基于Landsat TM的城镇信息自动提取研究

Study on Retrieving Urban Residential Areas Automatically Based on Landsat TM Data

为实现城镇信息的自动提取,在Erdas的Modeler模块下构建计算模型,选取大城市、小城市和乡镇各一个研究区进行实验,通过试验发现了常见的基于LandsatTM的城镇信息提取方法的不足。运用了综合阈值法进行城镇信息自动提取,条件为NDVI<0、TM1/TM4>T和TM2+TM3<TM4+TM5。由于TM图像成像时间的差异,T值在1.5～1.9之间,春夏季在1.5附近,秋冬季在1.9附近,该值也可根据实际情况灵活调整。该方法在3个研究区效果较好,消除了山体阴影、水体及其边界、城市绿地、河流等的影响,精度较理想。

In order to extract the urban residential areas automatically, this paper chose one of metropolis, small cities and towns respectively to carry out research experiment based on the calculation models built using the spatial modeler of Erdas Imagine. The results showed that the traditional extraction methods based on Landsat TM had disadvantages. This study established a comprehensive threshold method to retrieve urban residential areas automatically under the conditions of NDVI〈0, TM1/TM4〉T and TM2 ＋ TM3〈TM4＋TMS. T took values in the interval of 1.5 to 1.9 at different acquisition date of TM images. For example, T took values about 1.5 in spring and summer but 1.9 in autumn and winter. Of course, the value could also be adjusted flexibly according to actual situation. This method had a better accuracy in the three study areas because it rednced the impacts of some features such as the mountain shadows, water and its boundary, urban grass land and river.