摘要
为了解决灾区的特殊地形与复杂环境给救援工作造成难度大的问题,研究了深度学习在无人机救援中的应用。固定翼无人机结构简单、成本低、部署快,能够高效地完成飞行任务。对比环形搜寻、扩展方形搜寻和“8”字形搜寻3种搜寻方式优势,搭载深度学习智能系统,通过引入Faster RCNN算法、多尺度特征提取融合算法、RPN网络优化算法和区域的目标识别增强算法能够提高救援的精准度和效率,为促进无人机救援的进一步发展提供参考。
In order to solve the difficult problem caused by the special terrain and complex environment in the disaster area,the application of deep learning in UAV rescue was studied.The fixed-wing UAV has simple structure,low cost,fast deployment,and can efficiently complete the flight mission.Compared with the advantages of the three search methods of ring search,extended square search and"8"shaped search,equipped with a deep learning intelligent system,the introduction of Faster RCNN algorithm,multi-scale feature extraction fusion algorithm,RPN network optimization algorithm and regional target recognition enhancement algorithm can improve the accuracy and efficiency of rescue,and provide a reference for promoting the further development of UAV rescue.
作者
徐超
XU Chao(Hebi Polytechnic,Hebi 453080,China)
出处
《科技创新与生产力》
2023年第2期86-89,共4页
Sci-tech Innovation and Productivity
基金
河南省科技攻关项目(212102310550,212102310488)
鹤壁职业技术学院科技重点课题(2021-KJZD-006)。
关键词
固定翼无人机
救援
特征提取融合
网络优化
目标识别增强
fixed-wing UAV
rescue
feature extraction and fusion
network optimization
target recognition enhancement
