面向专利发布特殊字符的识别系统-控制科学与工程专业毕业论文.docxVIP

面向专利发布特殊字符的识别系统 面向专利发布特殊字符的识别系统 SPECIAL CHARACTER RECOGNITION SYSTEM FOR PATENT ISSUED ABSTRACT This article is mainly concerned about the auto-recognition of special characters about Greek letters and Rare Chinese Characters in patents issued, this system is developed for secondary recognition after the Hanwang recognition. This article analyzes the feasibility of the system theoretically and initially designs a character recognition system of BP neural network. This article introduces the image pre-processing of image binary and normalization, which are well-prepared for feature extraction. And in the part of feature extraction, grid feature and projection feature are both designed with, that would take into account the overall situation features and the local situation features of the characteristics, which has laid a good foundation for good recognition rate. A lot of research work has been done in the design of BP network classifier, the valuable experience of our predecessors are referenced, many good improve programs based on many actual tests are selected. The structure design process of BP network input layer, hidden layer and output layer, the process of setting network parameters and the design process of error criterion function are all described by detailed language. At last, this article describes the overall system design and performance analysis. The system design procedure contains the pre-processing and feature extraction, BP network training, the use of BP network to recognition and database operations. The design flow charts and the main function code are given. This paper describes the training and recognition of the Greek letters and Chinese characters cut plans from the patent literature, and shows the system has the advantages of recognizing right of the character with slightly deformed and has higher sensitivity of Chinese characters through the test results. KEY WORDS: system of recognition, image processing, feature extraction, BP neural