Research On Image Scoring Algorithm Of Intelligent Point Detector

With the popularity of automated banking machines and economic development, smart banknote counter has large number of applications in banks and other financial institutions. Sorting method of smart banknote counter is also in constant development, most started with a voice recognition method, magnetic recognition method and so on, and now more and more using image based method. How to use the images for efficient sorting banknotes is a very important issue, it has great theoretical and practical significance.This paper studies the smart banknote counter image clearing algorithm. This paper is mainly about value recognition, face recognition, recognition of old and new, damaged identification, crown word number identification, which also uses a secondary color image sorting. The main work is as follows:1) Banknote image preprocessing. Including banknote image binarization based on histogram equalization image enhancement, after scanning up and down about the use of the Hough transform boundary line detection, correction and extraction zone banknotes image tilt, bilinear interpolation scale normalization banknote.2) Banknote value recognition and face recognition.In the face value recognition, analysis of several common banknotes recognition method, proposed a bill based on the length of the range for the face value of recognition algorithm, and also proposed a formula based on the color image color banknotes recognition methods, and the experiment. In face identification, notes there are four in the image-oriented, mainly:being put positive; positive upside down; being put negative; negative upside down. In this paper, the bill mesh feature extraction, and then use SVM (Support Vector Machine) method on the bill face identification.3) The identification of old and new banknotes and damaging smear identified. In recognition of old and new, we use edge intensity histogram to quickly identify new and old banknotes, specifically to determine the mean intensity banknote image edge, and then find the edge intensity histogram with the old template image edge intensity histogram correlation coefficient, then combining these two data to determine the old and new paper money. On the damaged transmission image recognition is used to select a threshold calculating damaging rate after binarization. In addition to this paper, the use of a scanner to scan a color image on banknotes defaced identification was attempted.4) Notes crown word number recognition. In this paper, the horizontal projection numbers were down crown word boundary detection, the use of the vertical projection of the number of characters were crown word segmentation, using the KNN (K nearest neighbor method) notes crown word number of character recognition.5) Image sorting online debugging subsystem design and implementation. For the above proposed algorithm, we design and implement image sorting online debugging subsystem, respectively from the hardware design, software design, image acquisition format and the realization of the system are discussed.