基于弹跳射线法的海上目标快速成像与识别算法

合成孔径雷达(Synthetic Aperture Radar,SAR)因为全天候、全天时的观测能力在军事和民用领域有着广泛的应用背景。考虑到SAR研究的成本和效率以及SAR图像在目标检测领域的应用,SAR图像的仿真技术发挥了重要优势。针对传统成像方法耗时较长的问题,本文利用基于弹跳射线法(SBR)的快速成像技术以达到快速获取大批量SAR图像的目的。为了更加精准地识别SAR图像中的目标,在Faster RCNN目标检测网络的基础上,根据真实舰船目标的尺寸改变候选框的初始尺寸以及利用特征融合的方式对原算法架构加以改进。最后,在Faster RCNN框架中加入特征金字塔结构(Feature pyramid networks,FPN),进一步提高算法对SAR图像中的舰船目标检测和识别的能力。

Research on Recognition of Maritime Targets Based on the SBR Imaging Algorithm

Synthetic Aperture Radar (SAR) has a wide range of applications in military and civilian sectors with all-weather, all-day observation capabilities.Considering the cost and efficiency of SAR research and the application of SAR image in the field of target detection, SAR image simulation technology plays an important role.In view of the time-consuming problem of traditional imaging methods, this paper uses the rapid imaging technology based on SBR to obtain SAR images in large quantities quickly. The data source generated by this method is more complete than the measured data set.In order to identify targets in SAR images more accurately, the original algorithm framework was improved by changing the initial size of the candidate box according to real ship targets and using feature fusion based on the target detection network of Faster RCNN.Finally, Feature pyramid networks (FPN) is added to Faster RCNN framework to further improve the ability of target recognition algorithm to detect and recognize ship targets in SAR image.