基于双阶段特征匹配的非同源SAR船只跟踪方法

Non-Homologous SAR Ship Matching and Tracking Method Based on Two-Stage Feature Matching

开展星载合成孔径雷达(Synthetic Aperture Radar,SAR)船只匹配跟踪研究是海上船只监测的重要内容,对提升海上目标监测管控能力有重要意义。当前利用SAR卫星进行船只目标匹配跟踪时,由于船只目标尺寸小且船只目标的运动会在SAR图像中产生几何畸变,导致船只目标在SAR图像中难以准确实现匹配跟踪。基于此,本文提出了基于全局和局部双阶段特征匹配的SAR船只目标匹配跟踪方法,该方法能够实现TerraSAR-X和RadarSat-2等不同波段、不同平台(非同源)SAR图像的船只目标匹配跟踪。为验证本文算法的性能,采用了2对TerraSAR-X和RadarSat-2 SAR数据将本文提出的方法与SURF(Speeded Up Robust Features)、Harris和PCA-SIFT(Principal Component Analysis-Scale Invariant Feature Transform)等经典船只目标匹配跟踪算法进行了对比。实验结果表明,本文提出的基于全局和局部双阶段特征匹配船只目标匹配跟踪方法,对运动船只的跟踪性能可达83%,而SURF、Harris和PCA-SIFT算法精度仅为36.2%、27.7%和23.4%,可见本文算法高于这些经典船只目标匹配跟踪算法。本文还分析了船只目标的运动在非同源SAR图像中的几何特征(周长、面积、长宽比等)差异,以及船只目标匹配跟踪算法在非同源SAR中的适用性评估,结果表明,本文方法在船只几何特征形变较大时,也能很好地实现船只目标匹配跟踪。

Study on satellite-borne SAR ship matching and tracking is an important part of maritime ship monitoring and has great significance for improving the ability of maritime target monitoring and control.However,when SAR satellite is used for the ship target matching and tracking,geometric distortions may possibly be generated in the SAR images due to the small size and movement of the ship target,which makes it difficult to accurately match and track the ship target in the SAR images.For this reason,a method for SAR ship target matching and tracking based on global and local two-stage feature matching is proposed.This method can make it possible to realize the ship target matching and tracking on the SAR images of different bands such as TerraSAR-X and RadrarSat-2 and different platforms(non-homologous).In order to verify the performance of the algorithm proposed in the paper,two pairs of TerraSAR-X and RadarSat-2 SAR data are used for comparing the method proposed in the paper with the classic algorithms such as SURF(Speeded Up Robust Features),Harris and PCA-SIFT(Principal Component Analysis-Scale Invariant Feature Transform).The results show that the method proposed in the paper can make the tracking performance for moving ships reach to 83%,while by using SURF,Harris and PCA-SIFT algorithms the accuracy are only 36.2%,27.7%and 23.4%,respectively.It can be seen that the algorithm of this study is much better than those classic algorithms.In addition,the differences in geometric features(e.g.perimeter,area,aspect ratio,etc.)of the ship target motion in the non-homologous SAR images are analyzed and the applicability of the ship target matching and tracking algorithm in the non-homogeneous SAR images is evaluated.The results indicate that by using the method of this study,the ship target matching and tracking can also be achieve better,even when the geometric characteristics of the ship undergo a larger deformation.