SUMMARY
The study aimed to development a prediction model for soil erosion degree by image analysis techniques. The spectral information was obtained by image analysis in the RGB and HSB color system, and by calculus resulted rgb normalized values. Specific indices were calculated: intensity (INT), normalized difference index (NDI) and dark green color index (DGCI). The correlation analysis emphasized the existence of high levels of interdependence between specific indices and normalized color data rgb, respectively luminance (L). The regression analysis has enabled the creation of estimation models for soil erosion degree (DSE), in the form of linear equations in relation to luminance (R2=0.999, p<<0.001, RMSEP=25.5766) and INT (R2=0.998, p<<0.001, RMSEP=25.5833), and 2nd degree polynomial equations in relation to DGCI (R2=0.768, p<0.001, RMSEP=28.3275). Clustering analysis facilitated the grouping of the studied cases in two distinct clusters with four sub-clusters, under conditions of statistical accuracy, Coph. corr. = 0.831.