Dense 3D Reconstruction of Non-cooperative Target Based on Pose Measurement

Abstract

To accurately identify the three-dimensional (3D) structure information of non-cooperative targets in on-orbit service tasks, this paper proposed a 3D reconstruction method combining pose measurement and dense reconstruction. Firstly, based on the depth camera, the Iterative Closest Point (ICP) algorithm is used to realize the pose measurement between each frame of the non-cooperative target. Then, the loopback detection based on the bag-of-words method is adopted to eliminate the cumulative error of pose measurement. At the same time, the sparse 3D model of the non-cooperative target can be obtained by combining feature point extraction and matching. Furthermore, secondary sampling is performed in the tracking and local mapping threads by increasing the dense mapping thread. Then, the loop thread is used to update the pose in turn. After the global pose optimization, a more accurate 3D reconstruction model can be obtained. By using point cloud optimization, the accurate dense 3D reconstruction model of a non-cooperative target is constructed. Finally, numerical simulations and experiments verify the accuracy and effectiveness of the proposed method.