Neural Networks Applied to the Estimation of Object Orientation

We present in this paper a first approach to the use of artificial neural as a tool to determine the orientation of objects moving on a conveyor belt in a car assembly line. The capability of neural networks to generalise is a key element in the calculation of an object’s orientation. In this sense, a neural network with Competitive Hebbian Learning can identify the angle of a part never used in its training process. The equilibrium between exactitude and processing time is also studied.