Mapping and classification of crops using high resolution satellite image

Abstract

In the present study an attempt was made to perform land use land cover classification at Level-III in order to discriminate and map individual crops. IRS Resources at 2 LISS IV sensor imagery (5.0 m spatial resolution) of September 2014 was utilized for the study. A hybrid classification approach of unsupervised classification followed by supervised classification was adopted to identify and map the crop area in Kodumudi block, Erode district of Tamil Nadu. Signature evaluation was carried out to study the class separability and through cross tabulation and the accuracy was assessed by error matrix. The signature separability analysis to classify various land cover classes indicated that the class viz., waterbody, settlement, sandy area and fallow land were better and for vegetation sub-classes viz., individual crops were poor, which means classification of individual crops was a challenge. The overall accuracy with three different algorithms varied from 56 to 65 per cent and this low accuracy was due to the problem in discriminating the tonal variation and spectral pattern of individual crops in the study area. Thus, classification of vegetation categories into individual crops using LISS IV data resulted in moderate classification accuracy in areas with multiple cropping.