Mapping specific crop- A multi sensor temporal approach

This study explores the applicability of temporal and multi sensor data for specific crop mapping. For this, temporal data from a single sensor (LISS III from IRS- P6 satellite) was used and classified after selecting the best dates for mapping. In the second case a Landsat- 5 TM image (other sensor/ multi sensor approach) is added to the selected best LISS III temporal dates combination and classified again for evaluating the effect of the addition of a another sensor data (i.e. Landsat- 5 TM) on the overall accuracy of classification. A Possibilistic c-Means (PCM) classification technique has been used for extracting single class of interest (Sugarcane-ratoon) and for including the mixed pixels occurring in the heterogeneous landscape of the study area. In the absence of reference data, evaluation of the soft (fuzzy) classified outputs was done as an entropy measurement, where entropy provides an indirect absolute measurement of the classification accuracy in the form of an uncertainty measure.