摘要
为了实现载人航天器综合测试期间交会对接灯故障的快速准确识别,提出了一种应用改进YOLO v8模型+注意力机制的交会对接灯故障目标检测模型。该模型在改进YOLO v8模型的基础上,在加强特征提取网络上引入注意力机制模块,对故障目标给予更多关注,提高了密集小目标识别精度。同时,在自制灯数据集上进行仿真试验。试验结果表明:文章提出的模型平均准确率提高为0.91,可以对灯的部分点失效故障目标进行快速、有效识别,为载人航天器综合测试期间交会对接灯智能巡检提供了一种新思路。
In order to achieve rapid and accurate identification of rendezvous and docking light faults during comprehensive testing of manned spacecraft,a fault target detection model for rendezvous and docking light using improved YOLO v8 model and attention mechanism is proposed.Based on the improved YOLO v8 model,the new model proposed introduces an attention mechanism module into the strengthened feature network to pay more attention to faulty targets and improve the recognition accuracy for tense small targets.At the same time,simulation experiments are conducted on a self-made dataset,and the results show that the model proposed has an average accuracy of 0.91,which can effectively identify the partial fault targets of light and can provide a new way for intelligent inspection of rendezvous and docking light during manned spacecraft comprehensive testing.
作者
黄连兵
薛霞
李立凌
尹桂松
齐天哲
芮康
HUANG Lianbing;XUE Xia;LI Liling;YIN Guisong;QI Tianzhe;RUI Kang(Beijing Institute of Spacecraft System Engineering,Beijing 100094,China)
出处
《航天器工程》
CSCD
北大核心
2024年第4期130-136,共7页
Spacecraft Engineering
