The Detection of Non-Cooperative Targets in Space by Using 3D Point Cloud

Abstract

In the space operations fields, spatial targets that do not have cooperative signals and capture characteristics are called non-cooperative targets. The use of non-cooperative target docking capture and space robot technology is an effective way to carry out the tasks such as satellite unattended maintenance and orbit adjustment. The traditional method of detecting non-cooperative targets using RGBD mainly uses the color and texture information of the objects. However, unstable space light will lead to mismatch or poor robustness of traditional methods. To solve this problem, this paper proposes a non-cooperative target detection and transformation matrix estimation method by 3D point cloud information from the sensor. First, use the depth camera to sample the object at different angles of equidistance to create a template point cloud composed of point cloud models of different poses. Then, the scene is sampled by the sensor and the FPFH feature is extracted. Finally, based on the FPFH feature of the point cloud, the SAC-IA and the ICP algorithm are used to obtain the pose information of non-cooperative targets in the scene. On this basis, the robotic arm can accurately capture non-cooperative targets in space.