一种针对复杂背景下高分辨率SAR图像河道检测算法

A River Extraction Algorithm for High-resolution SAR Images with Complex Backgrounds

针对复杂背景下高分辨率SAR图像中河道轮廓提取问题,在分析前人已有成果的基础上,提出一种基于分块直方图及区域特征的河道信息检测方法。首先通过小波变换技术在保持图像边缘特性的同时对图像去噪处理,进而对图像分块后利用统计直方图初步确定河道标记点位置。在此基础上采用基于标记点的分水岭变换进行初始分割,最后利用区域邻接图(RAG)的区域合并策略得到河道检测结果。实验结果表明:与戴光照等提出的采用直方图阈值快速分割提取算法相比,该算法在完整提取河道轮廓的同时显著提高了提取精度,同时可进一步用于河道桥梁提取,具有良好的可用性与有效性。

This paper proposes a river detection method based on sub block histogram and regional charac- teristics for the river extraction in high-resolution SAR images with complex background. Firstly, using wavelet transform reduces speckle and at the same time keeps the edge features of the image. Then utilizing statistics histogram initially fix the markers of river on the sub-block image. On that basis, it executes the marker-based watershed transformation to acquire the initial segmentation of the river. Finally it gets the detection results by the regional adjacency graph （RAG）. Compared with the algorithm proposed by Guan- Zhao Dai,etc. which used the statistics from histogram for fast segmentation the method in this paper that can quickly determine the position of the river and extract the river body with higher extraction accuracy. This method also can be further used for bridge detection. The experiments show that the good usability method can effectively extract the outlines of river in high-resolution SAR images with complex back grounds.