Research On The Algorithm And Realization Of Bill Character Recognition System

To increase the automation level of China printing industry and to aimed at the serious lag of theories study and system development to online character recognition of high-speed bill image, real-time processing algorithm and theory of motion image that based on machine vision technology is studied in the thesis, and some key technologies of high-speed presswork online character recognition are realized.According the analyses to the technical requirement and application feature of bill image character recognition, total technology scheme, the flow and research technique of online character recognition of bill image is depicted. Through reading a lot of related literatures around the world, technology of character recognition is analyzed and summarized from the aspects of its present development and their promising future.A piecewise linear transformation is chose to enhance the contrast of gray-scale image, and an optimum gray value threshold processing method based on iterative operation is selected to apply to image segmentation. A new skew correction based on directional projection is adopted to get the image corrected .And a directional complement algorithm that base on the mathematic morphology is proposed to eliminate motion blur in bill figures.A dynamic localization algorithm is introduced to reduce the possibility of error segmentation. According to feature of printed bill image, character segmentation algorithm that based on projection histogram is chose to character division accurately by comparing with connected domain algorithm based connectivity of the character. This paper also introduces the method of cracked character segmentation. According to feature extraction, this paper introduces grid feature and the moment features which are invariant to rotation in theory.Contrasting the two character recognition algorithms: template matching and BP neural networks algorithm, neural networks algorithm is adopted. A modified BP algorithm is designed and applied on printed number recognition.At last, the character recognition system of bill image is realized. Now the system is test running on the fieldwork of industry successfully.