1998 Volume 60 Issue 1 Pages 85-98
Studies were conducted to establish the feasibility for a charge simulation retina model to identify shape and size. The retina consisted of sensory cells which detected image features generated by different 3-D objects that were arbitrarily located in the retina's field of view. The image features were compressed by a charge simulation method algorithm by computing output signals at work cells located in the retina. With these signals, neural networks were used to classify each image, to identify shape and size. Classification rates above 75% for both shape and size were obtained, showing that it is feasible for the retina to identify shape and size. Since object displacement affected the performance of the retina, to minimise misclassification, it is necessary that the centre of area of each object is kept within a circle whose radius is one-tenth that of the retina and is measured from the centre of the retina base.