Research On Auto-recognition For Chinese Hand-written Commercial Characters In Form Processing

Presently, bank system was trying to raise the efficiency in its everyday affairs. In its affairs, a lot of kinds of form are applied. These forms always need to be handled by people in the past. The procedure of inputting is not only slow but also easy to make wrong. The commercial form processing system does be constructed to satisfy the need of bank system. The commercial form processing system uses computers and other equipment to facilitate inputting date, and constitute the date base to manage the date. But how to input important information into computer is still a different technical problem. Normally, high-speed scanners are used to scan forms and input their image into computer. Then computers make sure the location important information and abstract it. In the important information, Chinese hand-written commercial character is one main kind. If computer can be used to recognize the image of these Chinese characters, the capability of processing system is to be raised. This paper focuses on the problem of auto-recognition for financial characters.Through past several decades, the OCR technology has improved rapidly. Many kinds of OCR technology has gotten very high recognition rate. But recognition for hand-written characters is still difficult. Although the set of financial character is not very big, recognition for this is difficult because they are hand-written. Different people have different style in writing, and the different writing tools also enhance the difficult. According to these features, we made the arithmetic for financial Chinese characters. The characteristics of this arithmetic are listed below.Firstly, we use gravity-center-transfer method to deal with the problem of out-of-shape for hand-written characters.Secondly, we use 16-scale method to describe the structure of Chinese characters, and this method makes abstracting the feathers of characters more easily.Thirdly, we use panels (7*7+8*8) to count the features and this method makes the result more robust.Finally, we raised a method to make the recognition system more rapidly and more accurately.We gathered more than 40,000 hand-written Chinese characters to make our research more efficiently. According to the test, the recognition rate is above 95%. The result is satisfiable.