一种基于概率潜在语义模型的高分辨率遥感影像分类方法  

A Probabilistic Latent Semantic Analysis Based Classification for High Resolution Remotely Sensed Imagery

作　　者：陶超[1,2] 谭毅华[1,2] 彭碧发[1,2] 田金文[1,2] 

机构地区：[1]华中科技大学图像识别与人工智能研究所,湖北武汉430074 [2]华中科技大学多谱信息处理技术国防重点实验室,湖北武汉430074

出　　处：《测绘学报》2011年第2期156-162,共7页Acta Geodaetica et Cartographica Sinica

基　　金：国家863计划(2007AA12Z153)

摘　　要：针对高分辨率遥感影像中"同谱异物","同物异谱"现象对影像分类过程造成的干扰,将文本分析中的概率潜在语义模型应用于高分辨率遥感影像分类,提出一种无监督的遥感影像分类新方法。该方法首先利用均值漂移分割方法对影像进行分割构建图像区域集合,然后提取集合各区域中每个像元的Gabor纹理特征,并对这些特征进行聚类形成视觉词汇,最后利用PLSA方法对各区域进行分析,找出其最可能属于的主题或者类别,从而完成影像分类。文中GeoEye-1和IKONOS影像试验结果表明,该方法能有效提高高分辨率遥感影像分类精度。A new unsupervised classification algorithm for high spatial resolution remotely sensed imagery,which combines Gabor texture feature and PLSA model（probabilistic latent semantic analysis）,is presented.Firstly,homogeneous segments are extracted from original imagery through mean shift（MS） segmentation.Secondly,Gabor texture features of every pixel in each region are extracted,and clustered into several visual words.Thus,the imagery segments correspond to the documents,the visual words used to describe the segments correspond to the words in the documents,and the categories to be discovered for each segment correspond to the topics of the documents.Finally,PLSA model are used to analyze each segment,and achieve the image classification by assigning the most likely category for them.The experimental results of GeoEye-1 and IKONOS imagery have shown that the approach can outperform the existing algorithms in terms of classification accuracy.

关 键 词：高分辨率遥感影像分类 概率潜在语义模型 视觉词汇 GABOR纹理特征 

分 类 号：P237[天文地球—摄影测量与遥感]