基于多元数据和不同分类算法的遥感影像信息提取及精度评价 ——以祁连山东段为例

以祁连山东段典型山地系统为研究区,通过提取研究区TM影像的主成分、各类植被指数、基于灰度共生矩阵的影像纹理特征以及研究区地形特征等数据,应用最优波段指数方法得到最优波段组合,并运用非监督分类、最大似然法、支持向量机分类法、决策树分类法对上述最优波段进行分类研究。结果表明多尺度数据挖掘有利于分类精度的提高,同时选取合适的判断标准的决策树分类方法在遥感信息提取中有比较直观意义和较高的分类精度。在上述分类方法中分类精度由高到低为决策树分类>支持向量机法>最大似然法>非监督分类法。决策树分类总体分类精度为94.50%,kappa系数为0.9122。


RS Information Extraction and Accuracy Evaluation Based on Multi-elements Data and Different Classify Method——A Case Study of the Eastern Section of Qilian Mountains

Taken the eastern section of Qilian Mountains,Shiyanghe Basin as the study area,the land use and land coverage was classified using TM data in the typical mountainous system. The study area is located in 37°5'37"N-37°6'30"N,102°3'37.2"E-102°6'E with elevation 2 016～4 318 m. Firstly,the retrieved data from TM image such as principal components, vegetation indexes, image textural feature by using gray co-occurrence matrix, the topographical data from DEM (elevation,aspect) were obtained in this study. The optimal bands combination was acquired using optimal band index method. The optimal bands combinations was used to classify the land use and coverage based on non-supervised classification,maximum likelihood method, support vector machine and the decision tree classification method. The conclusion can be drawn from the approach:① the multi-scale data mining is useful to improve the accuracy of classification; ② by comparing the four methods,the decision tree classification method had been selected due to its high accuracy which depended on the appropriate criteria selected.