Research On Pose Estimation And 3D Reconstruction Of Non-cooperative Target Based On Structured Light Vision

In the process of exploring space,although human beings have achieved excellent research results,a lot of trash have been left in space.Space trash will collide with the International Space Station,spacecraft and artificial satellites,which causes great potential safety hazards.How to clean up space trash has become one of the most important and widely-discussed topics in the world.However,the unknown motion state and 3D structure of space trash make it difficult for space robots to clean them.Therefore,it is necessary to obtain the motion and structure information of space trash by some specific methods.Many scholars have found that the pose estimation and 3D reconstruction technique based on vision is an effective solution to this problem.In this paper,the structured light 3D vision system is used to track the target.Based on the pinhole imaging model of the camera,the principle of structured light depth camera is introduced,the structured light 3D vision model is established,and the method of obtaining 3D coordinates of spatial points by structured light 3D vision system is deduced.After obtaining the images of the target,the image preprocessing is conducted,which mainly includes image segmentation and image morphology processing.The feature points of the target surface need to be acquired after finishing the image preprocessing.By analyzing and comparing the extraction effects of several commonly-used feature extraction algorithms,the SIFT algorithm which can get the most uniform distribution of the feature points is finally selected.For the topic of target tracking,a pyramid LK optical flow tracking method based on S IFT algorithm is adopted to realize the matching of feature points between the front frame and the back frame.In order to achieve the real-time tracking,this paper uses a camera in the moving state to track the target.The pose estimation and 3D reconstruction method of the target based on EKF algorithm is applied.First,the principle of EKF algorithm is introduced,and the iterative estimation formulas are deduced.After that,an estimation algorithm based on EKF-IMU is proposed.The motion state of the camera is estimated accurately by EKF algorithm with the motion parameters obtained by IMU.Then,based on the principle of EKF algorithm,the state model and observation model of the target and the camera are established.In order to reduce the dimension of the state vector and solve the coupling estimation problem of Joint EKF algorithm,a Tri-EKF algorithm for sequential estimation is proposed.Rotational motion filter,translational motion filter and structure parameter filter are used to complete the estimation process in turn.Finally,the effects of Tri-EKF algorithm and Joint EKF algorithm on pose estimation and 3D reconstruction of non-cooperative target are compared through MATLAB simulation experiments.The experimental results show that Tri-EKF can achieve higher convergence speed and estimation accuracy for the translational motion of the target.Finally,an experimental scheme is designed to verify the algorithm.The overall framework of the proposed algorithm is summarized,and the experimental platform which includes the camera motion platform,the non-cooperative target motion platform and the structured light 3D vision system is built.The pose estimation and 3D reconstruction of the non-cooperative target are realized for different motion forms,such as translational motion,rotational motion,and the compound motion of translation and rotation.The proposed algorithm is evaluated by analyzing the experimental results.