3.4. Supervised classification and validation
Subsequently, we performed a supervised classification of the values of the NDVI and IC obtained, relying on data collected in the field, we performed the classification of maximum of likelihood using regions of interest (ROIs tools). These are selected carefully on level of the study area using the satellite data processing software (ENVI 5.2). The combination of the two classifications has given maps of vegetation types and the maps of soils in our study area (Figure 2). The validation of the adopted classification of all the years of study was estimated by using points of truth on ground (Figure 3) and confusion matrix. The latter was calculated for each card taking into account the percentage of classification and the accuracy coefficient named Kappa coefficient. According to Girard and Girard, 1999; the scale of this coefficient is as followings:
(i): Excellent classification :> 0.8
(ii): Good classification: between 0.8 and 0.6
(iii): Average classification: between 0.6 and 0.2