Researches On Principle And Algorithm Of ID Card Recognition System

The information acquirement of ID Card plays an important role in establishing the information file database in police, presently the personal information in ID Card is mostly inputted by manual work, it's very inefficient. With the rapid development of OCR technology, the printed character recognition technology has been mature and applied widely. Therefore, if the image processing technology can be used to automatically segment the scanned image of ID Card, and then recognize the segmented characters automatically, the work efficiency would be improved remarkably.The research emphases of this paper include three parts: the orientation of character region, character segmentation and character recognition. Firstly, using synthetically the image processing and analyzing technology to implement the image preprocess, taking the self-traits of the scanned image (disturbances produced by strong shadings and laser falsification-preventing signs) into consideration.Secondly, because of the binarization is the key step in image preprocesses, in order to decrease the effect of noise, a segmentation template is designed to separate the gray image into several subblocks, on the basis of lucubrating the existing image binarization methods, this paper puts forward an image binarization algorithm of ID Card based on iterative threshold, improves on the horizontal projection algorithm and the vertical projection algorithm to accomplish the accurate character line orientation and individual character segmentation precisely.Finally, after normalization and thinning in the individual character image, a 22-dimension mixed feature vector is selected which can best represent the number's characteristic of construction and statistics. Euclidean Distance is chosen as a measurement, then it puts forward the minimum classifier to complete the number recognition. Furthermore, according to the meaning represented by the ID Card number, as well as the generative algorithm of the check code, it can be used to verify the recognition result, in this way the recognition veracity could be improved to some extent. Experiments suggest that the algorithms are simple and easy to implement, the recognition effect is better.