Structural Components Recognition Method for Malfunctioned Satellite based on Domain Randomization

In the process of long distance non-cooperative target recognition, due to the lack of prior information, it is necessary to identify the structural components of the non-cooperative target in order to formulate the next step of approaching and capturing tasks. Most of the existing satellite recognition algorithms are based on known targets, which have poor generalization ability and cannot cope with the complex shape characteristics of space targets. This paper proposes a satellite structural components recognition algorithm based on domain randomization. Firstly, SolidWorks software is used to build the three-dimensional model of the satellite. Then, the satellite structural components dataset is generated by randomly rendering the structural components and background of the satellite 3D model. The labeled model is trained by YOLOv4 network. Finally, a physical experiment is carried out, and the experimental results show that this method provides a feasible intelligent engineering scheme.