Principal component-based algorithm on multispectral remote sensing data for spectral discrimination of tree cover from other vegetation types

摘要

Remote sensing technology offers an effective tool for mapping of forest resources in a cost-effective manner. The Forest Survey of India (FSI) carries out forest-cover mapping of the entire country once in two years, using remote sensing data. Delineation of forest-cover from contiguously occurring agriculture or bushy vegetation on remote sensing imagery is a problem, which is often faced in classification of remote sensing data by digital method. There are several indices, including NDVI for highlighting vegetation areas on the scene. However, these indices have not been found very effective in discriminating tree cover from other vegetation types. In this study, a new approach based on principal components has been tried for the above-stated purpose. A newly defined index has been found to yield good results in separating pixels having tree cover from the other types of smaller vegetation.