Discrimination of mango orchards in Malihabad, India using textural features

Abstract

Malihabad in Lucknow district of Uttar Pradesh, India is world renowned for its ‘Mango Malihabadi Dusseheri’, a Geographical Indicator (GI) tagged variety of mango. In this study, a strategy has been demonstrated to discriminate the mango orchards using two dates LISS IV data (spatial resolution: 5.8 m; bands: green, red, NIR) in the selected study region of Malihabad and neighbouring areas of mango-growing belt. As mango orchards have unique textural patterns when seen through space, textural features computed using grey tone spatial dependence matrices have been used as an addition to spectral features for better discrimination. Sparsely and densely populated mango orchards have been taken as separate classes. Classification algorithms of Maximum Likelihood Classifier and Support Vector Machine have been compared for their effectiveness as well as majority analysis has been used as a post-classification process to improve classification accuracy. Classification accuracy increased by many folds from 71 to 80% using only spectral bands to 85–89% using both spectral and textural features and majority analysis as additional post-classification process. Final distribution map of sparsely and densely populated mango orchards have been prepared after intersection of respective classes in the classified maps generated using two dates.