Banknotes Multi-Spectral Image Analysis

Banknotes recognition is a popular issue in recent years, because it has a great future in application. The banknotes sorter developed by designer plays an important role in bank area. With the application of color sensors, due to the requirement of high speed of banknotes sorter, how to recognize paper currency under multi-spectral image is an important problem. The key technology of the classifier system is real-time image processing and image recognition. Banknote image analysis is of vital importance to the currency supervision.In this paper, the muti-spectrum images of banknote are systematically investigated in depth, The algorithms of banknote image analysis are first studied including fast banknote classification, detection of image details (defect feature or old degree feature).In this work , several strategies of improvement are provided according to the characteristics of multi-spectral image: In the multi-spectral image preprocessing stage, a new way is proposed to compensate the image and get the image at the same time. This method improves the speed of compensation and enhance the stability of acquisition. And use of the brightness component of HSI to do edge detection on the image. Because the image of banknote in the system may be deformation, a method is advanced to extract adaptive grid feature. According the traditional method, it has a better stability. And the use of color and geometry information to classify the banknote images improves the accuracy of classification denomination. According to question that in the actual system the distance in the species may be very big and the distance between the species may be very small, A way is advanced to combine the result of distance classifier and the result of artificial neural network. It has a higher recognition rate than the traditional method.In testing the old and new version of banknotes, a multi-spectral color image based on old and new identification methods, to make up for the gray image in the new and old test results when the vulnerability of the image sensor consistency distinguish differences and equipment during the operation of the image brightness changes. This paper advances a chromatic homogeneity based algorithm for banknote defect detection, and enhanced the detection capabilities. It is very effective to sort the version of RMB, in the method based on gradient of feature area.