基于知识的洞庭湖湿地遥感分类方法

ON THE REMOTE SENSING CLASSIFICATION METHOD OF DONGTING LAKE WETLAND BASED ON KNOWLEDGE

湿地遥感影像分类是遥感研究的一大难题。分析洞庭湖不同湿地类型在遥感影像上的光谱曲线规律,利用两个季节的洞庭湖ETM数据,并辅助以物候特征和地面GIS信息,通过遥感软件Erdas Image的专家分类知识库建立决策树分类方法,结合研究区的DEM进行洞庭湖湿地的影像分类.通过专家分类器分层次实现了包括水体、泥沙滩地、防护林滩地、湖草、芦苇滩地和苔草滩地以及其他水体7种湿地类型的分类。相比传统分类方法,专家分类过程以规则为基础,可以同时利用多个条件进行分类,减少了数据处理时间,同时还提高了分类精度,最终得到试验区较为可靠的遥感分类图像。

Remote sensing data based land cover classification in wetland areas is a difficult problem in the research area of remote sensing. In this paper, the rule of hyperspectral curves of different types of wetland in Dongtinghu Lake were analyzed, a model of decision tree for the final classification was constructed by the Erdas Image expert classification,and vegetation classification supported by GIS data and phonological information was made by using 2-seasonal ETM data of Dongtinghu Lake wetland. By the use of expert classification system provided by Erdas Image software,the image was classified into water area, mudflat, protection forest beach, Carem spp beach,Phragmites beach, Carex beach and other water body according to layered algorithm. The expert classification based on rules can extract wetland types at one time using multi-conditions, save the time of data processing and improve the precision of classification compared with other ways. The final classification result by the proposed method is reliable.