Land Cover Classification In Zhejiang Province Based On MODIS Satellite Data

As one of the most advanced ground observation methods,remote sensing has a continuous coverage of the space and a continuous imaging phase.The use of remote sensing for land cover mapping has become an important land cover types of data access methods. It plays an important role in the establishment of land cover data set, which is in terms of different scales.In this paper, use the MODIS satellite data products, choose the year2008, eight days500m synthetic surface reflectivity data.The data pre-processing, including generating mask data, overlaying the data, the generating of vegetation index data sets,and smoothing filter reconstruction data based on Savitzky-Golay of the VI time series. After filtering get the EVI time series data which is reflect the seasonal variation of vegetation more better.The core of the classification method of the remote sensing image is converting the remote sensing data into object information.In this paper, based on the traditional classification model, use two classification models:A kind of decision tree classification based on MODIS-VI series timing and a land cover classification based on SVM. At last,finish the classify, and do the classification accuracy for the validation and analysis the results.The results indicate that:1. Decision tree classification, by setting the EVI threshold, is feasible.2. Compared with traditional classification methods, decision tree algorithm reduces thecomputational complexity, at the same time improving the classification accuracy greatly.3. SVM classification is more simple than decision tree, and it’s easier to achieve, but it will cost more time.