Design And Realization Of Commercial Invoice Handwritten Number Recognition System

At present, receipt processing is a arduous work in many commercial and government organs, communication organs, hospital and trade of insuring. For a long time, because of the management method is backward, each kind of receipt printing, arrangement, binding and filing needs to waste the massive manpower and resource. So in order to using computer to instead people to do these arduous work, this paper according to the commercial invoice uses image process and pattern recognize method, realizes one receipt automatic processing system, discuses every procedure module and provide the concrete solution.At first, analyzing and study the invoice layout, confirm the relative position of static area and dynamic area, calculate the size of form and find interest area etc. Using these information form the template for different invoice. So when we know the type of the invoice, we can match the templates and get the information we interest.Use improved Hough algorithms to orient the invoice. We can also calculate the slope angle during the Hough algorithms. If the invoice with too big slope angle, it needs to correct the slop to avoid the influence to handwritten number recognition.Emphasize researching the method that taking off the form without harming the frame of handwritten figure and cutting apart of pen figure connecting. Use the morphology principle to compensate and reduce the noise for the pick-up figure. Combine OPTA thinning algorithm and Hilditch thinning algorithm, make the thinning result more smooth and more level. Draw the macroscopical statistics characteristics and micro structure characteristics of the handwritten figure, and through calculating, get rid of the false characteristic .According to different characteristic, different complementarity of different classify, use the method to make up of different classify, combine believe degree theory, put forward the concept of every classification degrees. Every classification degrees reflect the credibility of classify divide sample into different classes. According to the credibility, dynamically make up the combination of classify. Give full play to the strong point of each classifying device on eachcategorised performance, has improved the discerning rate of classification device after making up.Through analyzing and comparing the information of invoice, to discern the validity of the result. Effective reduce the wrong rate of the result.