Selection strategy of classification methods for hyperspectral remote sensing data

In the present stage, there are a large amount of classification methods for hyperspectral remote sensing data. It has always been a crucial problem that how to select a suitable classifier because these methods may have different application scopes. Based on Pushbroom Hyperspectral Imager (PHI) data that includes 80 spectral bands and covers 8 vegetation classes, this study performed a serious classification experiments to test the characteristics of some popular classifiers (Maximum Likelihood, Spectral Angle Mapper, Decision Tree Classifier and Support Vector Machine) with different training sampling numbers and dimensions of feature bands. According to the classification results, this paper analyses the application of different classifiers on the sophisticated vegetation classification with hyperspectral data.