High Precision Structural Recovery And Relative Position And Posture Measurement For Space Non-cooperative Targets

Because the structure and motion characteristics of space non-cooperative targets are unknown,the relative pose measurement of space non-cooperative targets has always been a difficult problem.At present,the relative pose measurement of space non-cooperative targets mainly uses stereo vision method at home and abroad,but this method is only applicable to the relative pose measurement of space non-cooperative targets in short distance and small scale.In order to solve this problem,a high-precision structure recovery and relative pose measurement method based on monocular and laser fus ion,which is insensit ive to distance and scale,is proposed.This method realizes the three-dimensional structure recovery of the target and the relative space non-cooperative target pose measurement of the satellite in the process of flying around and approaching the space non-cooperative target with camera load.The main work carried out in this paper is as follows:(1)Tradit ional monocular algorithms need a lot of feature point descriptors and there are a lot of mismatching problems.In this paper,a feature extraction and matching method based on FAST corner and LK optical flow is proposed.After obtaining the feature points and matching relations of the image sequence by this method,the target three-dimensional point cloud and camera parameters are obtained by using the RANSAC five-point method for the first pair of images.For the subsequent added images,the corresponding camera parameters of the newly added images are solved by using the RANSAC EPNP algorithm,and the target three-dimensional point cloud is restored continuously.Finally,BA optimization algorithm is used to improve the accuracy of the target point cloud,and loop detection and optimization algorithm are used to eliminate the cumulative drift error.Experiments show that this method is more accurate than tradit ional methods.(2)In view of the lack of real scale and low accuracy of traditional monocular vision,this paper proposes a relative pose measurement method based on the fus ion of monocular camera and laser ranging.Firstly,the method of projection laser ranging points is used to establish the correlation between laser ranging points and key points of recovery depth.Then,the distance between laser ranging and key points to the center of light of the current camera is used to introduce the true scale for monocular vis ion.Then,the error terms are added to the global BA by using the three-dimensional coordinates of laser ranging points and the three-dimensional coordinates of key points to obtain higher accuracy.Finally,a relocation method is proposed to solve the problem of transforming the camera coordinate system to the target body system.(3)The simulation experiments on the relative circumference and approximation condit ions of space non-cooperative targets show that the three-axis position and the three-axis Euler angle obtained by the fusion method of monocular camera and laser ranging have high accuracy,which verifies the algorithm proposed in this paper.