Optimizing classification using multi-classifiers for spaceborne hyperspectral dataset

A method for enhancing the classification accuracy from hyperspectral remote sensing images using multiple classifiers and accuracy assessments is presented. The methodology is implemented in four steps. First, classification images are generated using a variety of classifiers. Second, accuracy assessments are performed for each class and overall classification, for every classifier. Third, an overall within-class accuracy index is estimated by weighting class accuracy with the overall accuracy for each class and every classifier. Finally, each pixel is assigned to a class for which it has the highest overall within-class accuracy index amongst all classes in all classifiers. The method was tested on Hyperion image from a study area over Udaipur City in southcentral Rajasthan, India. The results indicate that the method is effective in increasing classification accuracy.