Railway Ticket Automatic Identification Technology

With the fast development of computer science, pattern recognition and digital image processing technology have been widely applied to many fields such as industry, agriculture, transportation, national defense, scientific research and so on. They are the some of the hottest topics which are being researched and applied.The automatic recognition of railway passenger tickets is the premise of the automatic control in the tickets-inspected job of railway. Aimed at the current condition in Railway Corporation, we bring forward automatic recognition system facing to image processing, pattern recognition and automatic control. The research topic is involved the problems like image segmentation, matching and character recognition etc. The research center focuses on whether can find a valuable and effective algorithm to realize the passenger tickets recognition.The automatic recognition system has some important aspects: passenger tickets code character segmentation and recognition, passenger rickets barcode segmentation and recognition and control platform of entry gate. The dissertation mainly focuses on automatic recognition technology of passenger tickets. First, we introduce several methods common used in the image processing. Then, based on the current theoretic algorithm, much attention has been given to the research center. The innovation and the major contribution of this thesis include:Several image pre-processing methods aiming at gray image are studied, including histogram equalization, filtering, binarization, character normalization, etc.The passenger tickets code can't be solely segmented using conventional gray image processing method because the tickets contain many gray levels and inherited background. In order to solve the problem, the HSI color space transform is used. What's more, we propose a discrete KL transform color image processing method. The experiments results show it plays valuable in the segmentation.The template correlation matching algorithm and neural network algorithm of character recognition are studied: Much analysis is given to template correlation matching algorithm, which is relevantly improved the calculating speed.The particle swarm optimization (PSO) algorithm is used to optimize the initial parameters of the BP neural network, including the weights and biases. The hybrid algorithm takes full advantages of the better performance of global optimized search of PSO and the local optimized search of BPNN, which can improve the tickets code character recognition rate.We propose a max continuous string segmentation method to realize the barcode segmentation. According to the work experience, some useful designs of entry gate control platform are listed.In the last part of the dissertation, conclusions are made. The deficiency of our research is indicated and the directions of future research in the field are further proposed.