RMB Code Recognition Based On Image

The uniqueness of the RMB code decides it could be as the identification of RMB, and then monitoring the flowing and using of RMB, which make the RMB circulation market to be more security and standardization. In this paper, the contents of the study can be divided into three parts, which are preprocessing, segmentation positioning and identification.Analyzing the content of the image preprocessing, taking into account the particularity of the RMB image preprocessing, the edge detection algorithm of slope linear fitting method combining grouping average method is used, which is simple and accurate. In the image geometric distortion correction, not only the image tilt correction is considering, while also the image shear correction is adopted, making the system more adaptable.Research the algorithm of RMB code location and segmentation. A location and segmentation method is adopted, which is preliminary rough location, then fine positioning, and finally fine segmentation. Rough positioning algorithm is implemented by fixed ratio algorithm presented, this algorithm uses the results of edge detection and geometric correction, which is simple, less computation, strong adaptability, and will not be impacted by the changing of image size. Fine positioning and fine segmentation are implemented by X axis projection method and Y axis projection method.Code recognition section is divided into two parts, namely, digital recognition and English character recognition. Digital recognition part is implemented by two kinds of feature extraction algorithm, intersection variation algorithm and improved eight-neighbor intersection algorithm. Intersection variation algorithm is a simple comparison operation, which has high recognition rate and short recognition time for the RMB image without large tilt. Improved eight-neighbor intersection algorithm overcomes some shortcomings of intersection variation algorithm in large tilted image recognition, making the system have a strong adaptability. For English character recognition, we offered PNN algorithm based on FICA, the algorithm is superior to other neural networks algorithm considering the accuracy and rapidity for pattern classification.