New and Old Banknote Recognition Based on Convolutional Neural Network

Recognition for new and old currency is a key function of the paper currency sorter. How to discriminate unfitness banknotes which became rough and fuzzy, even be damaged is an important task in financial institution. Different from traditional fitness banknote recognition based on extracting feature manually, a method based on convolutional neural network was proposed to identify fitness banknotes in this paper. Firstly, we preprocess Ukrainian banknote image and train letnet-5 model to identify fitness and unfitness currency. Secondly, after optimizing the network structure from the network layer and convolutional kernel size, we determine the best structure and performance parameters. Finally, compared with the traditional fitness banknotes recognition methods, optimized structure achieves higher recognition rate. It owes better result to combining multiple features such as holes, stains and so on. In a word, the method proposed has considerable advantages in accuracy.

[1]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[2]  Satoshi Naoi,et al.  Beyond human recognition: A CNN-based framework for handwritten character recognition , 2015, 2015 3rd IAPR Asian Conference on Pattern Recognition (ACPR).

[3]  Kang Ryoung Park,et al.  Efficient Banknote Recognition Based on Selection of Discriminative Regions with One-Dimensional Visible-Light Line Sensor , 2016, Sensors.

[4]  Chulhee Lee,et al.  Efficient multi-currency classification of CIS banknotes , 2015, Neurocomputing.

[5]  Baiqing Sun,et al.  The Recognition of New and Old Banknotes Based on SVM , 2008, 2008 Second International Symposium on Intelligent Information Technology Application.

[6]  Jan-Mark Geusebroek,et al.  Learning banknote fitness for sorting , 2011, 2011 International Conference on Pattern Analysis and Intelligence Robotics.

[7]  Shutao Li,et al.  A Classification Method for the Dirty Factor of Banknotes Based on Neural Network with Sine Basis Functions , 2008, 2008 International Conference on Intelligent Computation Technology and Automation (ICICTA).

[8]  Ching Y. Suen,et al.  Multi-class SVM based gradient feature for banknote recognition , 2017, 2017 IEEE International Conference on Industrial Technology (ICIT).

[9]  Kang Ryoung Park,et al.  A High Performance Banknote Recognition System Based on a One-Dimensional Visible Light Line Sensor , 2015, Sensors.

[10]  Jan-Mark Geusebroek,et al.  Brain2: Machine Learning to Measure Banknote Fitness , 2012 .

[11]  Dumitru Erhan,et al.  Going deeper with convolutions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[12]  Sigeru Omatu,et al.  Neuro-Classification of Currency Fatigue Levels Based on Acoustic Cepstrum Patterns , 2000, J. Adv. Comput. Intell. Intell. Informatics.

[13]  Kang Ryoung Park,et al.  Recognition of Banknote Fitness Based on a Fuzzy System Using Visible Light Reflection and Near-infrared Light Transmission Images , 2016, Sensors.

[14]  Tien Dat Nguyen,et al.  Recognizing Banknote Fitness with a Visible Light One Dimensional Line Image Sensor , 2015, Sensors.

[15]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[16]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[17]  Kang Ryoung Park,et al.  A Survey on Banknote Recognition Methods by Various Sensors , 2017, Sensors.

[18]  Jürgen Schmidhuber,et al.  Deep learning in neural networks: An overview , 2014, Neural Networks.

[19]  Kilian Q. Weinberger,et al.  Densely Connected Convolutional Networks , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).