模糊ARTMAP神经网络在TM影像分类中的应用  

Application of Fuzzy ARTMAP Neural Network in TM Image Classification

作　　者：韩敏[1] 程磊[1] 唐晓亮[1] 李金祥[2] 

机构地区：[1]大连理工大学电子与信息工程学院自动化系,辽宁大连116023 [2]水利部松辽水利委员会,吉林长春130021

出　　处：《信息与控制》2004年第5期623-627,共5页Information and Control

基　　金：国家自然科学基金重点资助项目 ( 5 0 13 90 2 0 )

摘　　要：要 :阐述了模糊ARTMAP网络结构及其采用的算法 ,提出一种引入遥感图像判读结果的警戒系数自动调整算法 ,能够解决人为选择警戒参数效率低、难以取得合适数值的问题 .仿真结果表明 ,具有警戒系数调整功能的模糊ARTMAP神经网络能够有效地对向海自然保护区的TM影像进行分类 ,它与最大似然法和传统的模糊ARTMAP神经网络相比 ,对样本的依赖程度较低 ,分类精度较高 .The structure and algorithm of fuzzy ARTMAP neural network are described in detail, and an automatically adjusted algorithm of vigilance parameters is presented. The automatic adjustment has the advantages of inducting classification results into the selection of vigilance parameters and avoiding the problem of getting the optimum value artificially. The classification results show that the fuzzy ARTMAP neural network with the automatically adjusted algorithm can effectively classify the thematic mapper(TM) image which covers Xianghai Nature Reserve, and it depends less on samples and gets higher accuracy of the classification than maximum likelihood classifier and conventional fuzzy ARTMAP neural network.

关 键 词：遥感 TM影像 分类 模糊ARTMAP 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]