Research on band selection of MODIS data

MODIS belonged to the new generation of optical and thermal remote sensor that combines both images and spectrum is one of the most important sources for observing land surface change and analyzing the development of terrestrial biosphere. It is very important to determine the optimal band sets from its 36 bands according to specific application purpose for further management analysis and information extraction of multi-spectral dada. Based on the analysis of selection criteria, a subspace-division-based (SD) method was proposed after the comprehensive contrary of current modish methods. Then one swath MODIS data covered Beijing was disposed for supervised classification experiment to validate the suitability of this method. The validation demonstrated that the accuracy of land using cover classification using the 12 bands selected by SD method is 92.04% without other auxiliary data in medium-scale spatial resolution.