文献类型：期刊文献

英文题名：A Pose Estimation Method for Non-Cooperative Spacecrafts Based on 3d Bounding Box Detection

作者：Chen, Liwei[1]; Wu, Bin[2]; Yi, Jianjun[1]; Xia, Wenxuan[1]; Su, Lin[1]; Wu, Hailei[3,4]

机构：[1] Department of Mechanical Engineering, East China University of Science and Technology, Shanghai, 200237, China; [2] Aerospace System Engineering Shanghai, NO.3888 Yuanjiang Road, Minhang District, Shanghai, 201109, China; [3] Shanghai Aerospace Control Technology Institute, Shanghai, 201109, China; [4] Shanghai Key Laboratory of Aerospace Intelligent Control Technology, Shanghai, 201109, China

年份：2023

外文期刊名：SSRN

收录：EI(收录号：20230029312)

语种：英文

外文关键词：Orbits

摘要：Relative pose estimation of a non-cooperative spacecraft is the basis of many in-orbit space tasks. In many cases, only visible light based observation methods can meet the real-time and feasibility requirements of space tasks. In this paper, we design a neural network with image as input, which can effectively extract the image and global structure features of a spacecraft with the symmetrical structure of double panels to estimate the positions and relative depths of 9 key points of the 3D bounding box fitting the main body of the spacecraft. On this basis, our post-processing method is able to further optimize the detection results by recovering the scale of relative depth detection result and a simplified ICP registration method, so as to obtain the accurate pose estimation from the detection results with errors. Experiments show that our method performs well on non-cooperative spacecrafts under different lighting conditions. ? 2023, The Authors. All rights reserved.