Forest Vegetation Types Classification Based On Modified Vegetation Index

Forest is regarded as "the lung of the earth",it's importance is self-evident,and it has the practical significance in obtaining the information of forest vegetation quickly and accurately.In addition to the traditional field measurement method,remote sensing technology can be used to obtain forest vegetation information quickly in a large amount and in a timely manner.In this study,the study area is located in Hexigten Banner,Chifeng City,the Inner Mongolia Autonomous Region,China.The data sources are 2 remote sensing images of Landsat 8 and GF-2 satellite,acquired in June 2015.To acquire the forest cover information,the forest is divided into broad-leaved forest,coniferous forest and shrub forest.Four kinds of frequently-used supervised classification methods are used to extract forest cover information,including maximum likelihood method,minimum distance method,neural network method and support vector machine method.The modified NDVI(Normalized Difference Vegetation Index)is used to the decision tree method to extract forest vegetation type information.The evaluation result shows that the mean overall classification accuracy of the decision tree method using modified vegetation index is 77.36%,and the modified vegetation index improves the accuracy of decision tree method apparently.And the overall classification accuracy of supervised classification methods is from 74.400%?85.52%.The decision tree method has the simpler principle and the less parameters(with less uncertainty due to parameter setting),and is easy to implement,though the accuracy of supervised classification method is slightly higher.