Image analysis methods were developed and diversified greatly in recent years due to increasing speed and accuracy in providing information regarding land cover and vegetation in urban areas. The aim of this paper is to process satellite images for monitoring agricultural areas. Satellite images used in this study are medium and high resolution images taken from QuickBird and SPOT systems. Based on these images, a supervised classification was performed of a very large area, having as result the land use classes. Supervised classification can be defined as the ability to group the pixels that compose the satellite image, digitally, in accordance with their real significance. Gaussian algorithm of maximum similarity (Maximum likelihood) was used, referred to in the specialty literature as maximum likelihood method or probabilistic classification, and based on the use of probability theory (function Gaussian) to compare the spectral values of each pixel in hand with statistical " fingerprint "of each area of interest. Practically, conditional probabilities were calculated of belonging to one class or another. The points in the middle of the group have a higher probability of belonging to the certain class, probability intervals (concentric isolines or contours of equal probability) being delimited graphically by izocontours expressing spectral variations within each set of training.