The Research Of ID Card Recognition System

The ID card is an important certificate for the adults above 18 years of our country. It has the holder's basic personal information and has a unique ID card serial number. Since most of the ID cards are registered by manual now, a more convenient method is expected. With the progress of computer science, image processing and character recognition, it has been possible for us using computer to identify the information in ID card image automatically. The ID card recognition system can be applied in many departments, such as transportation, public security office etc. Since it can save large quantity of manpower, it has vast of application foreground.In this thesis, the ID card recognition system based on a simplification model is investigated. It includes the detail process: line segmentation, image preprocess, recognition preprocess, feature extraction, recognition and the serial number's after-recognition. To fit the special character recognition, preprocess sequence is adjusted as: line segmentation under the gray level image, image enhance, binary, character segmentation. We provide the algorithm of line segmentation under gray level image and character segmentation based on the biggest width backtracking algorithm to get the accurate position of characters. Gray image is converted to binary one by suitable threshold. A 31-dimension feature vector is selected which can best resemble the digit's characteristic of construction and statistics. Gross and fine classification methods are used for Chinese character recognition, respectively with a 4- dimension and 72-dimension vector. Factor analysis method is applied on the 31-dimension vector to get independent variables. It decreases the relations among the variables, but also changes their weights to get the bigger variance among different classes, which benefits for the recognition. We process the after-recognition for the serial number with pre-knowledge from the ID card. So we can know whether the number recognized is correct, and revise some wrong decisions. Experiments verify the validation of the algorithms.