Haar-SVM for Real-time Banknotes Recognition

摘要

Haar wavelet is the simplest possible wavelet, and Support Vector Machine (SVM) is an effective classifier. This paper proposes a new method that combines the Haar wavelet and SVM for the first time to solve the problem of small denomination banknotes recognition with small computations, real timing and excellent performance. The key idea of this method is to extract the waveform features and transform them to the digitalize support vectors with fixed length. We illustrate our method on a dataset that comprises 220,000 samples of 2,200 different banknotes, and the approach achieves an average recognition rate of 99.9877%. The experimental results show our method is also simple, fast and robust. Our method has been applied in JBYD-8801A RMB-banknote counter.