Research On Visual Measurement Method Of Non-cooperative Spacecraft

Capturing technology of non-cooperative target, aiming at on-orbit service ofthe spacecraft, eliminating the obsolete satellites, space attack and defense, has beena research focus in the field of space robot technique, recently. In order to measurethe relative pose between the two non-cooperative spacecrafts in the final approachnavigation of capture task, the method based on stereo vision measurement for thenon-cooperative spacecraft is studied in this article.Firstly, the mathematical model for the stereo vision system, which isnecessary for the following research, is introduced. Each camera is modeled asnonlinear camera perspective projection mode. Then the mathematical model of thebinocular stereo vision is stated in detail according to the epi-polar geometry, andthe method for the3D reconstruction is presented.Secondly, a novel method based on information fusion is proposed to measurethe relative pose between the two non-cooperative spacecrafts. As, there is noartificial patterns on the targets, two kinds of features on the spacecraft are chosenas the recognized features. A method which combined point feature detection withthe line feature detection is utilized to recognize the non-cooperative target. Inaddition, the polygons recognition algorithm and line feature detection algorithm ismodified to improve the processing speed and robustness of the vision system.In order to achieve the relative pose measurement in the dynamic tracking ofthe non-cooperative target, a new method which combined the Kalman filter methodwith pyramid Lucas-Kanade optical flow algorithm is proposed. This method usesPLK algorithm to track the strong feature points dynamically and takes the output ofthe PLK algorithm as the observation of Kalman filter to estimate the optimal state.Then, the optimal state estimation value is taken as the initial PLK values of thenext frame image tracing processing. In this way, the real-time performance of thevision system can be improved effectively.Finally, the experimental platform of vision measurement system is built up toverify the pose estimation algorithm and the dynamic tracking algorithm. Theresults show that the accuracy of the measurement, robustness and the real-timeperformance of the proposed algorithms can satisfy the measurement accuracy anddynamic tracking requirements in the final approach navigation of capture task.