Abstract: RMB image pre-processing and serial number identification are achieved.The sequential minimal optimization algorithm of support vector machine(SVM) in statistical learning theory is mainly studied.The support vector machine is used in serial number identification,and it solves the problems of seared samples,nonlinearity and high dimensions in serial number identification on RMB banknote.The experimental result indicates that this money recognition method based on SVM has higher feasibility and identification precision than the simple BP(backpropagation) neural network.