星载合成孔径雷达图像的飞机目标检测  

Aircraft Target Detection from Spaceborne Synthetic Aperture Radar Image

作　　者：郭倩[1] 王海鹏[1] 徐丰[1] GUO Qian;WANG Haipeng;XU Feng(Key Laboratory for Information Science of Electromagnetic Waves,Fudan University,Shanghai 200433,China)

机构地区：[1]复旦大学电磁波信息科学教育部重点实验室,上海200433

出　　处：《上海航天》2018年第6期57-64,共8页Aerospace Shanghai

基　　金：国家自然科学基金(61571132);上海航天科技创新基金(SAST2017078;SAST2016061)

摘　　要：针对大场景下星载合成孔径雷达(SAR)图像中飞机目标检测问题,提出一种端到端的飞机目标检测算法。先在大场景SAR图像中对机场目标进行粗检测,定位机场区域,再通过精确分割算法获得机场的精细区域。对机场区域中的飞机目标进行检测,采用一种基于Canny算子的边缘检测与卷积神经网络结合的飞机目标检测算法。通过飞机边缘检测、边界框预处理等操作确定潜在飞机目标在机场中的位置范围,采用基于GoogLeNet的卷积神经网络对可疑目标进行鉴别。利用星载合成孔径雷达数据对算法进行验证,证明该方法的有效性与实用性。Aiming at the aircraft target detection in big scenes from spaceborne synthetic aperture radar(SAR)image,an end-to-end algorithm is proposed in this paper.It firstly locates the airport area from the large SAR image via rough detection method,and then obtains an accurate mask of the airport by using accurate segmentation method.Within the airport area,aircraft targets are detected via a two-stage method.In the first stage,potential targets are identified based on Canny edge-detection,and in the second stage,true aircraft targets are discriminated via GoogLeNet convolutional neural network.The effectiveness and practicality of the method are verified by collected spaceborne SAR image data.

关 键 词：合成孔径雷达 替代滤波 飞机检测 边缘检测 卷积神经网络 

分 类 号：TN911.73[电子电信—通信与信息系统] TP391.9[电子电信—信息与通信工程]