Research And Application On Document Image Recognition Technology

Abstract:The improvement of the hardware performance and the rapid development of digital image processing technology have contributed to the document information collection system. As a major function of document information collection system, document image processing affects the final effects. For such a document information collection system, a corresponding research on document image acquisition technology has been done.The thesis elaborates the design and implementation of the document image processing system, including the pretreatment, the correction of fisheye, the processing of document outline, the incline correction of document outline and image segmentation. WB is finished in the pretreatment. By the fisheye, it solves the problem of the width of the products on the market. And the barrel distortion can be avoided by using the cubic spline curve to fit the image and the bilinear interpolation to make the gray interpolation. The system can extract the outline of the document by the Freeman chain code method and finish the OCR of the document by calling the OCR engine in the Microsoft MODI. To identify the human face in the document, the thesis takes the RBF neutral network to classify the mode. The RBF neutral network is good at nonlinear mapping and can solve the nonlinear problem in the face recognition. It also uses the Supervised Clustering Algorithm to initialize the center of hidden layer and the Gaussian width as well as the Hybrid Learning Algorithm to practice the RBF network. By analyzing the cluster performance and classification performance of the RBF network, the simulation in the ORL face database shows that this system gets a high recognition rate in the face recognition.The system can get the image conveniently and extract the information and collect the face quickly and exactly. It can also check and correct the information that had been identified. The recognition rate of OCR can reaches95%.Now, the system has been operated in a unit and gets a good feedback. It is valuable to apply to the other relevant areas.