OBJECT RECOGNITION OF ONE-DOF TOOLS BY A BACKPROPAGATION NEURAL-NET

Abstract

We consider the recognition of industrial tools which have one degree of freedom. In the case of pliers, the shape varies as the jam angle varies, and a feature vector made from the boundary image varies with it. For a pattern classifier able to classify objects without regard to angle variation, we have utilized a back propagation neural net. Feature vectors made from Fourier descriptors of boundary images by truncating the high frequency components were used as inputs to the neural net for training and recognition. In our experiments, the back-propagation neural net outperforms both the minimum-mean-distance and the nearest-neighbor rule widely used in pattern recognition. Performances are also compared under noisy environments and for some untrained objects.