面向滇中高原湖泊区地表覆盖遥感探测的分类精度对比研究

Comparison of Accuracy of Land Cover Classification by Remote Sensing Monitoring Methods in Yunnan Plataea Lake Area

遥感图像分类是遥感信息提取的重要技术手段,目前已有多种遥感图像分类方法运用在地表覆被分类中,而分类精度的提升是遥感数据处理及其应用的核心所在。因此,针对具体研究区环境特性,并通过比较分析各种遥感图像分类方法的精度状况,成为了如何选择最合适遥感应用方法的根本依据。本文选取具有云南省滇中高原湖泊区域典型代表性的星云湖及周边湖滨带径流区域为研究试验区,通过采用监督分类、BP神经网络分类和面向对象分类等三种遥感分类方法,开展研究区地表覆被类型的遥感图像信息提取,并对三种方法所获取结果的精度状况进行对比分析研究。结果显示,面向对象方法相对其余两种方法更适用于星云湖流域土地利用分类。研究结果对于推进面向云南滇中高原湖滨带地表覆盖信息的遥感快速提取与动态监测具有现实科技指导作用,同时对研究云南滇中高原湖泊区域的资源环境利用与保护管理等具有重要的科学意义。

Remote sensing classification is an important technology used for the extraction of information, there are varieties of remote sensing methods used in the classification, accuracy is the represent of the effect of classification which makes the improvement of accuracy become the core of remote sensing data processing and applications. Therefore, the analyst of accuracy by comparing the different classification methods in the specific study area has become the fundamental basis of remote sensing application. The typical representative area of Central Yunnan Plateau lakes region, Xing Yun lake and the runoff area were selected as study area in this paper. Three methods of unsupervised classification, BP neural network classification and object-oriented classification were used for the extraction of land cover information in remote sensing image, and comparing the three classification methods by accuracy evaluation. The results show that the object-oriented method is more suitable for land use classification in the Xingyun Lake Basin than the other two methods. There is practical guidance for the results in promoting quickly extraction and dynamic monitoring of land cover information, and important scientific significance for the study of resource and environment management, utilization and protection in Central Yunnan plateau lake.