FPGA Implementation Of The Characters Recognition Based On Neural Network

Neural network in machine vision， mechanical automation， Aeronautics andAstronautics，remote sensingand other fields have a great application value. Theartificialneural network, based on theBiology, got the attention of the whole scientific communityfrom the beginning. FPGA, withhardware programmable and parallel computing, provides anew approach for the ANN’s realization. It also solves the problems of the general computer’sreal-time performance, high costs. Proposed in this paper on the FPGA realization method ofthe BP neural networkbased on this background,is designed to deal with identification of thesmall class group of characters image such as the license plate number, theindustrialinstrument.Based on the others’ work, aiming at the small class group of characters, animplementation based on digital image processing and BP network on FPGA for automaticidentification is proposed. Firstly, gathering a large number of samples and realizing thenonlinear mapping between the image and the characters based on the BP (back propagation)neural network. Then, considering the feasibility of the nonlinear mapping on FPGA, acommon method that has a high-speed identification of the characters image on FPGA isproposed. Result shows that this scheme can guarantee the stability of system design andrealize high-speed identification.