The Design And Implementation Of Maintenance Information System On Passenger Train Wheel And Axle

Research of pattern recognition has been widely used in text classification, speech recognition, image recognition, video recognition, information retrieval, medical image analysis and data mining etc.. Character recognition is an important content in the research of pattern recognition, the printed character recognition at home and abroad after years of research and have made use of very good results, the production of a number of companies out of the character recognition of appropriate equipment put on the market, has brought great convenience to the text entry. Irregular handwritten character recognition has a large number of random and stroke, is more difficult for handwritten character recognition, handwritten character recognition is also a big challenge to researchers is very exciting, is a research hotspot in recent years. With the introduction of the in-depth study and a large number of technology, researchers handwritten letter recognition proposed many solutions, produced a variety of recognition algorithm. At present, there are a variety of character recognition methods,such as the identification of template method is applicable to printed text; statistical decision method is a statistical classification method of pattern, cannot reflect the fine structure pattern; syntactic structure is pattern recognition using pattern and sub pattern of hierarchical tree structure information completed, anti interference the ability is too weak. All of these methods and the corresponding algorithm, are under certain conditions is feasible, and each has its advantages and disadvantages, still did not develop into a unified, efficient and can be used for the handwritten pattern recognition letters recognition model. Integrated neural networks have become the core now handwritten character recognition research, because it can greatly improve the ability of generalization of the system in the.On the handwritten letter recognition at present is the use of genetic algorithm,which is also based on the genetic algorithm in order to improve the recognition ability of BP(Back Propagation) neural network algorithm, and this article is through the study of relevant information to analyze the both the effect of use. Recognition system two plate learning and recognition, in which the learning part is composed of 7 BP neural network learning and training the composition, adaptive training is training according tothe reference values of selected, calculating the output error, then according to the output error calculation of local optimal neural genetic algorithm network or global optimum, give full play to the role of computing ability of artificial neural algorithm of local search and genetic algorithm for global optimization. Recognition component includes a feature extraction module and recognition module, used to identify the letters just through the network to forward calculation obtains the final recognition results. For the feedforward direct network digital identification calculation to get the final recognition results, without the use of more complex algorithm.The realization of genetic neural network algorithm and the combination of letter recognition experiment with VC++ programming, handwritten letter recognition test average recognition rate of 71.92%, the recognition results are quite satisfactory. The results show that: there is a direct relationship between recognition and handwritten letter specification, when handwritten letter is standardized, good recognition effect.