基于二维模糊熵的GA在SAR图像分割中的应用  

Application of Genetic Algorithm Based on Two-dimensional Fuzzy Entropy in SAR Image Segmentation

作　　者：张红蕾[1] 宋建社[1] 张宪伟[2] 

机构地区：[1]第二炮兵工程学院信息工程研究所,西安710025 [2]二炮装备研究院信息所,北京100085

出　　处：《计算机工程》2007年第5期158-160,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60272022)

摘　　要：SAR图像中目标和背景的分割是SAR图像分割中的重要内容,阈值方法是其中比较简单实用的方法。阈值分割的核心问题是如何选择合适的阈值,最简单和常用的方法是从图像的灰度直方图出发,得到各个灰度级的概率分布密度,依据某一准则选取一个或多个合适的阈值,以确定每个像素点的归属。该文借助遗传算法工具,对基于二维直方图的模糊熵法做了改进和设计,提出了寻找最优阈值的分割算法,经MSTAR数据测试,对于含噪SAR图像目标和背景的分割具有很好的效果,抑噪功能强。The segmentation of target and background is an important content of SAR image segmentation. Threshold method is a kind of simple and practical method. The key problem of threshold segmentation is how to choose appropriate threshold value, and the ordinary method is utilizing grey histogram to get probability distribution density of each grey level, and then choose one or more suitable threshold value according to a certain rule to define adscription of each image pixel. This paper improves the fuzzy entropy method based on two-dimensional histogram, and puts forwards a new segmentation algorithm to seek optimum threshold value in the presence of genetic algorithm. The experimentation shows that it is an effective method for speckled SAR image to segment object and background.

关 键 词：合成孔径雷达 遗传算法 模糊熵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]