基于深度学习的改进卫星检测方法  

An Improved Satellite Detection Method Based on Deep Learning

作　　者：翁岩松 杨洁[1,2] WENG Yan-song;YANG Jie(School of Artificial Intelligence,Hangzhou Dianzi University,Hangzhou 310018,China;Research Center for Intelligent Transportation,Zhejiang Lab,Hangzhou 311121,China)

机构地区：[1]杭州电子科技大学人工智能学院,杭州310018 [2]之江实验室智慧交通中心,杭州311121

出　　处：《航空电子技术》2024年第4期1-8,共8页Avionics Technology

基　　金：自组织任务分配机制及算法研究基金(2023AIASS0601)。

摘　　要：针对航天活动频繁导致空间物体数量激增及安全管理问题日益凸显的现状,本文提出了基于深度学习的卫星识别方法。通过改进YOLOv8模型,增加了全局注意力机制模块,提升了模型深度挖掘和提取图像中的细微特征的能力,能够较好地实现对卫星和其部件的识别与分类。实验结果表明,改进后的模型在识别精度上有较大的提升,优化了卫星识别过程,帮助地面人员减少了信息解读的误差,为航天领域的安全管理提供了更加可靠的技术支持。In response to the rapid increase in space objects due to frequent space activities and the growing safety management challenges,a deep learning-based satellite recognition method is proposed.The approach improves the YOLOv8 model by incorporating a global attention mechanism module,enhancing the model's ability to deeply mine and extract subtle features from images.This method effectively identifies and classifies satellites and their components.Experimental results demonstrate a significant increase in recognition accuracy with the improved model,optimizing the satellite recognition process and reducing interpretation errors for ground personnel.This provides more reliable technical support for safety management in the aerospace sector.

关 键 词：人工智能 目标检测 卫星识别 深度学习 

分 类 号：V52[航空宇航科学与技术—人机与环境工程]