A generalized orthogonal subspace projection approach to unsupervised multispectral image classification

Orthogonal subspace projection (OSP) has been successfully applied in hyper spectral image processing, In order for the OSP to be effective, the number of bands must be no less than that of signatures to be classified. This en sures that there are sufficient dimensions to accommodate orthogonal projec tions resulting from the individual signatures. Such inherent constraint is not an issue for hyperspectral images since they generally have hundreds o f bands, which is more than the number of signatures resident within images . However, this may not be true for multispectral images where the number o f signatures to be classified is greater than the number of bands such as t hree-band pour l'observation de la terra (SPOT) images. This paper presents a generalization of the OSP called generalized OSP (GOSP) that relaxes thi s constraint in such a manner that the OSP can be extended to multispectral image processing in an unsupervised fashion. The idea of the GOSP is to cr eate a new set of additional bands that are generated nonlinearly from orig inal multispectral bands prior to the OSP classification. It is then follow ed by an unsupervised OSP classifier called automatic target detection and classification algorithm (ATDCA), The effectiveness of the proposed GOSP is evaluated by SPOT and Landsat TM images. The experimental results show tha t the GOSP significantly improves the classification performance of the OSP .