摘要
该文提出了一种在SAR图像中有效地得到连通河流的检测方法。算法基于二进小波变换的边缘提取和脊线跟踪技术。该文应用小波变换多尺度特性进行边缘检测,采用小波域系数直接相关的方法提取边缘。该方法比WTMM获取边缘点的算法更加简单高效。在解决连接不连续边缘点和水域连通问题时,该文采用了指纹提取中的基于灰度级图像的脊线跟踪技术,从而避免了已广泛应用于海岸线连接中的Snake模型的一些缺点。结果证明该文算法更高效,对复杂的河流边缘适应性更强,算法得到的河流检测结果连通性更好,定位准确。
This paper deals with automatic river detection in SAR images. A new method is proposed using the boundary extraction in wavelet domain and merging the water region by ridge tracing. In the method, edge detection is approached by direct spatial correlation of wavelet transform at several adjacent scales. It is much easier than the algorithm of WTMM proposed by Mallat and proved to he more efficient. And the ridge tracking based on gray scale image , which is applied in fingerprint tracing, is introduced here to complete the river hank linking and river connecting. This improvement avoids the disadvantages of the Snake model widely used in linking the coastline. The river detection result from the method is good at connection and location. And the method is much more efficient and adaptive to the complicated bank.
出处
《电子与信息学报》
EI
CSCD
北大核心
2007年第3期524-527,共4页
Journal of Electronics & Information Technology
关键词
SAR图像
河流检测
边缘提取
脊线跟踪
连通
SAR image
River detection
Boundary extraction
Ridge tracing
Connection
作者简介
胡正磊：女，1982年生，硕士生，研究方向为合成孔径雷达图像处理．
孙进平：男，1975年生，副教授，主要研究领域为高分辨率雷达成像及图像处理．
袁运能：男，1962年生，副教授，主要研究领域为小波分析及其应用、高分辨率雷达成像．
毛士艺：男，1935年生，教授，主要研究领域为高分辨率雷达成像、信号建模与谱分析、信号检测与分类、多目标跟踪和数据融合等．
