基于改进的粒子群和蚁群算法的高光谱森林聚类研究  

The Research of Hyperspectral Image Clustering for Forest Type Based on Ant Colony Algorithm and Improved PSO

作　　者：李琰[1] 邢艳秋[2] 王立海[2] 

机构地区：[1]东北林业大学信息与计算机工程学院,黑龙江哈尔滨150040 [2]东北林业大学森林作业环境研究中心,黑龙江哈尔滨150040

出　　处：《安徽农业科学》2014年第27期9615-9618,共4页Journal of Anhui Agricultural Sciences

基　　金：国家自然科学基金面上项目(41171274);国家林业"948"项目(2014-4-78)

摘　　要：高光谱图像分类可分为监督分类与非监督分类,聚类分析进行非监督分类是一种现今比较受研究者广泛关注的技术。粒子群算法具有自适应、自组织性、可同时进行局部和全局搜索等特点;蚁群算法通过智能个体间不断进行信息交流和传递,具有较强的发现最优解的能力。提出一种基于改进的粒子群和蚁群算法的高光谱图像聚类方法,设计其模型并将其应用在森林类型分类问题上,提高分类精度,减少人工干预。以吉林省汪清林业局为研究区,通过修改粒子群的惯性系数,得出最优解集,然后利用蚁群寻优的过程对阔叶林、针叶林、混交林、水体进行聚类分析,区分精度达到85%证明,该方法能较好地识别森林类型。Hyperspectral image classification can be divided into supervised classification and unsupervised classification.Unsupervised classification was conducted on cluster analysis,which is a relatively modern technology by researchers attention.PSO algorithm with adaptive,self-organization,can be used for local and global search simultaneously.Ant colony algorithm has a strong ability to find the optimal solution through continuous exchanging the information and transmission between intelligent individual.This paper presents an improved method for hyperspectral image clustering based on improved PSO and ant colony algorithm,designs the model and applies it in the forest type classification problems for improving classification accuracy,reducing manual intervention.Taking Wangqing Forestry Bureau in Jilin Province as the study area,by modifying the inertia coefficient of PSO,the optimal solution set was obtained.Using optimization process of the ant colony,clustering for forest types can distinguish better between broad-leaved forest,coniferous forest,mixed forest and water bodies.The classification accuracy is 85％.The method can identify forest types.

关 键 词：粒子群优化算法 蚁群算法 遥感图像 高光谱 聚类 森林类型 

分 类 号：S126[农业科学—农业基础科学]