Land cover classification of aerial and satellite remote sensing data has become a challenging problem due to the recent advances in remote sensor technology that led to higher spatial and spectral resolutions. Conventional classification using spectral features can be greatly improved by integration of spatial information. Extraction of spatial information can be computationally expensive. This research paper presents a novel sensor independent classification system for dealing with the challenges of spectral spatial classification of high volume remote sensor data. Using the proposed band reduction method, the dimensionality of the input image can be, optionally, reduced. The band reduced image is then split into two mutually disjoint pure and mixed pixel subsets by a pixel segregator built using extended mathematical morphology techniques. A support vector machine based classifier then classifies the pixel subsets by adaptively including the usage of expensive spatial information based on the pixel categorization is proposed. The computational efficiency obtained by adaptive inclusion of spatial features, flexibility and accuracy of the proposed system are demonstrated by evaluating the classification results using several hyperspectral and multispectral data sets.
Indian Member 40.00
Others Member 3.00