Classifications of remote sensing images using fuzzy multi-classifiers

Fuzzy methods have been widely applied in image classification, which are believed to be more appropriate for handling uncertainty in remote sensing. This paper presents an algorithm integrating fuzzy multi-classifiers in classification. Traditional Mahalanobis distance classification (MDC) and maximum likelihood classification (MLC) are fuzzified by using fuzzy means and fuzzy covariance matrices, resulting in two fuzzy partitioned matrices. The output membership degrees matrix is generated by combining the previous two fuzzy partitioned matrices, with the pixels being classified into the category having the maximum membership degrees. Experimental results indicate that this new method can increase the classification accuracy. Further research is needed for increasing the algorithm´s efficiency.