基于决策树的地质灾害信息提取研究

以汶川县为研究对象,选取2009年landsat5TM影像为数据源,研究地质灾害及其背景地物在TM影像的光谱特征及其波段间的相互运算。研究发现,短波红外波段(TM5)和热红外波段(TM6)是提取地质灾害信息的关键波段,构建了决策树地质灾害提取模型(TM5-TM6)/(TM5+TM6),实现了对地质灾害信息提取,对结果进行了精度评价,kappa系数为0.92,生产者精度为93.2%、用户精度为97.3%,满足要求。解译结果中发现,地质灾害的分布与区内地质构造密切相关,随距断裂带距离的增加而减少、呈线性分布,且84.5%集中分布在距断裂带0~5km区域内,建议震区公路选址应尽量绕避该区域。

Studies on information extraction of geological hazard based on decision tree classification

Taking Wenchuan as study area, make research on the bands operation and finding out the spectral characteristics of geological hazard and other objects in landsat5 TM remote sencing image. The research shows that shortwave infrared band （TM5） and thermal infrared band （TM6） are the key bands for extracting the information of geological disasters. Based on this ,construct a decision tree model （（TM5-TM6）/（TM5＋TM6）） to extract geological hazard information and make accuracy evaluation for the results with a kappa coefficient of 0.92, the producer accuracy of 93.2% and an users accuracy of 97.3%, which meet the requirements. Disaster interpreting results shows that 84.5%geological disaster area are concentrated in the region of 0~5km away from the fault zone. The distribution of geological disasters is linearly, which is closely related to the geological structure of the region, and decreases with increasing distance from the fault. Road location around the earthquake zone should be avoided in these regions.