Design Of Latin Character Recognition Module And Application In Business Card Recognition

Business card exchange is an important way of communication in people’s daily life. It’s still great demands for people to transform business card to electronic information. This process is based on the support of business card recognition technology. Some research has been made in the Held of business card recognition technology. This article will apply the existing technology to business card recognition during the process of edge detection, layout analysis and character recognition, to improve some of the module mentioned above, especially the character recognition module.Canny edge detector and Hough transform are modified and used to the edge detection of business cards. Connected Component Analysis is used to analyze the layout. The Deep Neural Network technology which rose again in recent years is used to character recognition module. Stacked Denoising Sparse Auto-encoder Network is used to train the model of feature extraction and Softmax Regression is used to classification. The priority of Deep Neural Network to shallow neural network technology and the promotion of recognition rates for Deep Neural Network from denoising technology will be shown in the quantified experiment.