tower point after fusion, (x1, y1) and (x2, y2) are the coordinates of the tower point before fusion, and c1 and c2 are the confidences of the two lines before fusion.

[0049] As a specific implementation method provided in this case, the detection result data in S103 is fused with the first fusion data, including:

[0050] The data stored in the first storage area and the fourth storage area are converted to the same three-dimensional coordinates, and the method is the same as the fusion of the data stored in the first storage area and the fourth storage area;

[0051] Power line matching, the high voltage line data {L l 1.L l 2, ..., L l n} and the high voltage line data {L g 1,L g 2,…,L g n} Perform traversal processing. If the average distance between the two lines is less than the set average distance If the angle between the two lines is less than the preset angle θ<6°, the two lines are matched successfully;

[0052] After the matching is successful, the high-voltage line is fused. If the distance between the two power lines is less than the preset distance threshold, the coordinates of the tower points of the power lines are weighted averaged according to the confidence level to obtain the coordinates of the new tower points after fusion, thereby obtaining the new fused high-voltage line.

[0053] As a specific implementation method provided in this case, in S103, the second fusion data is used to update the data stored in the second storage area and the third storage area respectively, and the second fusion data is sent to the avionics display system of the carrier aircraft by setting a confidence threshold, including:

[0054] The high-voltage line of the confidence level of the data in the second fused data c≥15 updates the data in the third storage area; the high-voltage line of the confidence level of the data in the second fused data 3.5≤c<15 updates the data in the second storage area;

[0055] The power line is transmitted to the avionics display system. If the confidence level of the data in the second fusion data is c>6.5, it is sent; otherwise, it is not sent. It should be pointed out that the "scan line" high-voltage line detection detects high-voltage towers and power lines from the radar echo of the current antenna scanning line according to the geometric and electromagnetic characteristics of high-voltage towers and power lines, and identifies them as high-voltage lines {L1, L2, ..., L n}, and calculate their confidence {c1,c2,…,c n}, at the same time, obtain the reference point information Ref of the current scanning line carrier, including latitude and longitude, altitude and heading information.

[0056] In general, the common anti-collision safety perception methods first analyze the target features based on the echo, and then extract the features for target detection and identification. These methods mostly consider the more ideal terrain and obstacle features. Due to the large variety of ground targets and complex features, this will lead to the problem that the unconsidered targets are difficult to detect, and cannot meet the use requirements of all dangerous obstacle detection and perception in low-altitude environments. Compared with traditional methods, the advantages of this invention are:

[0057] First, by processing with the point cloud method, a large amount of useful information can be obtained from the radar echo, and the utilization rate of information is much higher than that of the traditional method; secondly, through point cloud filtering processing, the surrounding space information is fully utilized, and interference such as false alarms and side lobes can be easily removed; thirdly, through the method of the present invention, ground segmentation and non-ground target segmentation can be achieved respectively, and target detection and reconstruction in the whole scene can be achieved, which greatly improves the ability of collision avoidance safety perception. The method of the present invention can be applied to radar signal processing and data processing, providing strong support for improving the low-altitude collision avoidance safety perception ability.

[0058] In the above method, the acquisition of point cloud data in S101 includes:

[0059] S801: Set an 8dB signal-to-noise ratio threshold, extract echo point information from each frame of radar echo, and record the latitude, longitude, altitude and heading information of each frame of the carrier aircraft to form point cloud data containing target information and carrier aircraft information;

[0060] S802: point cloud preprocessing, including, for the acquired point cloud data, using the first frame of the aircraft position as a reference point, and compensating for the point cloud data movement according to the position change of each other frame of the aircraft;

[0061] The outlier removal method is used to remove outliers with a distance of more than 100m;

[0062] S803: Point cloud ground segmentation, including:

[0063] Construct a voxel map, where the grid map resolution is designed to be 1.0° in azimuth and 12m in distance according to the azimuth resolution and distance resolution of the radar, and the point cloud data is quantified and filled into the grid map;

[0064] Plane fitting, where the voxel map is divided into regions of 5° and 60m, and the plane model is estimated using the RANSAC algorithm within the region;

[0065] Ground likelihood estimation, where for each area, ground features are calculated based on the estimated plane model, including platform normal verticality, average height deviation and flatness, preset parameter values, including verticality threshold of 75°, average height deviation threshold of 10m and flatness threshold of 3m, and the areas that meet the preset parameter values ​​are recorded;

[0066] Ground height calculation, where, for an area that meets the preset parameter value, the average of the calculated ground points is selected as the height of the area.

[0067] S804: Non-ground segmentation, including:

[0068] Construct a voxel map, wherein, according to the azimuth resolution and distance resolution of the radar, a grid map resolution of 1.0° in azimuth and 12m in distance is designed, the grid position of the point cloud data is calculated, and it is stored in the corresponding grid map;

[0069] Gridding to select seed voxels, where the gridding processing resolution (seed spacing) is set to 5 times the grid division rate, the voxel closest to the center in the grid is selected as the initial seed voxel, and the isolated seed points in the voxel space are deleted;

[0070] Voxel growth, where the voxel is grown from the seed voxel through the 26-neighbor search method. If there is a voxel in the adjacent range, the voxel is grown into the current supervoxel, and the adjacent search is continued until the current search rule cannot be used to grow further, and the next supervoxel is grown until there are no unsearched voxels in the space;

[0071] Supervoxel growth, in which adjacent, highly similar supervoxels are connected and grown together;

[0072] S805: power line detection and identification based on heuristic search, wherein supervoxels that meet the geometric features of the high-voltage tower are screened out as input for power line identification;

[0073] According to the geometric distribution characteristics of power lines, high-voltage towers are connected as power lines, their confidence is calculated, and the reference point information of the current scanning line carrier is obtained to form point cloud data.

[0074] The above is only a specific implementation of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present disclosure should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A full-scene target perception method based on radar point cloud data processing, suitable for detection by airborne radar, the radar is a millimeter wave radar and includes a processor, characterized in that: The processing machine includes a first storage area, a second storage area, a third storage area and a fourth storage area, and the method includes: S101: extracting point cloud data from radar echoes, and determining detection result data of power lines based on a point cloud reconstruction method, and storing the data in the first storage area, with the detection result data serving as current data; S102: The second storage area and the third storage area are initially cleared, and the data stored in the second storage area and the third storage area can be matched and fused in real time to form first fused data, and the first fused data is stored in the fourth storage area; S103: The detection result data is fused with the first fused data to form second fused data, and the second fused data covers or replaces the first fused data; the second fused data is used to update the data stored in the second storage area and the third storage area respectively, and the second fused data is sent to the avionics display system of the carrier aircraft by setting a confidence threshold.

2. The power line detection and identification method according to claim 1, characterized in that: Extract point cloud data from radar echoes, including: A power threshold value is preset, echo point information higher than the preset power threshold value is extracted from each frame echo of the radar, and the latitude, longitude, altitude and heading information, echo point information, latitude, longitude, altitude and heading information of each frame of the carrier are recorded to form point cloud data.

3. The power line detection and identification method according to claim 2, characterized in that: The detection result data of power lines is determined based on the point cloud reconstruction method, including: S301: preprocessing the point cloud data, performing motion compensation on the point cloud data according to the position of the aircraft in each frame, and performing denoising processing; S302: performing ground segmentation processing on the pre-processed point cloud data, including constructing a voxel map, designing the voxel map according to the azimuth resolution and distance resolution of the radar, determining the position of the point cloud data in the voxel map, and storing it accordingly; S303: plane fitting, dividing the voxel map into multiple small areas, and using the RANSAC algorithm in the small areas to estimate the plane model and the ground features of the plane model, wherein the ground features include the platform normal verticality, average height deviation and flatness; S304: Ground likelihood estimation: if each small area meets a preset feature threshold of ground features, the small area is used as ground point cloud data, and non-ground point cloud data in the point cloud data is filtered out according to the ground point cloud data; S305: Segmentation of non-ground point cloud data, determining the position of the non-ground point cloud data in the second grid in the voxel map, and storing the non-ground point cloud data in the corresponding second grid; Any one of the second grids is used as an initial seed voxel, the initial seed voxel is used as an initial super voxel, and the super voxel is grown by an adjacent search method; S306: Power line detection and identification based on heuristic search, including: Get the geometry of high voltage towers and power lines; Screening out supervoxels that meet the geometric features of the high-voltage tower as input for high-voltage line identification; According to the geometric distribution characteristics of the power lines, high-voltage towers are connected as power lines, and the confidence of the power lines is calculated. At the same time, the reference point information of the current scanning line carrier is obtained; the power lines and their confidence and reference point information are used as the detection result data.

4. The power line detection and identification method according to claim 4, characterized in that: In S305, the supervoxels that are adjacent in position and highly similar are connected and grown together.

5. The power line detection and identification method according to claim 3, characterized in that: S102 forms first fusion data, including: The data stored in the second storage area and the third storage area are converted to the same three-dimensional coordinates, and all high-voltage line data in the second storage area are obtained. And retrieve the corresponding reference point information Ref l ; Take out all high voltage line data from the third storage area {L g 1,L g 2,…,L g n }, and the high voltage line data {L g 1,L g 2,…,L g n The absolute coordinates are transformed to the reference point Ref corresponding to the data in the second storage area. l down and at the same reference point; The data stored in the second storage area and the third storage area can be matched and integrated in real time. include, Power line matching, high voltage line data {L l 1,L l 2,…,L l n } and high voltage line data {L g 1,L g 2,…,L g n } Perform traversal processing. If the average distance between the two lines is less than the set average distance And the angle between the two lines is less than the preset angle θ<6°, then the two lines are matched successfully; After the matching is successful, the high-voltage line fusion is performed. If the distance between the two power lines is less than the preset distance threshold, the coordinates of the tower points of the power lines are weighted averaged according to the confidence level to obtain the coordinates of the new tower points after fusion, thereby obtaining the new high-voltage line after fusion, which is expressed as: Where (x′, y′) is the coordinate of the tower point after fusion, (x1, y1) and (x2, y2) are the coordinates of the tower point before fusion, and c1 and c2 are the confidences of the two lines before fusion.

6. The power line detection and identification method according to claim 5, characterized in that: The detection result data in S103 is fused with the first fused data, including: The data stored in the first storage area and the fourth storage area are converted into the same three-dimensional coordinates; Power line matching, the high voltage line data {L l 1.L l 2, ..., L l n } and the high voltage line data {L g 1.L g 2, ..., L g n } Perform traversal processing. If the average distance between the two lines is less than the set average distance If the angle between the two lines is less than the preset angle θ<6°, the two lines are matched successfully; After the matching is successful, the high-voltage line is fused. If the distance between the two power lines is less than the preset distance threshold, the coordinates of the tower points of the power lines are weighted averaged according to the confidence level to obtain the coordinates of the new tower points after fusion, thereby obtaining the new fused high-voltage line.

7. The power line detection and identification method according to claim 6, characterized in that: In S103, the second fused data is used to update the data stored in the second storage area and the third storage area respectively, and the second fused data is sent to the avionics display system of the carrier aircraft by setting a confidence threshold, including: The high-voltage line of the confidence level c of the data in the second fused data is ≥ 15, and the data in the third storage area is updated; The high-voltage line of the confidence level of the data in the second fused data is 3.5≤c<15, and the data in the second storage area is updated; The power line is transmitted to the avionics display system. If the confidence level c of the data in the second fusion data is greater than 6.5, it is sent; otherwise, it is not sent.

8. The power line detection and identification method according to claim 5, characterized in that: The acquisition of point cloud data in S101 includes: S801: Set an 8dB signal-to-noise ratio threshold, extract echo point information from each frame of radar echo, and record the latitude, longitude, altitude and heading information of each frame of the carrier aircraft to form point cloud data containing target information and carrier aircraft information; S802: point cloud preprocessing, including, for the acquired point cloud data, using the first frame of the aircraft position as a reference point, and compensating for the point cloud data movement according to the position change of each other frame of the aircraft; The outlier removal method is used to remove outliers with a distance of more than 100m; S803: Point cloud ground segmentation, including: Construct a voxel map, where the grid map resolution is designed to be 1.0° in azimuth and 12m in distance according to the azimuth resolution and distance resolution of the radar, and the point cloud data is quantified and filled into the grid map; Plane fitting, where the voxel map is divided into regions of 5° and 60m, and the plane model is estimated using the RANSAC algorithm within the region; Ground likelihood estimation, where for each area, ground features are calculated based on the estimated plane model, including platform normal verticality, average height deviation and flatness, preset parameter values, including verticality threshold of 75°, average height deviation threshold of 10m and flatness threshold of 3m, and the areas that meet the preset parameter values ​​are recorded; Ground height calculation, where for an area that meets the preset parameter value, the average of the calculated ground points is selected as the height of the area; S804: Non-ground segmentation, including: Construct a voxel map, wherein, according to the azimuth resolution and distance resolution of the radar, a grid map resolution of 1.0° in azimuth and 12m in distance is designed, the grid position of the point cloud data is calculated, and it is stored in the corresponding grid map; Gridding to select seed voxels, where the gridding resolution is set to 5 times the grid division rate, the voxel closest to the center in the grid is selected as the initial seed voxel, and the isolated seed points in the voxel space are deleted; Voxel growth, where the voxel is grown from the seed voxel through the neighbor search method. If there is a voxel in the neighbor range, the voxel is grown into the current supervoxel, and the neighbor search is continued until the current search rule cannot be used to grow further, and the next supervoxel is grown until there are no unsearched voxels in the space; Supervoxel growth, in which adjacent supervoxels with high similarity are connected and grown together; S805: Power line detection and identification based on heuristic search, wherein supervoxels that meet the geometric features of the high-voltage tower are screened out as input for power line identification; According to the geometric distribution characteristics of power lines, high-voltage towers are connected as power lines, their confidence is calculated, and the reference point information of the current scanning line carrier is obtained to form the final point cloud data.

Images