Research And Design On Relative Pose Measurement And Tracking For Non-cooperative Target

With the exploration of outer space, the incessant development of on-orbit satellitedefense and satellite repair and space debris cleaning missions, the importance ofnon-cooperative target tracking and pose measurement technology is becoming moreand more significant. In this dissertation, we have mainly studied the tracking and therelative position and attitude measurement for non-cooperative target based on thebinocular vision. In addition, the tracking and measuring system have been designedand achieved in this dissertation.Firstly, the development status of the non-cooperation measure and tracking isintroduced in this dissertation. The models for binocular vision measurement and3Dreconstruction have been also established, especially, the method of3D reconstruction isstudied in the case of epipolar constraint. Then, the binocular stereo calibration andstereo rectify experiment have been completed in this dissertation.Secondly, a scheme for non-cooperative target’s relative position and attitudemeasurement is proposed, which is based on the binocular vision. In the proposedscheme, the typical characteristics of the spacecraft are used as the recognitioncharacteristics, which the model of non-cooperation measurement is established by. Inorder to ensure the real-time and stability of attitude measurement, the combinationmethod of the panel contour and nozzle contour has been adopted.The combinedmethod can ensure the accuracy of non-cooperative target recognition. The noisesuppression function has been introduced into SSD（Sum of Squared Differences）matching method, which can improve the speed and accuracy of image matching.Thirdly, in order to realize the dynamic tracking for non-cooperative target, theMean Shift tracking algorithm has been studied in-depth. An improved Mean Shiftalgorithm (LBP and Scale adaptive Mean Shift, LSMS) is proposed by adopting theadaptive scale and the LBP feature. Multiple sets of standard image sequences havebeen used to test and verify the effectiveness of LSMS algorithm. The simulation resultsshow that the improved algorithm is superior to the original Mean Shift algorithm inadaptive scale, the number of iterations,the tracking error aspects and so on.Finally, the binocular vision measurement and tracking of physical experimentplatform is constructed by using VC++and OpenCV. At last, the physical simulationand error analysis on the non-cooperative measurement scheme proposed in thisdissertation and the target tracking based on LSMS method is carried through.