西安电子科技大学学报（自然科学版）2023，Vol.50Issue(6)：120-132,13.DOI:10.19665/j.issn1001-2400.20230307
一种超轻量化卫星目标形心定位算法
Lightweight centroid locating method for the satellite target
摘要
Abstract
The space-based photoelectric detection unit is vital in satellite identification and positioning.It has a large field of view,a small load,and flexible maneuvering properties.However,the computational capability of the CPU mounted on an on-orbit satellite could be much higher,which can hardly afford the necessity of deep learning networks.In this paper,we analyze the character of space targets deeply and designed a lightweight real-time processing algorithm.We specifically design feature extractors for the line and satellite contour patterns in the algorithm and propose a minimum bounding box calculation strategy.The algorithm is tested on the simulation dataset and verified on the images captured by the physical emulation platform.Testing results demonstrate the effectiveness of our algorithm.The detecting accuracy of our algorithm is better than YOLOv5n,and the computational load is only 10%that of the competitive methods.We transplant the algorithm to an on-orbit real-time processing platform.The positioning speed reaches 5 120×5 120@5fps.The accuracy for angle measurement based on the centroid positioning results is more significant than 0.05°,which meets the requirement for real application systems.关键词
空间站/卫星/目标定位/目标检测/实时处理Key words
spacecraft/satellites/target positioning/object detection/real-time processing分类
信息技术与安全科学引用本文复制引用
刘路远,韩璐瑶,李娇娇,夏晖,饶鹏,宋锐..一种超轻量化卫星目标形心定位算法[J].西安电子科技大学学报（自然科学版）,2023,50(6):120-132,13.基金项目
芜湖-西电产学研合作专项资金(XWYCXY-012021002) （XWYCXY-012021002）
国家自然科学基金(61901343) （61901343）
111计划(B08038) （B08038）
地理信息工程国家重点实验室开放基金(SKLGIE2020-M-3-1) （SKLGIE2020-M-3-1）
中国博士后科学基金(2018T111019) （2018T111019）
陕西高校青年创新团队 （）
