关键词: 稀土矿区, 遥感监测, CART决策树, 纹理特征, 裸土指数, 遥感影像分类

Abstract: In order to accurately reflect the mining status of rare earth mining area in Southern Jiangxi Province，taking Xunwu County of Jiangxi Province as the study area and selecting Landsat-8 multi-spectral image as the data source，through the extraction of the feature information of mean texture，bare soil index （BSI） and normalized difference vegetation index （NDVI），the mining information ore rare earth of the study area is identified by using the classification method based on multi-source data CART decision tree.The results show that the overall accuracy of the classification is 89.43% and the classification precision of the mining area is 88%.The classification accuracy is better than the ones of CART decision tree classification method based on spectral information and maximum likelihood classification method.Based on the above discussion results，remote sensing dynamic monitoring for the rare mining area in the study area from 2013 to 2016 is carried out.It is found that the increasing mining area is mainly distributed within the scope of mining boundary，the reduced mining area is distributed within and outside the mining boundary and the degree of reduction is 41%，which further indicated that the government and related departments have played an important role in developing a healthy and orderly rare earth industry.The above study results show that the classification method based on multi-source data CART decision tree has certain feasibility in information extraction and dynamic monitoring of rare earth mining area.

Key words: Rare earth mining area, Remote monitoring, CART decision tree, Texture characteristics, Bare soil index, Remote sensing image classification