郭智^{1, 2},
宋萍^{1, 2, 3,,},
张义^{1, 2},
闫梦龙^{1, 2},
孙显^{1, 2},
孙皓^{1, 2}
1.中国科学院电子学研究所 ??北京 ??100190
2.中国科学院空间信息处理与应用系统技术重点实验室 ??北京 ??100190
3.中国科学院大学 ??北京 ??100049
基金项目:国家自然科学基金(41501485)

详细信息

作者简介:郭智：男，1975年生，研究员，研究方向为地理空间信息综合处理与应用
宋萍：女，1991年生，硕士生，研究方向为机器学习与遥感图像智能解译
张义：男，1987年生，助理研究员，研究方向为阵列信号处理
闫梦龙：男，1985年生，副研究员，研究方向为机器学习与遥感图像智能解译
孙显：男，1981年生，副研究员，研究方向为机器学习与遥感图像智能解译
孙皓：男，1984年生，副研究员，研究方向为机器学习与遥感图像智能解译

通讯作者:宋萍　 pingsong2014@163.com

中图分类号:TP753

计量

文章访问数:2660
HTML全文浏览量:1034
PDF下载量:160
被引次数:0

出版历程

收稿日期:2018-01-26
修回日期:2018-06-06
网络出版日期:2018-08-30
刊出日期:2018-11-01

Aircraft Detection Method Based on Deep Convolutional Neural Network for Remote Sensing Images

Zhi GUO^{1, 2},
Ping SONG^{1, 2, 3,,},
Yi ZHANG^{1, 2},
Menglong YAN^{1, 2},
Xian SUN^{1, 2},
Hao SUN^{1, 2}
1. Institute of Electronics, Chinese Academy of Sciences, Beijing 100190, China
2. Key Laboratory of Technology in Geo-spatial Information Processing and Application System, Chinese Academy of Sciences, Beijing 100190, China
3. University of Chinese Academy of Sciences, Beijing 100049, China
Funds:The National Natural Science Foundation of China (41501485)


摘要
摘要:飞机检测是遥感图像分析领域的研究热点，现有检测方法的检测流程分为多步，难以进行整体优化，并且对于飞机密集区域或背景复杂区域的检测精度较低。针对以上问题，该文提出一种端到端的检测方法MDSSD来提高检测精度。该方法基于单一网络目标多尺度检测框架(SSD)，以一个密集连接卷积网络(DenseNet)作为基础网络提取特征，后面连接一个由多个卷积层构成的子网络对目标进行检测和定位。该方法融合了多层次特征信息，同时设计了一系列不同长宽比的候选框，以实现不同尺度飞机的检测。该文的检测方法完全摒弃了候选框提取阶段，将所有检测流程整合在一个网络中，更加简洁有效。实验结果表明，在多种复杂场景的遥感图像中，该方法能够达到较高的检测精度。
关键词:遥感图像处理/
飞机检测/
密集连接卷积网络
Abstract:Aircraft detection is a hot issue in the field of remote sensing image analysis. There exist many problems in current detection methods, such as complex detection procedure, low accuracy in complex background and dense aircraft area. To solve these problems, an end-to-end aircraft detection method named MDSSD is proposed in this paper. Based on Single Shot multibox Detector (SSD), a Densely connected convolutional Network (DenseNet) is used as the base network to extract features for its powerful ability in feature extraction, then an extra sub-network consisting of several feature layers is appended to detect and locate aircrafts. In order to locate aircrafts of various scales more accurately, a series of aspect ratios of default boxes are set to better match aircraft shapes and combine predictions deduced from feature maps of different layers. The method is more brief and efficient than methods that require object proposals, because it eliminates proposal generation completely and encapsulates all computation in a single network. Experiments demonstrate that this approach achieves better performance in many complex scenes.
Key words:Remote sensing image processing/
Aircraft detection/
Densely connected convolutional network



PDF全文下载地址:

https://jeit.ac.cn/article/exportPdf?id=80a9d416-0528-46e8-b61c-52672547031b