摘要： 提出一种SAR图像水体自动检测方法。该方法采用树型小波对SAR图像进行分解，提取样本图像与待检测图像的中频信息，并进行纹理分析，利用模糊加权支持向量机对样本图像的纹理进行训练，从而获得判别水体的决策函数，通过该决策函数能够检测出图像的水体区域。该方法结合了图像的灰度与纹理信息，减少了SAR图像中的噪声影响，能够适用于大幅面范围的SAR图像水体自动检测。

关键词: SAR图像, 水体检测, 纹理分析, 树型小波, 模糊加权支持向量机

Abstract: A method of water automatic detection for SAR images is presented, which uses tree wavelet to decompose the SAR images, obtains the medium-frequency imformation from the sample image and the test image, and analyzes the texture. The texture information of the sample image is trained by using Fuzzy Weighted Support Vector Machine(FW-SVM), and the water decision function is got, through which the water region can be detected. This method combines the gray and texture information of the SAR image, reduces the noise effect, and is suitable for the water automatic detection of large area SAR image.

Key words: SAR image, water detection, texture analysis, tree wavelet, Fuzzy Weighted Support Vector Machine(FW-SVM)

中图分类号: 

• TP391.72