基于YOLO的月表撞击坑检测实验方案设计

Experimental scheme of Lunar craters detection based on YOLO network

月表撞击坑的自动提取可为星际探测器着陆的自动选址提供可靠参考,可直接服务于军民用航天领域的重要需求。然而,月表撞击坑尺度多变、分布不一的特性给有效的撞击坑自动提取方法的开发带来了挑战。该文针对航空航天信息类专业实践培养需求,设计了基于YOLO的月表撞击坑检测实验方案。首先构建月表撞击坑训练用数据库,然后搭建基于YOLO系列网络的深度学习模型,并对撞击坑检测效果进行测试,最终所训练的网络模型在测试集上的检测精度能达到97.7%。通过该实验方案的设计,可以使学生在实践中深入理解航空航天领域问题研究的思路和方法,提高学生科研探索的兴趣和动手能力。

The automatic extraction of impact craters on the Lunar surface provides a reliable reference for the automatic location of interplanetary probes for landing, and can directly serve the important needs of the military and civil aerospace fields. However, the variable scale and uneven distribution of Lunar craters make it challenging to develop an effective automatic impact crater extraction method. In this paper, a Lunar surface impact crater detection experiment scheme based on YOLO is designed to meet the practical training needs of aerospace information majors. First, the lunar crater training database is built, then the deep learning model based on YOLO series network is trained, and the effectiveness of impact crater detection is tested. Finally, the detection accuracy of the trained network model on the test set can reach 97.7%. Through the design of the experimental scheme,students can deeply understand the ideas and methods of research in the aerospace field in practice, and improve their interest and practical ability in scientific research and exploration.