Design And Implementation Of Image Processing System For Personal Certificates

With the social development of real-name system, information on the documents quickly and accurately capture become an increasingly important issue. Performance of hardware and digital image processing technology to improve the rapid development of information collection documents contributed greatly increase system performance. Of documents as a function of information collection system as a whole image processing part of the personal documents of the system to the effect of large, and, according to the different systems, and its appropriate treatment are different. This topic will focus on the project team designed a set of documents, information collection system, personal documents to present its corresponding design and implementation of image processing. The focus of this thesis include the three parts of the image acquisition, processing and management. First of all the integrated use of image processing and analysis techniques, and the characteristics of the documents, scanned images, the image preprocessing. Secondly, since the binary is a key step in the pretreatment in order to reduce the impact of noise, set the partition template gray-scale image into several sub-blocks, in-depth study of the existing image binarization method based on the proposed a bilinear interpolation method for gray-scale interpolation image binarization algorithm and improved the horizontal projection algorithm and the vertical projection algorithm were used to complete the accurate segmentation of the exact location of the line of text and individual characters. Finally, in a single character images owned by one and refinement process for the identification numbers, to extract better reflect the digital statistical and structural properties of multi-layer network based on forward structure of the BP learning algorithm, the recognition results can carry out an initial check processing, and so to some extent, improve the recognition accuracy of the number. The experiments show that the algorithm is simple, easy to implement, identify better.At present, the system has been in trial operation of the unit, users reflect the good. The system can easily obtain the final image to achieve fast and accurate extraction of text messages and documents the collection head, and the information on the identified manual verification and correction. OCR accuracy rate of 95% or more, can be better put to use, while the use of this system has great value, the system can also be easily extended to other applications.