We propose a neural network based vision system for attending micropieces manufacturing process in micromechanics. The system permits us to recognize the shape of the micropieces (3mm diameter screws) in order to get information for controlling and improving the manufacturing process. The neural classifier used for the shape recognition task is termed Limited Receptive Area Grayscale (LIRA Grayscale). The developed vision system has recognition rate of 96.88%. This work is motivated by the idea of obtaining an automated control system for micromachines. This paper contains a detailed description of the model and learning rules, and discusses future perspectives.
|