摘要
为探究深度卷积神经网络在舰船检测与识别中的应用,研究了基于深度学习方法的可见光图像舰船目标检测与识别,总结了适用的可见光图像舰船数据集与针对舰船目标的网络优化方法。研究表明,迁移学习、先验框改进、特征优化等方法均能提升舰船检测与识别的准确率。未来应结合多源特征的融合,对轻量化舰船识别、细粒度舰船分类等方向进行研究。
To explore the application of deep convolutional neural networks in ship detection and recognition,this paper studies the ship detection and recognition in visible images based on deep learning,summarizes applicable ship datasets of visible images and optimized network method for ship target.The research shows that methods such as transfer learning,anchor improvement,feature optimization can all improve the precision of ship detection and recognition.Further research should integrate with the method of multi-source characteristics fusion,focus on light-weight ship recognition and fine-grained ship classification,etc.
作者
岳瞳
杨宇
YUE Tong;YANG Yu(Engineering University of PAP,Xi’an 710016,China)
出处
《舰船电子对抗》
2021年第2期77-82,95,共7页
Shipboard Electronic Countermeasure
基金
武警工程大学基础研究基金,项目编号:WJY201906
武警部队军事理论研究计划,项目编号:WJJY19-134
装备军内科研项目,项目编号:WJ20182A620020-2。
关键词
深度学习
先验框改进
特征优化
舰船检测与识别
可见光图像舰船数据集
deep learning
anchor improvement
feature optimization
ship detection and recognition
ship datasets of visible images
