一种改进的蚁群算法在高光谱遥感图象分类中的应用  

An improved ant colony algorithm used in Hyperspectral remote sensing image classification

作　　者：孙可[1] 殷守林 刘杰[1] 

机构地区：[1]沈阳师范大学科信软件学院,沈阳110034

出　　处：《沈阳师范大学学报（自然科学版）》2016年第3期354-358,共5页Journal of Shenyang Normal University:Natural Science Edition

基　　金：国家自然科学基金资助项目(60970112)

摘　　要：针对原始蚁群算法在高光谱遥感图象分类中收敛速度慢,结果不精确的缺点,提出了一种改进的蚁群算法,并把基于改进蚁群算法的特征选择应用在高光谱遥感图象分类中,从而建立一种新的高光谱遥感图象分类模型。模型分为3部分:基于传统蚁群算法原理,提出新的蚁群算法信息素更新方法;使用改进后的蚁群算法,令蚂蚁在平面上随机选择一个随机投影到平面上的特征,在所有特征中,使用判别函数来确定哪条路径最优,随后形成特征组合;根据特征组合,使用极大似然分类器对遥感图象进行分类。实验结果表明,基于改进的蚁群算法更能有效的对高光谱遥感图象进行分类。When original ant colony algorithm is used in Hyperspectral remote sensing image classification, which has slow convergence speed and results inaccuracy. The author proposes an improved ant colony algorithm in this paper. And the feature selection based on new scheme is applied into Hyperspectral remote sensing image classification, so it builds a new Hyperspectral remote sensing image classification model. This model includes three parts： firstly, it presents new pheromone updating methods based on traditional ant colony algorithm; secondly, each ant stochastically selects a feature on the plane（features are randomly projected on a plane） based on new ant colony algorithm. In all features, it uses discrimination function to determine which road is the best; thirdly, according to the characteristics of combination, it uses the maximum likelihood classifier to classify remote sensing image. Finally, experimental results show that the new scheme based on improved ant colony algorithm is more effective to deal with Hyperspectral remote sensing image classification.

关 键 词：蚁群算法 特征选择 高光谱遥感图象分类 信息素 极大似然分类器 特征组合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]