遥感信息专题分类不确定性的可视化研究

摘要

Nowadays extract thematic categories of information from remote sensing data is one of the most important application areas. As the remote sensing classification of thematic information used in various fields, its data quality has paid more and more attention. Evaluation of classification uncertainty is the most important topics in class data quality. In this paper we mainly used Keriya county’s 2006 years remote sensing image as the main data sources, in ENVI software platform support of study area, shear, registration, interpretation, classification processing, it is concluded that the classification figure and basic space information. On this basis in MATLAB Programming computing the entropy of the image and analyzing. In ArcScene realize the visualization of uncertainty in thematic classification of remote sensing information. The results show that: water has smaller uncertainty, cause under natural environmental conditions, Regardless the depth of water, the angle of specular reflection, the reflection of the visible and near infrared bands are quite small, absorbed almost all the incident energy. The characteristics of remote sensing image recognition of water to bring convenience, no matter what a band, water body image are dark, compared with the surrounding surface features it has a tone contrast. In arid area, cotone is a transitional zone between desert and oasis, part of it plants has high similarity with Oasis, Easily be divided into oasis, the plan-tlees part has a high similarity with desert, easily be divided into the desert, therefore has low resolution, high uncertainty. Interlaced zoon has larger uncertainty, the large uncertainty distribution on the edge of category, and the smaller uncertainty distribution on the internal of category.%利用遥感技术（RS）从遥感影像资料中提取专题类别信息是当前遥感数据主要的应用领域之一。由于遥感分类专题信息广泛应用于各领域，其数据质量受到越来越多的关注。而不确定性是评价分类专题类别数据质量最主要的方面。鉴于此，本文以于田县2006年的遥感图像为例在ENVI软件平台支持下，对研究区进行剪切、配准、解译、分类处理，得出分类图和基础空间信息，在MATLAB中计算图像的熵值，进行不确定性分析；并用ArcScene来实现遥感信息专题分类不确定性空间分布规律的可视化。实验结果表明：水体和绿洲有较小的不确定性，交错带有较大的不确定性，大的不确定性分布在类别边缘区域，而小的不确定性主要在类别内部。