The Band Selection Algorithm in Supervised Classification Using Mixed-Pixels and Canonical Correlation Analysis

The commonly used methods for the optimum band selection in supervised classification of multi-spectral data are Divergence, Transformed divergence (TD) and Jeffreys-Matusita distance (JM distance). But those methods might be ineffective when there is a need to change in the number of bands used and some spectral information in multi-spectral data can be redundant in classification process. This study introduces new algorithm with “bands variables set” and “classes variables set”, the canonical correlation analysis is made use of feature classification. Using the canonical cross-loadings we can orderly identify the bands correlation that largely affects the remotely sensed data. To verify the suitability of the new algorithm, the classifications using the each best band combination through TD, JM distance and new method were performed and the accuracy was assesed. As a result of classification accuracy assessment, overall accuracy and k^ for the new method were superior to TD’s and had competitive results to JM distance method.