An Effective Procedure to Build Space Object Datasets Based on STK

With the development of space technology, deep learning methods, with their excellent generalization ability, are increasingly applied in various activities. The object data is difficult to obtain, which greatly limits its application images existing public spacecraft dataset mostly rendered, not only lack physical meaning but also have limited data. In this paper, we propose an effective construction procedure build a based on STK, can help break limitations methods Firstly, conduct orbit simulation for 24 targets and establish dataset; secondly, use 600 6 typical label them real-shot validation dataset. Finally, constructed STK verified be through six semantic segmentation networks, used train real spacecraft’s segmentation. Lots experiments show that accuracy migrating training results shooting slightly reduced, mPA still greater than 85%. particular, after adding orbital physics data, generally improved. Therefore, STK-based method construction.