论文部分内容阅读
文章以安徽省淮南市为例,采用2005年Landsat-5TM多光谱数据,分析地物谱间关系,选择改进归一化差异水体指数(MNDWI)、归一化植被指数(NDVI)、归一化建筑指数(NDBI)、TM4+TM5、TM4+TM5+TM7、DEM高程和坡度等特征值,构建决策树分类规则,完成研究区土地利用遥感分类。结果表明:MNDWI和TM4+TM5特征值能较好地区分水体与非水体;NDVI和NDBI可完成非水体区域植被与非植被信息分类;DEM和TM数据6波段和值可完成建筑用地、裸地和沙(旱)地分类;DEM和TM4+TM5+TM7能较好地解决耕地和园地混淆问题。决策树法分类总精度和Kappa系数分别为90.29%和0.87,相比较于最大似然分类法和基于特征提取波段的最大似然分类法,精度均有所提高。
Taking Huainan City of Anhui Province as an example, the Landsat-5TM multi-spectral data of 2005 were used to analyze the relationship between land-use spectra to improve the normalized difference water index (MNDWI), normalized difference vegetation index (NDVI) Building index (NDBI), TM4 + TM5, TM4 + TM5 + TM7, elevation and slope of DEM and other eigenvalues, the classification rules of decision trees are constructed and the remote sensing classification of land use is completed. The results show that: MNDWI and TM4 + TM5 eigenvalue can distinguish between water body and non-water body well; NDVI and NDBI can do classification of vegetation and non-vegetation information in non-water body area; DEM and TM data 6- And dry (dry) land; DEM and TM4 + TM5 + TM7 can better solve the problem of farmland and garden confusion. Compared with the maximum likelihood classification method and the maximum likelihood classification method based on feature extraction band, the accuracy of the decision tree classification method is 90.29% and Kappa coefficient respectively, and the accuracy is improved.