Dual‑domain Joint Dense Multiple Small Ship Target Detection Algorithm for Spaceborne SAR Images  

面向星载SAR图像的双域联合密集多小舰目标检测算法

作　　者：JIA Peng DONG Tiancheng WANG Taoyang ZHANG Guo SHENG Qinghong LI Jun 贾鹏;董天成;汪韬阳;张过;盛庆红;李俊(南京航空航天大学航天学院,南京211106;武汉大学测绘遥感信息工程国家重点实验室,武汉430079;武汉大学遥感与信息工程学院,武汉430079)

机构地区：[1]College of Astronautics,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,P.R.China [2]State Key Laboratory of Information Engineering in Surveying,Mapping and Remote Sensing,Wuhan University,Wuhan 430079,P.R.China [3]School of Remote Sensing and Information Engineering,Wuhan University,Wuhan 430079,P.R.China

出　　处：《Transactions of Nanjing University of Aeronautics and Astronautics》2024年第6期725-738,共14页南京航空航天大学学报(英文版)

基　　金：supported by the Foundation Strengthening Fund Project(No.2021-JCJQ-JJ0251);in part by the National Natural Science Foundation of China(Nos.42301384 and 42271448)。

摘　　要：Ship detection via spaceborne synthetic aperture radar(SAR)has become a research hotspot.However,existing small ship detection methods based on the radar signal domain and SAR image features cannot obtain highly accurate results because of the obvious coherent speckle noise at sea and strong reflection interference from near‑shore objects.To resolve the above problems,this study proposes a dual‑domain joint dense multiple small ship target detection method for spaceborne SAR image that simultaneously detects objects in the image and frequency domains.This method uses an attention mechanism module and algorithm structure adjustments to improve the small ship target feature mining ability.In the frequency‑based image generation,extreme signal strength values are detected in the azimuth and range directions,with the results of the two complementing each other to realize dual‑domain joint small ship target detection.The comprehensive qualitative and quantitative evaluation results show that the proposed method can attain a final precision rate of 92.25%and achieve accurate results for SAR ship detection in open‑sea,coastal,and port area ships.The test results for the self‑built SAR small‑ship dataset demonstrate the effectiveness and universality of the method.通过星载合成孔径雷达(Synthetic aperture radar,SAR)进行舰船探测已成为研究热点,但由于海上相干斑点噪声明显,近岸物体反射干扰强等问题,现有的基于雷达信号域和SAR图像特征的小型舰船探测方法无法获得高精度的结果。为解决上述问题,提出了一种针对空间SAR图像的双域联合密集多重小型船舶目标检测方法,可同时在图像域和频域检测目标。该方法利用注意力机制模块和算法结构调整来提高小船目标特征挖掘能力。在基于频率的图像生成中,检测方位角和测距方向的极端信号强度值,二者结果互为补充,实现双域联合小型舰船目标检测。定性和定量的综合评价结果表明,所提出的方法最终精确率可达92.25%,在开阔海域、沿海和港区船舶的SAR船舶探测方面取得了准确的结果。自建SAR小型船舶数据集的测试结果证明了该方法的有效性和通用性。

关 键 词：synthetic aperture radar(SAR) small ship detection deep learning attention module YOLO dual‑domain joint 

分 类 号：V19[航空宇航科学与技术—人机与环境工程]