Research On The Banknote Feature Recognition Technology Based On Image Processing

Nowadays rapid economic development has been driving the development of finance field, which led to the high improvement in the characteristics of the notes sorter. Recognition of paper notes feature based on image processing is the key technology in R&D (research and development) field of the notes sorter, and for this reason, it has been widen concerned and applied. Poor classification accuracy, restriction by external conditions and deficiency in deformed and defiled notes allocating pervade domestic recognition of paper notes feature based on image processing. This paper analysis and improve the recognition of currency image denomination, image-oriented, which make a myriad of preparation for pre-image recognition and get accurater result and faster recognition speed. Furthermore, with a view to notes sorter fail to the recognition of note number, image recognition aimed to note serial number recognition, which served as a recognition module will be mentioned in this paper.In the process of banknote feature recognition, domestic current fifth edition of the RMB serve as the original image, and base on RMB image gray feature signal, then apply digital image processing, pattern recognition and SOFM network to realize real-time identification.First, it gave a detail analysis of the two parts of note image: image enhancement preprocessing, which adopt histogram enhance, median filter and image sharpening; and image segmentation, which border detection and gray threshold segmentation. A great deal of work has been made in this process. Second, this paper analyzed the traditional algorithm which is based on dimension template matching, proposed and improve one-dimensional gray projecting algorithm. It shows new algorithm get faster recognition of the denomination of banknotes and SOFM network, which improved by this algorithm realize four image-oriented. It gets a satisfactory result from the recognition of letter and figure to BP neural network. At last, realize new and old, incomplete identification by the changing of histogram of different new and old, incomplete banknotes.The experimental result shows: banknote features recognition algorithm based on image processing satisfy the criterion of recognition and sorting technology, recognition accuracy rate is not less than 99%, and the favorable is accurate, reliable and real-time in the process of banknote recognition. This will make a progress on pushing the development of domestic sorter technology.