Optical Handwritten Digital Character Recognition Technology Research

The optical character, refers to the characters that can be observed by human’s naked eye and often print on the material such as paper. The process of translating Optical Character into the encoding that computer can identify is called Optical Character Recognition process.Optical character recognition has a very wide range of application. Such as mail sorting in logistics system, license plate recognition in intelligent transportation system, invoice processing in financial regulation, card recognition in public security management and so on. But for optical handwritten character, because of writing casually, it is more difficult to recognize than the normal printing characters. Therefore, for handwritten optical character recognition, especially in Chinese handwritten optical character recognition, the results are not satisfactory. This paper focus on the reseach of genetic optimizated error Back Propagation artificial neural network algorithm and support vector machine algorithm,which get better performance than other algorithms, compare the advantages and disadvantages of them on the handwritten numeric character recognition, hope to make guidance for project application.In this paper, C++programming language combined with OPENCV image processing open source library and Google MNIST handwritten digital character database and LIBSVM open-source library is used to do some experiment to test the performance of different algorithms which will be used in the process of OCR. The main work can be divided into three parts, the first part is image preprocessing, the second part is feature extraction, and the third part is identified. Involves the knowledge of binarization, character segmentation, character refinement, looking character outlines, polar coordinate transformation, support vector machines, artificial neural networks, genetic algorithms and so on. Using difference method to compares the features vector is an innovative feature points. Meanwhile, Thesis compare advantages and disadvantages of the previous classical algorithms using in the field of optical handwritten numeric character recognition, receive a relatively high practical value.