Hyperspectral sensors can provide each pixel with many continuous spectrum data, and make it possible for each pixel to give us information in different features and classes. In this article, two supervised classification methods, Cross-Correlation and Spectral Angle Mapper, are used for matching with four different features in AVIRIS image. The results from those two methods are similar. Among other things, the results of matching water and grass were good; the results of matching roads and buildings were fair, but further process (i.e. unmixing) may increase classification accuracy.