高分辨率遥感图像最近邻模糊分类器的研究及实现  

Study and implement of nearest neighbor fuzzy classifier for high-resolution remote sensing images

作　　者：朱国龙[1] 汪云甲[1] 乔浩然[1] 刘乐乐[1] 赵一清[1] 

机构地区：[1]中国矿业大学环境与测绘学院/江苏省资源环境信息工程重点实验室,江苏徐州221008

出　　处：《测绘科学》2010年第6期96-98,共3页Science of Surveying and Mapping

基　　金：国家自然科学基金资助项目(50774080);国家环保公益性行业科研专项(200809128);西部测图项目

摘　　要：本文针对高分辨率遥感影像快速高效萃取有用信息这一遥感技术应用的热点问题,探讨了一种适合于组合特征识别的遥感图像最近邻模糊分类器。该分类器首先把待识别目标的组合特征与训练模板中的组合特征样本的平均值一一进行比较,从而得到了一个特征差矩阵。用模糊分布函数在同类特征差之间进行处理,生成一个隶属度矩阵,然后用算术平均法对隶属度矩阵进行处理,并用最大隶属度准则来进行分类判决。以新疆和静县的SPOT5图像为例,应用此方法对其进行分类试验。结果表明:利用此分类方法对SPOT5遥感图像进行分类,不仅使分类结果具有丰富的语义信息,而且克服了由于特征选择的不稳定性对分类结果的影响,分类精度也得到了显著的提高。A nearest neighbor fuzzy classifier（NNFC） using combined feature was presented in the paper according to the hotspot of application of remote sensing technology that how to extract the valuable information from high-resolution remote sensing image.Firstly,the combined feature of the unknown target was compared with the mean of the samples of the training space in the NNFC,and a feature difference matrix was obtained.Then,fuzzy membership function was used to process the feature difference matrix and a membership degree matrix was obtained.The membership degree matrix was processed by averaging method,and the maximal membership degree rule was used to determine the classification of the target.Taking Hejing of Xinjiang province as the test area,a case study on SPOT5 image classification with NNFC was carried out.The case study showed that the application of NNFC on SPOT5 image classification can not only have more semantic information,but also overcome the impact of instability of feature selection to classification,and the overall classification accuracy could be improved as well.

关 键 词：高分辨率遥感 最近邻模糊分类器 组合特征 模糊分布函数 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]