Recently there has been an explosive growth in microelectronics and chips, the Internet, and new information technologies. In turn, among the many sections of IT-technology, the most progressive, both in the field of theoretical development and applications, is artificial intelligence. Pattern recognition in artificial intelligence refers to a wide range of problems: recognition of images, symbols, texts, smells, sounds, noise, situations. In the software market there are systems based on the recognition of traits, equipped with databases and knowledge, with the ability to adapt and learn.  This paper deals with supervised classification with ENVI software to recognize 5 classes from Landsat-7 satellite image. The paper applied the Mahalanobis Method to divide into classes and determined the relative accuracy of recognition for each class.