COMPARING MAXIMUM LIKELIHOOD AND MINIMUM DISTANCE CLASSIFIERS FOR LAND COVER MAPPING USING LISS-III IMAGERY

Abstract

Remote sensing is an important tool for producing land use and land cover maps through a process known as image classification. With the respect to this study compared the accuracy of two classification techniques as Maximum Likelihood classification (MLC) and Minimum Distance classification using multispectral images from remotely sensed (LISS-III) image of 05 Apr 2019. Here used study area of (upper Godavari basin) Chhatrapati Sambhajinagar and Ahmednagar region of upper Jaykwadi dam, Maharashtra. India. The area has been classified in five LULC classes as Waterbody, Vegetation, Fallow land, Bulit up area, Barren land. The overall accuracy of the classifier was found, and the Maximum Likelihood classifier produced suitable results. The overall accuracy of Maximum Likelihood was 92.29% with Kapp coefficient 0.86% and Minimum distance has over all accuracy is 78.81% with Kapp coefficient 0.71%. So Maximum Likelihood classification gives better accuracy than minimum distance classification techniques with Resoursat-2 LISS-III image. If the two classes are well-separated, then MLC and MDC will likely give the same result. However, if the two classes overlap, then MLC is more likely to give the correct result. This is because MLC takes into account the probability distribution of the data, while MDC does not.