Research on pattern recognition methods of TM remote sensing images

The pattern recognition of remote sensing images is based on the different spectral characteristics of surface features to identify the features type, mainly including the supervised classification and unsupervised classification. In this paper, the TM remote sensing images are preprocessed firstly, and then land-covered features are classified using the maximum likelihood method, the minimum distance method, parallelepiped method, K-means and ISODATA methods. Finally the results of pattern recognition about TM remote sensing images are analyzed, which shows that the classification accuracy of supervised classification higher than non-supervised classification, yet the latter can be used as classification aids.