An Improved Hyperspectral Mapping Using Multiple Classifier Combination

Matched filtering (MF) methods are widely used to detect specific materials based on matches to library or image endmember spectra. This paper used decision tree to combine three MF methods to extract the dioritic porphyrite from the hyperspectral remote sensing image. The study areas located at the Pulang porphyry copper and gold deposits in southwest of China. Firstly, the image was calibrated to apparent reflectance using the atmospheric correction model, and endmember was extracted by PPI algorithm from the intersection area of multi-segmentation and geology map. Then, dioritic porphyrite areas were extracted from hyperspectral remote sensing image using SAM, SFF and MTMF respectively. Finally, the three MF classification results were combined using decision tree. Comparing the classification results and geology map, it is concluded that combining multiple classifiers has the best classification performance and SFF has the better capable of pixel unmixed than SAM and MTMF.