利用遥感信息决策树方法分层提取荒漠化土地类型的研究探讨

Primary Study on the Multi-Layer Remote Sensing Information Extraction of Desertification Land Types by Using Decision Tree Technology

选择甘肃省民勤县绿洲作为典型荒漠化区域 ,根据荒漠化土地分类体系确定决策树的结构及各类地物在树形中的位置。基于各类地物的光谱反射特性和图象数据反映的综合特征 ,采取相应的识别和提取方法 ,以最大限度地利用遥感数据源。对于非荒漠化土地分类 ,利用土壤调节植被指数、阈值数字信号统计可以分离成一类树枝 ;而重点讨论的 3种荒漠化土地类型的分层分类 ,则相继采用光谱特征提取、几何特征提取、纹理特征提取、监督分类以及植被指数等复合识别指标进行分枝。结果表明 :利用决策树分层提取法可以有效地排除和避免提取地物时所有多余信息的干扰及影响 ,目标明确。同时 ,为提高分类的精度 ,开展野外遥感调查和特征分析是极其重要的。

This paper is given for studying and discussing the auto classification of desertification land types by using Landsat TM data in a typical desertification region of the oasis of Minqin county, Gansu province. Based on the desertification land classification system, the structure of the decision tree and the locations for each land covered in the tree are determined. According to the spectrum reflecting characteristic of targets and the integrated features of TM image data, identifying and extracting methods are correspondingly used in order that the remote sensing data are drawn on maximally. At first, the non desertification lands can be extracted from original TM image by using characteristics of Soil Adjusted Vegetation Index(SAVI) and Digital Numbers(DNs). Then the major research focuses on the multi layer information discernment and extraction of three types of desertification land covers and their respondent grades. Each grade is an object and is processed to yield a layer. Overlaying every layer can create a complete map of desertification land types. In the image processing, the spectrum reflecting properties, statistics training sites Supervised Classification(SC), geometry and texture properties analysis and Normalized Difference Snow Index(NDSI) have been respectively utilized for distinguishing and classifying the different land types. The results show the application of decision tree and multi layer technology could decrease the possibility of interaction and impact with others message and make target simplified in pixel identification. Meanwhile, the investigation and adjustment in the fields is quite important for analyzing the land cover distribution and comparing the difference between grades.