The Pose Measurement For Non-cooperative Target Based On Stereo Vision

Satellite missions in space include in orbit services,rendezvous and docking,space target monitoring,etc.And the accurate positioning of the target satellite relative to the position and attitude of the executing satellite is the necessary condition to complete these tasks.Based on the stereo vision measurement system,this paper designs a non-cooperative target pose measurement method based on the existing research results,and improved the traditional measurement method aiming at the specific problems in the experiment.a non-cooperative target pose measurement method based on stereo vision is designed,and it mainly includes the following aspects:1)The background and significance of the project are analyzed systematically.The research status of binocular stereo vision,multi-target tracking,binocular stereo matching and camera calibration at home and abroad are investigated,and the overall research ideas are established.2)The camera model,camera distortion model and spatial projective geometry principle are described.And the traditional calibration methods are summarized.In the process of experiment,the problem that the camera calibration error is too large under the condition of long depth of field is encountered.It is found that the main reason for this problem is that the calibration result of small target for long depth of field can only be local optimal.Aiming at this problem,a calibration method based on target with multi position and small standard is designed,which can effectively reduce the overall error of long depth of field calibration.3)The space background conditions of different situations are analyzed,and then the different processing results of different backgrounds in the experiment are analyzed.A set of image preprocessing method is designed under the condition of simple shooting background and good effect,and a kind of image preprocessing method is designed under the condition of strong light background.This method can effectively solve the problem that the traditional method cannot recognize the target in the strong light background.4)The physical characteristics of the target to be measured are analyzed,the feature points are identified by taking the straight line of the satellite rectangle and the ellipse of the docking ring as the target.After comparing several recognition methods,Hough transform is used to identify the rectangular line structure of satellite body,and use the method of Canny edge detection and ellipse fitting to recognize the ellipse of the docking ring and measure the coordinates of the center of the circle.By tracking the feature points of 3D reconstruction based on the global nearest neighbor filter,the method can complete the tracking of the target in the case of image frame missing to increase the accuracy and5)the whole experiment is carried out.The program is written by computer vision library Open CV3.4.1 in the development environment vs 2017.Firstly,two groups of non-cooperative target images with different positions and different postures were taken in the Physical experiment,then through preprocessing,feature recognition and 3D reconstruction,the real coordinates of feature points are restored.In order to verify the accuracy of the experimental results,another set of 3D reconstruction method of cooperative target is designed.According to the 3D reconstruction space distance between the target punctuation on the cooperative target and the corresponding non cooperative target satellite body corner point,the accuracy of the experimental results is verified by comparing with the real measurement distance.At the end of the paper,the achievements and existing problems in the overall design are summarized and prospected.