Research On Land Cover Classification Method Of Beijing-tianjin-hebei Region Using MODIS And TM Data

Land cover is an important part of land-based life support systems. The study of land cover as an important parameter, if timely and accurate understanding the surface of the land cover situation, it is very meaningful. With the continuous development of remote sensing technology, there are tens of thousands of remote sensing data every day. How to select from the mass remote sensing image for the characteristic parameters as the classification index, making remote sensing of land cover classification can be rapid, accurate, automated, which is today's hot issues.Currently, there are many domestic and international land cover classification system.Most of the land cover classification system was established to serve its own research objectives. However, as the study area, methods and purpose of the different, the classification system is not uniform.The current international, there is not an accepted classification system was uniform. Presents a variety of sensor resolution, hyperspectral, high phase, to play the advantages of remote sensing image itself will be a better application of remote sensing images of land cover classification studies.This article used as a research area of Beijing, Tianjin, Hebei region. Using the FAO of the dichotomy of principle and reference to Liu Jiyuan land resources classification system, combined with the natural characteristics of Beijing, Tianjin, Hebei region and tillage modes, a set of land cover classification system with Beijing, Tianjin, Hebei region. Selecting MODIS and Landsat-TM as the main data source for land cover classification. Classification Method: Select the vegetation index as indicators of MODIS image classification. As the seasons change, different land cover type of NDVI also change. While the same region, climate and environment, growth conditions, the same conditions, all land classes have similar NDVI curve. Play high temporal resolution MODIS images using the 12-month NDVI data accurately monitor the forest land, shrub land, grassland, arable land, no vegetation changes during the year, Automatic generation of remote sensing image classification tree by using the CART decision tree. As the Landsat-TM image in the spatial resolution of the advantages, manual interpretation of some of the land type (such as: urban land use, land villages and towns, rivers, water, sand, etc.) have certain advantages. In the 100m of the scale, the Landsat-TM on the MODIS image classification result, which improved the classification of certain lands in the interpretation of the results. Full advantage of the two images of the High time resolution, high-resolution features, than a single image with a sensor received the results of land cover classification more realistic reflection of the surface of the land cover conditions.Innovation in the thesis:(1) Proposed a set of land cover classification system be true about Beijing, Tianjin(2) Use of High time resolution MODIS images and high spatial resolution TM image as the main data source for land cover classification.(3) Select Vegetation Index - NDVI as a classification index, CART decision tree algorithm using MODIS images of land cover classification and manual interpretation of Landsat-TM combine the results of correction.