A New Framework for Effective Urban Land Use and Land Cover Classification: A Wavelet Approach

摘要

This study develops a new framework for multi-scale analysis procedures and a new operational wavelet-based algorithm to identify urban classes. For better evaluation of wavelets, traditional classifiers were used: maximum likelihood, Mahalanobis, and minimum distance to identify the same classes. Overall accuracy for maximum likelihood, Mahalanobis, and minimum distance for band-3 and multi-bands were 49%, 43%, 48%, 58%, 56%, and 55%, respectively. However, the wavelet approach with the above measures using Band 3 alone gave much higher accuracies (70%, 80%, and 81%). It can be concluded that the wavelet-based approaches are far more accurate than the traditional classifiers.