山区植被类型信息提取方法研究

根据遥感图像的光谱信息和空间信息特征及不同植被的分布规律,研究利用计算机处理技术提取山区植被类型的方法。分类过程采用四个步骤完成：①均一目标的象限四分树提取分类；②多光谱数据的最小距离分类；③综合利用波谱曲线的形态和地形数据进行分类；④高程数据修正分类。在分类处理过程中,分别利用了图像的空间信息、光谱信息以及地形数据。利用该分类方法在实验小区内进行植被类型提取试验,其精度为90%.与最大似然分类方法所得结果相比较,其分类精度提高了10%.

THE METHOD OF VEGETABLE CLASSIFICATION OF THEMATIC MAPPER IN MOUNTAIN AREA OF SOUTH-WEST CHINA

The major objective of this paper is to introduce a new method of classification to remote-sensing data, it is four-stages classifier, the classifier described here is based on four stages of operation:1) Quadtree segmentation and homogeneity;2) Minimum distance to means;3) Ancillary DTM data and spectral curves;4) Elevation data was used to correct the classification;The characteristic of this method is incorporates the information of spectral, spatial and ancillary DTM data in classication, in order to overcome the weakness in ordinary classification pixels by pixels. This method was used in the train area, the accuracy is 90%, the overall improvement in accuracy is 10% to compared to per-pixel maximum likelihood classifiers.