Research On 3D Reconstruction Method And Precision Analysis For Non-Cooperative Space Targets Using Image Sequences

The in-orbit space-related technologies and services of the high orbit malfunction satellite removal and waste objectives has become a hot topic States. The rail service typically required to obtain three-dimensional structure of the target model and size, to determine the appropriate capture sites on th e target, identify and repair the docking site which directly determines the success or failure of the arrest, docking, maintenance and other operations. However, most of the satellites in orbit servicing spacecraft, accepting services of non-cooperative target satellite in orbit. Therefore, to achieve a true space-orbit service, we must solve the three-dimensional structural model of non-cooperative target reconstruction. The monocular vision three-dimensional reconstruction method mostly using for ground processing target of the scene. However, taking into account specialty of the space environment and non-cooperative target, the target image feature point extraction and matching requirements are high, while the non-cooperative target of three-dimensional reconstruction accuracy is the most critical issue mission for designers, but also an important basis for guiding mission design and hardware selection, directly affects the follow-up mission successful implementation. We have yet to see a comprehensive a nalysis of various factors affecting research and error propagation mechanism. Therefore, reconstruction error propagation mechanism, analyzing the main sources of error and building a reconstruction precision to require hardware configuration and task design methods and processes seem to very essential. The main work of the thesis includes:Firstly, starting from the sequence image acquisition, expatiating spatial non-cooperative target feature point extraction and matching method. Based on affine optimization model, put forward the new method for the optimization of feature points matching. Meanwhile comparing the simulation results between SIFT algorithm and affine optimization.Secondly, teasing out spatial non-cooperative target sequence image 3D point clouds reconstruction method. Putting forward bundle adjustment method to optimize sparse three-dimensional point cloud reconstruction approaching.Thirdly, Focusing on the precision of 3D reconstruction for non-cooperative space target by image sequence, using a pinhole camera model derivation method and the re-projection method for systematic theory analyzing. Meanwhile, setting some comparison experiments for verification.Finally, doing 3D reconstruction experiments in many ways, come up with the reasons for departing from the experimental point of view, then making reasonable analysis and proposing methods for improvement.