ROBOTS ARE AN IMPORTANT MEANS OF INDUSTRIAL AUTOMATION. THEY POSSESS VARIOUS FEATURES THAT MAKE THEIR USE UNRIVALED BY OTHER MACHINES. GRASPING IS ONE OF THESE IMPORTANT FEATURES. FOR A ROBOT TO GRASP AN OBJECT, IT HAS TO LEARN THE LOCATION OF ITS GRASP POINTS. IF, DUE TO OPERATIONAL CONDITIONS, THE OBJECT ASSUMES A DIFFERENT POSITION, THE POINTS OF GRASP HAVE TO BE REDEFINED. THIS PAPER PROPOSES A METHODOLOGY FOR OBJECT RECOGNITION USING NEURAL NETWORKS IMPLEMENTED FOR ROBOT GRASPING.