Research On Binocular Vision Measurement For Non-cooperative Target Based On Point Cloud

With the growing of people's space exploration and space activities, Pose Measurement, as a critical technology of capturing mission to spatial non-cooperative targets, has been highly thought of by the main aerospace countries. And in this article, with having carried out extensive research on this issue, we proposed a pose estimation scheme using a binocular camera system and based on point cloud.By looking at non-cooperative target acquisition and the research progress of non-cooperative target visual measuring technique, analyze the existing idea, method of each scheme, technical feasibility and limitations, etc. On the basis of that, I have taken in-depth understanding of the measurement based on binocular vision visual theory. At the same time, using the planar target method to calibrate the binocular camera used in the experiment.Take the program binocular based on stereo matching to obtain point cloud, including the feature point extraction and feature-based stereo matching algorithm. Analyze and contrast the principle and performance of three characteristics extraction algorithm. Then combined with the BBF search and RANSAC algorithm, we get optimized matching results. Further we compute the three-dimensional coordinates of the corresponding spatial feature points according to the measurement principle, to obtain the target point cloud data. Next, we compare and analyze the two kinds of triangulation algorithms, the Crust algorithm and the fast triangulation algorithm based on Mesh-Growing. On this basis, we reconstruct the scattered point cloud to three-dimensional model.Finally, through the study of the principal component analysis, we designed pose measurement program based on Personal Computer Assistant(PCA). And using VC2010 and OpenCV libraries and PCL pose constructed measurement simulation platform to verify the effectiveness of this method.