基于MODIS数据的北京西北部地区土地覆盖分类研究

本文主要基于MODIS 16天合成的NDVI时间序列数据、8天合成 LST数据、1∶5万DEM数据以及其他辅助数据相结合，进行北京西北部地区土地覆盖分类的研究。首先选取适合于MODIS数据分类的土地覆盖分类系统，然后用PCA方法对NDVI时间序列数据进行信息增强与压缩处理，以排除各种干扰因素，提高分类精度。最后结合LST数据、DEM数据及降雨温度数据，利用?齂-均值非监督分类法，进行研究区的土地覆盖分类，经过分类后处理，得到北京西北部地区的土地覆盖分类图。分类结果表明，使用250m分辨率MODIS数据，结合本文所用方法，能够实现较大区域的土地覆盖分类，并且能达到较高的分类精度。

Land Cover Classification Based on MODIS Data in Area to the North-west of Beijing

This paper presented the research on land cover classification in area to the northwest of Beijing based on MODIS 16-day NDVI and 8-day LST time series datasets in 2002, 1: 50000 DEM and some ancillary data. We first selected a land use/land cover classification system suitable to MODIS remote sensing data, the classification system was built based on remote sensing data after comparing and summarizing the features of different remote sensing data. Secondly, in order to improve the classification accuracy, we performed a PCA transform on the MODIS NDVI time series data to remove noises, and selected the first four components as the basic land cover classification data. Finally, combined with MODIS LST data, DEM and other ancillary data of study area, we applied fuzzy K-means non-supervised classifier to get the land cover automatic classification in area to the north-west of Beijing. After post-classification process of the original classification result, the last land cover classification map of area to the north-west of Beijing in 2002 was got. Compared with the manual interpretation classification result based on the ETM remote sensing data, we had a classification accuracy evaluation about MODIS data. The final classification result showed that the 250m MODIS NDVI time series data and the classification method used in the paper not only can suit to large-scale land cover classification, but also can have better classification accuracy. Based on the classification result of area to the north-west of Beijing, we had a statistic about the area of seven land cover types and analyzed their geographic distribution.