In view of the high false alarm rate problem of the existed harbor detection method based on two-region segmentation and single-direction jetties scanning in synthetic aperture radar (SAR),a harbor detection method based on multi-region segmentation and multi-direction jetties scanning is proposed.The land covers are segmented to water region,strong backscattering man-made building and other land region using the multi-re-gion level set segmentation in polarimetric SAR based on the statistic characteristics of the coherency matrix at first.Then the regions of interest (ROIs)of harbor along the coastline are extracted from the sea-land segmen-tation result using the proposed multi-direction jetties scanning method.Finally,man-made harbors and some false naturally formed rugged terrains in the ROIs are distinguished according to the segmentation result of strong scattering buildings.Experimental results show that using polarimetric data acquired by RADARSAT-2 over some areas of Singapore and Zhangjiang of China to test the proposed method,man-made harbors along the coast are almost all correctly detected.%针对现有合成孔径雷达（synthetic aperture radar，SAR）图像中基于海陆两区域分割和单方向突堤扫描港口检测方法存在虚警率高的问题，提出一种基于多区域分割和多方向突堤扫描的人工港口检测方法。该方法首先使用基于极化 SAR 相干矩阵统计特性的多区域水平集分割方法将地物分割为水域、强散射建筑和其他陆地3类；然后在水陆分割结果上从多对正交方向对沿岸突堤进行扫描，从而确定出港口感兴趣区域；最后利用强散射建筑分割结果实现感兴趣区域中人工港口和虚警天然凸起地形的区分。基于 RADARSAT-2对新加坡和湛江地区实际观测的极化 SAR 数据进行实验，结果表明本文方法较好地实现了沿岸人工港口的正确检测。
展开▼
