Research On Pose Estimation Of Space Unknown Moving Targets Based On Binocular Vision

It is valuable to use robots to grasp space unknown moving targets or to perform on-orbit maintenance on faulty spacecraft,which needs to be studied.Among them,pose estimation is a key technology for robots to perform grasping tasks or on-orbit maintenance in close range,but there are some problems such as low accuracy of pose measurement,slow processing speed and sensitivity to occlusion.In this paper,a binocular vision measurement technique combined with iterative nearest point(ICP)algorithm and adaptive extended Kalman filter(AEKF)is proposed to estimate the relative pose of space unknown moving targets in close range.Firstly,binocular stereo vision system is used to obtain the 3D point cloud data of the target.The camera's imaging model is described.This paper studies the stereo matching technique,and improves the ELAS(Efficient LArge-scale Stereo)matching algorithm by selecting adaptive support points,thus improving the accuracy and speed of this algorithm.The segmentation of point cloud based on Euclidean clustering and the extraction of target based on the color and depth are studied.The target is extracted from the background,and the obtained point cloud is processed by noise removing and down sampling.Secondly,point cloud registration based on improved ICP algorithm is proposed.Aiming at the problem that traditional ICP algorithm is easy to fall into local optimum,this paper proposes an ICP algorithm initialization based on point cloud global registration,and obtains a better initial value.Aiming at the slow registration speed of traditional ICP algorithm,this paper proposes an acceleration of ICP algorithm based on K-D tree searching,which improves the registration speed.Then,pose estimation algorithm based on improved ICP algorithm and adaptive EKF is proposed.The initial pose of the camera and the target is obtained by initial pose estimation based on feature point registration.The pose of adjacent key frames is obtained by an algorithm based on improved ICP algorithm.Finally,the real-time pose of the camera and the target can be solved by ICP-AEKF algorithm.The closed-loop configuration of improved ICP algorithm and adaptive EKF can not only solve the initialization problem of ICP algorithm,but also filter out the noise and improve the accuracy.At the same time,continuous pose parameters can be estimated when the target is temporarily out of sight,thus improving the robustness of the system.Finally,the proposed pose estimation method is simulated and a physical experiment platform is built to verify the effectiveness of this proposed method.The target is fixed at the end of the manipulator,and the binocular camera is fixed at a certain position outside the manipulator,and the moving range of the target is ensured in the field of view of the binocular camera.Through the camera calibration,the camera model parameters are obtained,and the transformation between camera coordinate system and robot coordinate system is obtained by hand-eye calibration.And trajectory planning of UR manipulator are studied.The comparison experiments show that the proposed method is effective in accuracy,speed and robustness to occlusion.