Study On Land Cover/land Use Classification Based On MODIS Data

It is a fundamental research using remote sensing technology which can provide accurate and information in land cover/land use classification. As new generation remote sensing data, MODIS has a better spectral resolution and spatial resolution which can be used for macro-scale land cover/land use classification. MODIS designed seven bands which were specifically used in the land cover/land use classification and it has showed a strong advantage and wide range of application in land cover/land use. In this paper, Northeast of China was selected as the study area. MODIS data of 2005 were used based on a variety of classification methods for land cover/land use classification in this area. Research took following conclusions:1. Classification accuracy was directly affected by the selection of classification features. Features extraction based on MODIS data could significantly improve land cover/land use classification accuracy in regional-scale. In order to achieve a high accuracy of classification, MODIS1~7-band reflectance, vegetation index, water index and texture characteristics should be combined optimize used in land cover/land use classification.2. Combined various classification methods could greatly improve the land cover/land use classification accuracy using remote sensing data. The uses of multi-temporal MODIS data information and spectral information for different types of land cover/land use type in multiple classifications could improve the classification accuracy.3. The classification accuracy will be improved according to combine Landsat TM data and MODIS data. Training samples were extracted using the high resolution Landsat TM data and applied them in MODIS classification. It's also a feasible method to evaluate the MODIS land cover/land use classification accuracy.4. In the point of application, the 9 classes land cover/land use which base on MODIS data in Northeast China showed a great feasibility and practicability. The total accuracy and Kappa coefficient are 82.7822% and 0.8031, respectively. The results also coordinated with the the actual situation of land use in Northeast China. In this paper, MODIS multi-temporal data, including MODIS 7 bands of reflectance data, EVI, LSWI and texture characteristics are choosed for regional land cover/land use classification and have a good classification results. It showed that the multi-spectral and multi-temporal MODIS data have high practical value, high potential applications and well prospect in regional scale land cover/land use classification.