A spectral-spatial hyperspectral data classification approach using random forest with label constraints

A new classification approach using random forest with label constraints is proposed to deal with the underutilization effectively of spectral and spatial information for hyperspectral image classification. Firstly, the principal component analysis extraction method is adopted, and the extended morphological profiles of the image are extracted from the principle components images by using mathematical morphology method. Then random forest is constructed based on the extracted features. Finally, the label constraints based on space continuity is used to constraint the results by using the label information of its neighborhoods on image space. The classification result is decided by voting strategy. Experimental results of several real hyperspectral images demonstrate that the proposed approach outperforms the random forest method without constraint and the popular SVM classification method.