Research On Attitude Estimation Algorithm Of Scanning Imaging Lidar Sparse Point Cloud Data

With the gradual acceleration of space exploration,the relative pose measurement technology in the process of spacecraft rendezvous and docking is regarded as the focus of the research of the world’s aerospace powers,and gradually develops from cooperative target measurement to non-cooperative target measurement.Lidar is a measurement technology that actively obtains the three-dimensional space information of the target.It has the characteristics of small size,light weight,and high precision.It is a key measurement sensor for spacecraft rendezvous and docking.The beam scanning device is used by the scanning imaging lidar to obtain point cloud data based on the laser ranging system.It is the mainstream lidar at this stage because of its long detection distance and mature and reliable system.Because of the sparse single frame point cloud,small point cloud data size and fast calculation speed of scanning imaging lidar,it is more suitable for working environment where is required for rendezvous and docking in non-cooperative target attitude measurement.Attitude solution is the key to the space non-cooperative target attitude measurement technology.In recent years,scholars at home and abroad have proposed the method which used point cloud registration,model matching,Kalman filter and other methods to solve the problem of non-cooperative target pose estimation.There are still some problems to be solved such as the unperfect theoretical system of sparse point cloud processing algorithm,the difficulty of the absolute pose obtaining,and the low efficiency of algorithm.In order to solve these problems,the attitude estimation algorithm for non-cooperative targets based on sparse point cloud data of scanning imaging lidar has been researched in this paper.An attitude measurement method is proposed,which combines the absolute attitude acquisition of the initial frame and the fast and high-precision calculation of the relative attitude of the subsequent frames.The main contents include:1.The basic principles and main technical indicators of attitude measurement based on LiDAR sparse point cloud and the accuracy requirement of rendezvous and docking attitude measurement are introduced,the attitude measurement model of sparse point cloud is analyzed,and the research status of attitude measurement technology and point cloud registration methods at home and abroad is summarized.2.The characteristics of sparse point cloud data are studied,the principle of lidar imaging is introduced,and the characteristics of sparse point cloud data structure and noise are analyzed.A preprocessing algorithm based on grid reduction and compound filtering is proposed,based on existing noise reduction and reduction methods.3.The relative attitude measurement is divided into two stages: absolute attitude acquisition and relative attitude solution.Firstly,the attitude measurement method of sparse point cloud under the influence of target self-occlusion is studied.We analyze the application defects of existing registration algorithms in pose estimation of noncooperative targets.On this basis,an initial pose acquisition algorithm using point cloud segmentation and feature matching is proposed.The point cloud data is segmented by RANSAC and the region growing method,the point cloud feature confidence probability is calculated.Finally,the spatial transformation matrix of reference data with reference point cloud is solved to obtain the absolute attitude.The feasibility of the algorithm to obtain the initial attitude of the target by using the sparse point cloud is verified by the simulation experiment and the actual measurement experiment of the satellite model.It can obtain the absolute attitude of the target accurately and robustly.4.Aiming at the accuracy and efficiency of the registration algorithm in the relative pose tracking stage,we propose a point cloud registration algorithm based on feature vector extraction.And the relative pose is calculated by frame point cloud registration.The algorithm uses the point curvature and the number of points in the neighborhood to filter the feature points;then performs principal component analysis on the feature points to extract the feature vector and establishes the transformation relationship between the source point cloud and the target point cloud to achieve rough registration.In the fine registration stage,k-dimensional binary tree nearest neighbor search is introduced to improve the precision registration efficiency of the ICP algorithm.The effectiveness and feasibility of the algorithm are verified by the simulation and actual point cloud registration experiments.5.The satellite model attitude measurement platform is built and the proposed method is used to measure the satellite model attitude.Experimental results show that the proposed method can perform attitude measurement robustly when the target is at a large angle to guide the pursuit spacecraft to approach the target spacecraft,and it can complete attitude calculation quickly and accurately when the target is at a small angle,to achieve the requirements of rendezvous and docking attitude measurement.