The Crown Of The Number Of The Words Image Recognition Technology Research

RMB is the legal currency of China, because of the characteristics that each banknote has a unique crown word number, it can serve as the identity of the paper currency. Through the establishment of the banknotes’number sequence file, the financial departments can track the circulation of banknotes to achieve effective regulatory. Therefore, the establishment of automatic identification system of RMB crown word number is of far-reaching significance, and has a wide application prospect.In this paper, according to the characteristics of the crown word number images, combined with the theory knowledge in digital image processing and pattern recognition, it implements the algorithms as follows:the preprocessing of RMB images, the character segmentation of crown word numbers, the feature extraction and the character recognition. What’s more, it sets up an automatic RMB crown word number identification system. The main contents are as follows:(1) In the image preprocessing module of this paper, it puts forward the percentage threshold binarization method,as a result, the image processing effect is greatly improved; On account of the features of the perimeter and area of the noises in binary images, this paper adopts the chain code tracking algorithm to eliminate the interference of the noises. After that, it locates the crown word number and tilts correction for characters.(2) In the module of character segmentation for crown word numbers of this paper, it presents the method of fusing the binary images and grayscale images for image segmentation, which makes up for the deficiency of the character segmentation method that is only based on binary image. What’s more, it detailedly analyzes various situations that may effect the correctness of the character segmentation, improves the segmentation method and the segmentation effect is very good.(3) In the character feature extraction module, based on the eight directions for gradient feature extraction, this paper makes the linear transformation processing on the gray images, which effectively avoids the interference of background noise and improves the character recognition.(4) In the character recognition module, respectively, based on the nearest neighbor and support vector machine (SVM) classifier, it compares the experimental data and analyzes the cause of errors, and proves the effectiveness of the improved method in this paper.