Research On The Vision-based Measuring Method Of Non-cooperative Targets And System Implementation

Due to the characteristics of the low-cost, low-risk, high concealed ability and military value, on-orbit autonomous service has been the new research focus in the field of space technique. In order to solve the problem of the non-cooperative target relative pose measurement in final approach for the on-orbit service, a study on vision-based measuring method, specifically the stereo vision-based measuring method, was developed to measure the relative pose (position and attitude) of a non-cooperative target, and a measuring system was established based on DSP. The main content and contributions are as follows:1. A features detecting and tracking algorithm of non-cooperative targets was proposed based on the Hough transform. Because of the bad stability of point-features, the algorithm employed the high stability of line-features and the Hough transform was applied to recognize and detect the features. The experiment result show that the algorithm can locate the target exactly and stably2. A fast area-based matching algorithm based on local features was proposed. First, matching areas were built centered the local features. Then, the fast area-based matching algorithm was applied including epipolar line constraint and box-filtering technique. The experiment result showed that the algorithm can increase the speed remarkably.3. A 3D-reconstruction and pose solution and optimization method based on the generalized inverse was proposed. First, the generalized inverse was employed in the point reconstruction. Then, the generalized inverse was employed to improve the vectors-based pose solution. Finally, an optimization method was applied in the attitude matrix. The measuring experiment result showed that the method can obtain high measuring precision.4. A relative pose measuring system of non-cooperative target constituted of two CCD cameras, light sources and DSP processing system was established based the above research. The result of satellite model measurement demonstrate that the relative position errors are less than±20mm, relative attitude errors are less than±2°, and measuring speed is up to 8fps.