Research On Recognition Algorithm In Dynamic Detection Of Paper Currency

Research on image recognition in dynamic detection of paper currency is a more popular research direction in the field of pattern recognition in recent years. It has very practical applications, notes sorter based on this technology plays an important role at various financial institutions.Image recognition technology is one of the key technologies in notes sorter. Notes sorter mainly implements the sorting of paper currency in special version such as denomination, face, the degree of quality and impairment. This is an inevitable requirement for higher reliability and real-time. At present, the domestic notes sorter exist to classification of inconsistent and poor ability to deal with deficiencies in practical applications. The reason is that image recognition technology in the extension on the application of immature and mismatching. In this paper, we research and improve the problems in image recognition technology in dynamic detection of paper currency. That will make a inspiration and meaning in R&D and application of domestic notes sorter.In the research of recognition technology, the objects are the current circulation RMB of the fifth set in 05 version, to gray degree of image as a sample, using a certain degree of image processing techniques and pattern recognition techniques to achieve dynamic detection of note image recognition, the method used has strong portability.Pretreatment, the use of improved edge maintenance filtering method to remove the noise. The use of six-line access-point method for edge detection and positioning, and tilt correction. Tilt correction method has been simplified and improved, achieved some results.Impairment detected research, mathematical morphology-based analysis methods be used for the integrity of the overall geometry. Principal component analysis and the combination of Fisher discriminant criteria and the linear discriminant based on "characteristic distance" be used for the coupon pictorial clarity.Through simulation experiments show that the notes image recognition technologies and methods in this paper be to satisfy the note the technical requirements for dynamic detection and has practical application value.