| Currency recognition is a pattern recognition problem, which includes following three main parts, Data acquisition, Feature extraction and Classifier design. To improve the rate of currency recognition, based on the ideas of Canonical correlation analysis (CCA) and Geometric Interpretation of Support vector machine (SVM), after fusing two different kinds of Principal component analysis(PCA), the method based on Fuzzy support vector data description (FSVDD) was proposed to recognize the currency.To get more comprehensive information of feature extraction for currency recognition, based on the ideas of Rough set approach for attribute reduction and Canonical correlation analysis (CCA),after using two different kinds of Principal component analysis(PCA) to reduce the Dimension of paper features, attribution reduction was done based on the link between condition attributes and decision attributes. Then, the two different features were fused. The experimental results show that the method can effectively fuse different features. It is superior to previous algorithm with single PCA, and the training time is just the same.A new fast algorithm of SVM based on Geometry was proposed, the main idea was formed by The Nearest point algorithm (NPA) and DIRECTSVM algorithm. This approach with the low training time can reduce the large computation of Quadratic programming. The experimental results show that the performance of algorithm is high, and the training time is lower than SMO and DIRECTSVM.A multi-classification method based on Fuzzy support vector data description (FSVDD) was proposed to recognize the currency by its value. Especially, fuzzy membership operator was calculated by the tightness of points, and the proper penalty factor was chosen to reduce the imbalance of samples. The experimental results show that this approach can not only improve the currency recognition rate, but also reduce the training time and the effect of noise point than other SVM multi-classifications. |
PDF Full Text Request