摘要
目前,我国正在大力发展海洋武器装备,其无人化研究得到广泛关注,其中海上无人艇智能化是研究热点。针对我军对海上大中型目标检测和高精度定位的需求,进行基于深度学习的海上无人艇目标识别技术设计。设计了多源、多体制协同感知架构,以解决设备智能计算任务重复与资源浪费和深度学习加速问题;进行多层次特征提取、分析、融合技术设计,确定单/多传感器特征选取对象;开展基于深度学习的多特征目标检测、识别技术设计,建立基于深度学习网络的多源多维联合检测、识别处理方法。实验验证结果表明,所设计的方法对可见光图像识别率达99.7%以上,具有良好的识别效果。
At present,China is vigorously developing marine weapons and equipment,and the research on unmanned weapons and equipment has received extensive attention.The intelligentization of unmanned surface vessels is a research hotspot.To meet the detection and high-precision positioning requirements of large and medium-sized targets,this paper focuses on the design of target identification technique for unmanned surface vessel based on deep learning.Firstly,the multi-source and multi-system collaborative sensing architecture design is used to solve the problems of equipment intelligent computing task duplication and resource waste as well as deep learning acceleration.Secondly,multi-level feature extraction,analysis and fusion technique is designed to determine the features that should be selected for single/multi-sensors.Finally,the selected features are used to design multi-feature target detection and identification methods based on deep learning,and a multi-source multi-dimensional joint detection and identification processing method based on deep learning networks is established.The experimental results show that the recognition rate exceeds 99.7%for visual images,indicating that this technique has good recognition effects.
作者
王亮
陈建华
李烨
WANG Liang;CHEN Jianhua;LI Ye(Unit 91054 of PLA,Beijing 102442,China)
出处
《兵工学报》
EI
CAS
CSCD
北大核心
2022年第S02期13-19,共7页
Acta Armamentarii
关键词
无人艇
目标识别
特征提取
深度学习
unmanned surface vessel
target identification
feature extraction
deep learning
