Scanning laser radar point cloud data filtering method for spatial non-cooperative target three-dimensional reconstruction
By filtering and preprocessing the scanning data of the scanning lidar and establishing a spatiotemporal continuous point cloud filtering threshold, the problem of low point cloud resolution in the existing technology is solved, and the accuracy and quality of three-dimensional reconstruction of spatial non-cooperative targets are improved.
Patent Information
- Application Number
- CN202510616793.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-14
- Publication Date
- 2025-09-30
AI Technical Summary
Existing technologies make it difficult to improve the accuracy and quality of three-dimensional reconstruction of non-cooperative targets in space without changing the existing lidar hardware structure, especially because the point cloud resolution is low and the texture detail features are not detailed enough, which affects the feature extraction and reconstruction effects.
By filtering and preprocessing the scanning data of the scanning lidar, a point cloud filtering threshold based on spatiotemporal continuity is established, including distance and angle thresholds, to filter out erroneous point cloud data and improve the quality of point cloud data.
Without changing the hardware structure of the lidar, the quality of point cloud data and the three-dimensional reconstruction accuracy and quality of spatial non-cooperative targets are significantly improved, and the target outline is clarified.
Smart Images
Figure CN120726261A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a scanning laser radar point cloud data filtering method for three-dimensional reconstruction of spatial non-cooperative targets. The method is suitable for improving the quality of point cloud data by preprocessing the point cloud data without changing the hardware structure of existing laser radar products, thereby improving the accuracy and quality of three-dimensional reconstruction of spatial non-cooperative targets. Background Art
[0002] The rendezvous measurement sensor is an essential device for realizing the docking combination of spacecraft. It provides the spacecraft with navigation information such as the relative position and attitude of the docking target. On the basis of the original optical camera, the introduction of the laser measurement system can make the measurement system of the measurement sensor richer and the measurement range longer. It not only has the position and attitude measurement function of cooperative / non-cooperative targets, but also can realize the three-dimensional reconstruction of non-cooperative targets in space.
[0003] 3D reconstruction of non-cooperative targets in space is a challenging task. Compared to camera imaging systems, laser point clouds have lower resolution and a rougher depiction of target texture details. This significantly hinders feature extraction, matching, and reconstruction of non-cooperative targets. For scanning LiDAR, improving the 3D reconstruction of non-cooperative targets depends on a number of factors, including the number of point clouds collected, the laser's measurement accuracy, and the laser's angular measurement accuracy. Improving the accuracy and quality of 3D reconstruction of non-cooperative targets in space without changing the existing LiDAR hardware structure is a pressing issue. Summary of the Invention
[0004] The technical problem solved by the present invention is: to overcome the shortcomings of the existing technology and provide a scanning laser radar point cloud data filtering method for three-dimensional reconstruction of spatial non-cooperative targets. While keeping the state of the existing laser product unchanged, the scanning data of the scanning laser radar is filtered and pre-processed to improve the quality of the point cloud data collected by the laser radar, thereby improving the accuracy and quality of the three-dimensional reconstruction of spatial non-cooperative targets.
[0005] The technical solution of the present invention is:
[0006] A scanning lidar point cloud data filtering method for 3D reconstruction of non-cooperative targets in space, comprising the following steps:
[0007] The first step is to obtain the ranging accuracy σ of the scanning laser radar d and angle measurement accuracy σ x , σ y ;
[0008] The second step is to clarify the scanning lidar's on-orbit workflow, data acquisition parameters, and mission status.
[0009] The third step is to build a 1:1 model of the non-cooperative space target in the ground laboratory, scan the non-cooperative space target, and obtain a multi-frame, multi-angle scanning point cloud dataset P of the non-cooperative space target.
[0010] The fourth step is to extract the distance vector of the point cloud data from the data set P obtained in the third step and the x-direction angle vector y-direction angle vector
[0011] The fifth step is to analyze the configuration characteristics of the spatial non-cooperative target, evaluate the surface characteristics of the spatial non-cooperative target, and establish a target point cloud filtering threshold based on spatiotemporal continuity;
[0012] The sixth step is to filter the point cloud data extracted in the fourth step according to the established target point cloud filtering distance threshold τ1, point cloud filtering x-direction angle threshold τ2, and point cloud filtering y-direction angle threshold τ3 to form a filtered point cloud dataset.
[0013] Preferably, in the first step, the ranging accuracy and angle measurement accuracy of the scanning laser radar are obtained through the technical manual of the laser radar, or are measured through ground tests.
[0014] Preferably, in the fifth step, the method for determining the target point cloud filtering threshold τ1 is:
[0015] Based on the angular velocity ω of the laser radar scan, the data acquisition and processing frequency f of the laser radar, the number of laser radar single-frame data acquisition points n, and the working distance h between the laser radar and the target, the following formula is used to determine the maximum scanning distance s of two lasers on the target surface in adjacent time:
[0016] Taking the maximum scanning distance s of two lasers on the target surface in adjacent time as the scale, the distance between the two points with the largest surface drop within the circular area with a diameter of s on any surface of the spatial non-cooperative target is determined as the target point cloud filtering distance threshold τ1.
[0017] Preferably, according to the x-direction angular velocity ω scanned by the laser radar x and the y-direction angular velocity ω y , use the following formula to determine the target point cloud filtering thresholds τ2 and τ3;
[0018]
[0019] Preferably, the data set before point cloud filtering in the sixth step is The dataset after point cloud filtering is
[0020] in, Respectively represent the measured distance, x-direction measured angle, y-direction measured angle after filtering of a frame of scan data, Δd, Δθ x , Δθ y Respectively represent the measurement distance difference of adjacent scan data, the measurement angle difference in the x direction, and the measurement angle difference in the y direction, Represents the filtered point cloud data.
[0021] A scanning laser radar point cloud data filtering system for three-dimensional reconstruction of spatial non-cooperative targets, comprising a scanning laser radar task determination module, a scanning point cloud data set acquisition module, a point cloud data vector extraction module, and a point cloud data filtering module;
[0022] Scanning LiDAR mission determination module: used to clarify the scanning LiDAR on-orbit workflow, data acquisition parameters, and mission status;
[0023] Scanning point cloud dataset acquisition module: This module builds a 1:1 model of a non-cooperative space target in a ground laboratory, scans the non-cooperative space target, and obtains a multi-frame, multi-angle scanning point cloud dataset of the non-cooperative space target.
[0024] Point cloud data vector extraction module: extracts the distance vector, x-direction angle vector, and y-direction angle vector of the point cloud data from the point cloud data set acquired by the scanning point cloud data set acquisition module;
[0025] Point cloud data filtering module: establishes a target point cloud filtering threshold based on spatiotemporal continuity, filters the point cloud data extracted by the point cloud data vector extraction module, and forms a filtered point cloud data set
[0026] Preferably, the target point cloud filtering threshold based on spatiotemporal continuity includes a point cloud filtering distance threshold τ1, a point cloud filtering x-direction angle threshold τ2, and a point cloud filtering y-direction angle threshold τ3.
[0027] Preferably, the method for determining the target point cloud filtering threshold τ1 is:
[0028] Based on the angular velocity ω of the laser radar scan, the data acquisition and processing frequency f of the laser radar, the number of laser radar single-frame data acquisition points n, and the working distance h between the laser radar and the target, the following formula is used to determine the maximum scanning distance s of two lasers on the target surface in adjacent time:
[0029] Taking the maximum scanning distance s of two lasers on the target surface in adjacent time as the scale, the distance between the two points with the largest surface drop within the circular area with a diameter of s on any surface of the non-cooperative target in space is determined as the target point cloud filtering distance threshold τ1
[0030] Preferably, according to the x-direction angular velocity ω scanned by the laser radar x and the y-direction angular velocity ω y , use the following formula to determine the target point cloud filtering thresholds τ2 and τ3;
[0031]
[0032] Preferably, the data set before point cloud filtering is The data set after filtering by the point cloud data filtering module is
[0033] in, Respectively represent the measured distance, x-direction measured angle, y-direction measured angle after filtering of a frame of scan data, Δd, Δθ x , Δθ y Respectively represent the measurement distance difference of adjacent scan data, the measurement angle difference in the x direction, and the measurement angle difference in the y direction, Represents the filtered point cloud data.
[0034] Compared with the prior art, the present invention has the following beneficial effects:
[0035] The present invention analyzes the ranging and angular measurement error characteristics of scanning laser radars and studies the physical process of scanning space by scanning laser radars. Without changing the hardware structure of existing laser radar products, the present invention can improve the quality of point cloud data by preprocessing the point cloud data, thereby improving the accuracy and quality of three-dimensional reconstruction of non-cooperative targets in space. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 is a flow chart of the present invention;
[0037] Figure 2 The point cloud quality effects before and after point cloud filtering. DETAILED DESCRIPTION
[0038] The specific implementation of the present invention will be described in detail below with reference to the accompanying drawings.
[0039] For scanning lidar, the key to improving the three-dimensional reconstruction effect of non-cooperative targets in space is determined by key factors such as the number of point cloud collection, the measurement accuracy of the laser, and the angular measurement accuracy of the laser.
[0040] While keeping the status of existing laser products unchanged, this project proposes a method for filtering scanning lidar point cloud data for three-dimensional reconstruction of spatial non-cooperative targets based on filtering preprocessing of the scanning lidar data and research on the ranging and angle measurement error characteristics of the scanning lidar.
[0041] This method analyzes the ranging and angle measurement error characteristics of scanning laser radars and studies the physical process of scanning space by scanning laser radars. It can improve the quality of point cloud data by preprocessing the point cloud data without changing the hardware structure of existing laser radar products (such as Figure 1 As shown in Figure 3), the accuracy and quality of 3D reconstruction of non-cooperative targets in space can be improved.
[0042] The present invention provides a scanning laser radar point cloud data filtering method for three-dimensional reconstruction of space non-cooperative targets, such as Figure 1 As shown, the steps are as follows:
[0043] The first step is to analyze the ranging accuracy and angle measurement accuracy characteristics of the scanning laser radar and evaluate the ranging accuracy σ d , and angle measurement accuracy σ x , σ y For example, the ranging accuracy of a scanning laser radar is 0.05m, and the angular measurement accuracy is 0.002° and 0.002° respectively.
[0044] The second step is to clarify the scanning LiDAR's on-orbit workflow, data collection process, and mission status. This includes the LiDAR's operating distance from the target, the frequency of data collection and processing, and the point cloud scanning method. For example, a scanning LiDAR's mission condition involves scanning and measuring the target at a distance of 50 meters, circling the target clockwise, with a data collection and processing frequency of 6 Hz and a Lissajous point cloud scanning method.
[0045] The third step is to build a 1:1 model of the space non-cooperative target in the ground laboratory according to the actual mission conditions of the lidar described in the second step, scan the space non-cooperative target model, and obtain a multi-frame, multi-angle scanning point cloud dataset of the space non-cooperative target.
[0046] Step 4: Extract the distance vector of point cloud data and the angle vector Serves as the data basis for subsequent point cloud filtering.
[0047] The fifth step is to analyze the configuration characteristics of the non-cooperative spatial target, evaluate its surface features, and establish target point cloud filtering thresholds τ1, τ2, and τ3 based on spatiotemporal continuity. The filtering is based on the assumption that the distance between the surface of the non-cooperative spatial target and the lidar is spatially continuous and non-sudden during the transient acquisition time. This dictates that the distance values of two laser measurements taken within adjacent time periods should remain within a certain threshold range. Similarly, once the lidar scan rate is determined, the angle values of two laser scans taken within adjacent time periods should remain within a certain angular range. The target point cloud filtering threshold τ1 is determined using the following formula based on the lidar scan angular velocity ω, the lidar data acquisition and processing frequency f, the number of lidar data acquisition points n per frame, and the operating distance h between the lidar and the target.
[0048] The maximum scanning distance between two lasers on the target surface within adjacent time
[0049] Taking the maximum scanning distance s of two lasers on the target surface in adjacent time as the scale, the maximum surface drop is determined as the target point cloud filtering threshold τ1 when the distance s scale is determined on the non-cooperative target surface in space. At the same time, according to the x-direction angular velocity ω of the laser radar scanning x and the y-direction angular velocity ω y , determine the angle filtering thresholds τ2 and τ3.
[0050]
[0051] The sixth step is to filter the point cloud data according to the established target point cloud filtering thresholds τ1, τ2, and τ3 to form a filtered point cloud data set. The dataset before point cloud filtering is The dataset after point cloud filtering is
[0052]
[0053] in, Respectively represent the measured distance, x-direction measured angle, y-direction measured angle after filtering of a frame of scan data, Δd, Δθ x , Δθ y Respectively represent the difference between adjacent scan data, including measurement distance, x-direction measurement angle, y-direction measurement angle, Represents the filtered point cloud data, that is, retaining the difference between adjacent scan data, including the measured distance, x-direction measurement angle, and y-direction measurement angle, which are less than the filtering thresholds τ1, τ2, and τ3, respectively.
[0054] Figure 2 The point cloud quality effects before and after point cloud filtering. Figure 2It can be seen that the point cloud containing excessive errors has been effectively filtered out by the method of the present invention, and the contour of the target represented by the filtered point cloud is clearer and more accurate.
[0055] The present invention also provides a scanning laser radar point cloud data filtering system for three-dimensional reconstruction of spatial non-cooperative targets, including a scanning laser radar task determination module, a scanning point cloud data set acquisition module, a point cloud data vector extraction module, and a point cloud data filtering module.
[0056] Scanning LiDAR mission determination module: used to clarify the scanning LiDAR on-orbit workflow, data acquisition parameters, and mission operating status.
[0057] Scanning point cloud dataset acquisition module: By establishing a 1:1 model of a non-cooperative space target in a ground laboratory, the non-cooperative space target is scanned to obtain a multi-frame, multi-angle scanning point cloud dataset of the non-cooperative space target.
[0058] Point cloud data vector extraction module: extracts the distance vector, x-direction angle vector, and y-direction angle vector of the point cloud data from the point cloud data set obtained by the scanning point cloud data set acquisition module.
[0059] Point cloud data filtering module: establishes a target point cloud filtering threshold based on spatiotemporal continuity, filters the point cloud data extracted by the point cloud data vector extraction module, and forms a filtered point cloud data set
[0060] The target point cloud filtering threshold based on spatiotemporal continuity includes a point cloud filtering distance threshold τ1, a point cloud filtering x-direction angle threshold τ2, and a point cloud filtering y-direction angle threshold τ3.
[0061] The method for determining the target point cloud filtering threshold is the same as above.
[0062] The dataset before point cloud filtering is The data set after filtering by the point cloud data filtering module is
[0063]
[0064] The present invention provides a scanning laser radar point cloud data filtering method and system for three-dimensional reconstruction of non-cooperative spatial targets. The method analyzes the ranging and angle measurement error characteristics of the scanning laser radar and studies the physical process of scanning space by the scanning laser radar. Without changing the hardware structure of existing laser radar products, the method can improve the acquisition quality of point cloud data by pre-processing the point cloud data, thereby improving the accuracy and quality of three-dimensional reconstruction of non-cooperative spatial targets.
[0065] Parts of the present invention that are not described in detail belong to common knowledge among those skilled in the art.
Claims
1. A scanning laser radar point cloud data filtering method for 3D reconstruction of non-cooperative targets in space, characterized in that: Here are the steps: The first step is to obtain the ranging accuracy σ of the scanning laser radar d and angle measurement accuracy σ x , σ y ; The second step is to clarify the scanning lidar's on-orbit workflow, data acquisition parameters, and mission status. The third step is to build a 1:1 model of the non-cooperative space target in the ground laboratory, scan the non-cooperative space target, and obtain a multi-frame, multi-angle scanning point cloud dataset P of the non-cooperative space target. The fourth step is to extract the distance vector of the point cloud data from the data set P obtained in the third step and the x-direction angle vector y-direction angle vector The fifth step is to analyze the configuration characteristics of the spatial non-cooperative target, evaluate the surface characteristics of the spatial non-cooperative target, and establish a target point cloud filtering threshold based on spatiotemporal continuity; The sixth step is to filter the point cloud data extracted in the fourth step according to the established target point cloud filtering distance threshold τ1, point cloud filtering x-direction angle threshold τ2, and point cloud filtering y-direction angle threshold τ3 to form a filtered point cloud dataset.
2. The scanning laser radar point cloud data filtering method for 3D reconstruction of non-cooperative spatial targets according to claim 1, characterized in that: In the first step, the ranging accuracy and angle measurement accuracy of the scanning laser radar are obtained from the technical manual of the laser radar or measured through ground tests.
3. The scanning laser radar point cloud data filtering method for 3D reconstruction of non-cooperative spatial targets according to claim 1, characterized in that: In the fifth step, the method for determining the target point cloud filtering threshold τ1 is: Based on the angular velocity ω of the laser radar scan, the data acquisition and processing frequency f of the laser radar, the number of laser radar single-frame data acquisition points n, and the working distance h between the laser radar and the target, the following formula is used to determine the maximum scanning distance s of two lasers on the target surface in adjacent time: Taking the maximum scanning distance s of two lasers on the target surface in adjacent time as the scale, the distance between the two points with the largest surface drop within the circular area with a diameter of s on any surface of the spatial non-cooperative target is determined as the target point cloud filtering distance threshold τ1.
4. The scanning laser radar point cloud data filtering method for 3D reconstruction of non-cooperative spatial targets according to claim 1, characterized in that: According to the x-direction angular velocity ω scanned by the laser radar x and the y-direction angular velocity ω y , use the following formula to determine the target point cloud filtering thresholds τ2 and τ3; 5. The scanning laser radar point cloud data filtering method for 3D reconstruction of non-cooperative spatial targets according to claim 1, characterized in that: The dataset before point cloud filtering in the sixth step is The dataset after point cloud filtering is in, Respectively represent the measured distance, x-direction measured angle, y-direction measured angle after filtering of a frame of scan data, Δd, Δθ x , Δθ y Respectively represent the measurement distance difference of adjacent scan data, the measurement angle difference in the x direction, and the measurement angle difference in the y direction, Represents the filtered point cloud data.
6. A scanning laser radar point cloud data filtering system for 3D reconstruction of non-cooperative targets in space, characterized in that: It includes scanning laser radar task determination module, scanning point cloud data set acquisition module, point cloud data vector extraction module, and point cloud data filtering module; Scanning LiDAR mission determination module: used to clarify the scanning LiDAR on-orbit workflow, data acquisition parameters, and mission status; Scanning point cloud dataset acquisition module: This module builds a 1:1 model of a non-cooperative space target in a ground laboratory, scans the non-cooperative space target, and obtains a multi-frame, multi-angle scanning point cloud dataset of the non-cooperative space target. Point cloud data vector extraction module: extracts the distance vector, x-direction angle vector, and y-direction angle vector of the point cloud data from the point cloud data set acquired by the scanning point cloud data set acquisition module; Point cloud data filtering module: establishes a target point cloud filtering threshold based on spatiotemporal continuity, filters the point cloud data extracted by the point cloud data vector extraction module, and forms a filtered point cloud data set 7. The scanning laser radar point cloud data filtering system for 3D reconstruction of non-cooperative targets in space according to claim 6, characterized in that: The target point cloud filtering thresholds based on spatiotemporal continuity include a point cloud filtering distance threshold τ1, a point cloud filtering x-direction angle threshold τ2, and a point cloud filtering y-direction angle threshold τ3.
8. The scanning laser radar point cloud data filtering system for 3D reconstruction of non-cooperative spatial targets according to claim 7, characterized in that: The method for determining the target point cloud filtering threshold τ1 is: Based on the angular velocity ω of the laser radar scan, the data acquisition and processing frequency f of the laser radar, the number of laser radar single-frame data acquisition points n, and the working distance h between the laser radar and the target, the following formula is used to determine the maximum scanning distance s of two lasers on the target surface in adjacent time: Taking the maximum scanning distance s of two lasers on the target surface in adjacent time as the scale, the distance between the two points with the largest surface drop within the circular area with a diameter of s on any surface of the spatial non-cooperative target is determined as the target point cloud filtering distance threshold τ1.
9. The scanning laser radar point cloud data filtering system for 3D reconstruction of non-cooperative spatial targets according to claim 7, characterized in that: According to the x-direction angular velocity ω scanned by the laser radar x and the y-direction angular velocity ω y , use the following formula to determine the target point cloud filtering thresholds τ2 and τ3; 10. The scanning laser radar point cloud data filtering system for 3D reconstruction of non-cooperative spatial targets according to claim 6, characterized in that: The dataset before point cloud filtering is The data set after filtering by the point cloud data filtering module is in, Respectively represent the measured distance, x-direction measured angle, y-direction measured angle after filtering of a frame of scan data, Δd, Δθ x , Δθ y Respectively represent the measurement distance difference of adjacent scan data, the measurement angle difference in the x direction, and the measurement angle difference in the y direction, Represents the filtered point cloud data.