基于不良地质识别的分类方法研究与分析

Research and Analysis on the Classification Method Based on the Bad Geological Identification

中国西部的沙漠、冻土和盐渍土等典型的不良地质现象日益显现,基于遥感影像的不良地质识别已经成为遥感处理研究领域的一个热点和难点。以新疆尉犁县罗布人村寨为研究区域,针对当地典型的不良地质体遥感影像特征,主要探讨SVM分类、K均值分类以及基于因果关系的贝叶斯网络分类3种分类方法,初步尝试不同分类方法的融合,并通过实验对比分析了3种方法的分类效果和精度。结果表明:SVM分类结果为块状分布,K均值分类结果为点状分布,基于因果关系的贝叶斯网络分类取得了更好的分类精度,3种方法取得的影像融合更好地表达了多种不良地质体的识别效果。

Typical adverse geological phenomenon of western China such as the deserts,permafrost and salt marshes soil are commonplace for the area,and the remote sensing based on image recognition in these areas has become a hot and difficult research field of remote sensing processing.This article regards Yuri Rob village in Xinjiang as the study area,against the local typical adverse geological remote sensing images,investigating the three classification methods：SVM classification,K-means classification and Bayesian network classifiers based on causal relationship.A preliminary test of different integration of classification methods is made,and an analysis of the classification’s effectiveness and accuracy of the three methods by experimental and comparison is made.The experimental results show that the result of SVM classification is block distribution,K-means classification for dot is distributed,but Bayesian network classification based on causal achieves better classification accuracy,the results of image fusion got by three methods express a variety of adverse geological recognition effect.