Research On Motion Estimation Based On Binocular Vision And 3D Reconstruction For Non-cooperative Targets

Space robot technology has become an important means of spacecraft on track maintenance and space garbage cleanup,but the non-cooperative target motion and structure information are unknown,which brings great challenges to the capture work.Therefore,in order to achieve accurate capture of the target,we need to estimate the target motion and perform three-dimensional reconstruction in the space environment to obtain its motion and structure information.At present,the vision-based non-cooperative target measurement technology has become a hotspot in the field of space robot technology.Firstly,this paper analyzes the imaging principle of binocular stereoscopic vision.On this basis,the binocular visual mathematical model is established and the relationship between the coordinates of the spatial point and its image projection point is deduced.The binocular camera is calibrated and the distortion image is corrected with the obtained parameters,and the corrected target image is performed with target area segmentation and the interference of the environment noise is reduced.Then,the image feature matching and tracking problem are studied.For the feature matching,a gradual purification strategy based on the nearest neighbor ratio matching,cross validation matching and RANSAC optimization matching is adopted.The matching effect of the commonly used feature extraction algorithm is compared.Finally,the ORB algorithm is used as the feature extraction algorithm in this paper.For the feature tracking,combining the optical flow advantages of HS and LK,the joint-LK optical flow method is proposed,and the ORB feature point is used as the input of the Joint-LK optical flow to form the feature optical flow,thus realizing the accurate tracking of the matched feature point.Secondly,the target motion estimation and three-dimensional reconstruction method are studied.For the target motion estimation,the kinematic model of the non-cooperative target is established.Then,the optimal target function of the pose transformation matrix in adjacent time is established under the condition that the coordinates of the feature points are known.The SVD algorithm is used to optimize the target function,and the pose transformation matrix of the target in adjacent time is obtained,motion parameters of the whole motion process are obtained in further step.For the three-dimensional reconstruction of the target,the 3D coordinates of the feature points are transformed into the initial time by using the pose transformation matrix,and the three-dimensional point cloud is obtained.The three-dimensional reconstruction of the target is carried out by using the triangulation principle.Finally,in order to verify the feasibility and accuracy of the proposed method,the experimental study is carried out under different motion forms and different light sources.The experimental results show that the proposed method can complete the motion estimation and three-dimensional reconstruction of the non-cooperative target.