Currency Image Processing Algorithm

Along with the rapid development of global economy,the amount of paper money and the amount of circulation become great.It is necessary to develop the technology of paper money recognition in financial institutions.According to the actual market demand,this thesis focuses on the recognition algorithm of note serial number,banknote sorting algorithm,and the main achievements are as follows:Firstly,we improve serial number recognition method based on uniform grid feature and SVM classifier,and increase the identification accuracy by more than 30%to 99%.Secondly,we implement two methods of currency serial number recognition.We achieve recognition accuracy rate higher than 98%through character recognition method based on Canny edge and LBP feature,and higher than 99%through character recognition method based on HOG feature and SVM classifier.Thirdly,we propose a note quality analysis algorithm based on multidimensional linear regression and improve a note defect detection algorithm based on Gaussian function.And the detection accuracy is greater than 90%.Fourthly,we tansplant the sorting algorithm based on one class SVM classifier on DSP.This algorithm can achieve accuracy higher than 90%on testing images acquired from the same machine with training images.