The design of neural network configuration for object recognition

The design of a neural network configuration for object recognition is described. Recognition is achieved by determining the type and angle of orientation of the scene object. Supervised networks configured as a conventional classifier and three variations of a fuzzy classifier are investigated. Their performances are evaluated with the reference of correlation coefficients. Results demonstrate the superiority of the fuzzy neural network designs for predictive accuracy compared to the conventional neural network classifier. All network configurations yielded correct object angles