基于面向对象及光谱特征的植被信息提取与分析

Abstract

面向对象分类方法是实现精确详细提取遥感信息的新方法。以2010年TM影像作为基础数据,采用面向对象方法着重提取了西辽河流域平原的植被信息(耕地、林地、中生偏旱和中生偏湿草)。通过对不同地物的光谱和空间信息(位置、形状)进行比较分析,建立适当的隶属度函数和阈值范围,在草地的分类规则中,参考野外采集的草地植被光谱信息,辅助草地属性特征选择。最后构建分类决策树,完成自动分类,分类总体精度达到82.13%。在分类结果的基础上,讨论了植被分布特点,对不同类型的植被面积、不同区县植被面积以及植被与一级河流关系三方面进行分析。结果显示:研究区植被以林地和耕地为主,分别占总面积的38.9%和23.3%;耕地、林地、中生偏湿草地多分布于河流中下游区县,受人为因素影响较大的林地、耕地主要集中在科尔沁左翼中旗和科尔沁左翼后旗;林地、中生偏湿草较集中分布在一级河流的10km缓冲区内,耕地主要集中在5km缓冲区内,中生偏旱草与一级河流的关系不显著。;Object-oriented technique is a new method providing more accurate mapping product with higher detail. In this paper, TM images of 2010 were selected as the data sources. Based on object-oriented method, six categories information of west Liaohe river basin were extracted, including cultivated land, forest land, grassland (xero-mesophytes and hygro-mesophytes), water and unutilized land. Comparing and analyzing the spectral and spatial information among the categories, appropriate threshold were set in membership function. Meanwhile, the field spectral characteristics analysis was utilized to help in determining the appropriate grassland classification rules. Then combining classification tree, the automatic classification was accomplished and the general accuracy was 82.13%. Moreover, we discussed the characteristic of vegetation distribution in study area. Using spatial analysis of GIS to statistic different vegetation area, different vegetation area in different counties and relationship between vegetation and first class river, the results showed: (1) forest land and grassland were predominant vegetation in study area, which area accounted for 38.9% and 23.3% respectively. (2) Forest land, cultivated land and hygro-mesopgytes were mainly distributed in counties which located in middle and lower reaches of the river. Influenced by human impact, cultivated land and forest land concentrated distribution in Horqin Left Middle Banner and Horqin Left Back Banner. (3) Forest land and hygro-mesophytes were mainly distributed in 10 km buffer zone of the first class river, cultivated land concentrated distribution in 5km buffer zone of the first class river.