Folding Paper Currency Recognition and Research Based on Convolution Neural Network

With the rapid development of the internet financial era, automatic financial transaction equipment has become the indispensable tools for life. Aiming at the defects and deficiencies in the recognition function of folded paper currency of automatic financial transaction equipment, this paper proposes that the convolution neural network should be applied to the identification of folded paper currency, reference to banknote authenticity identification and damage criteria, specification the angle of folded banknote, manual collection of Ukrainian banknotes folded image, damaged image and normal image as experimental dataset. using the OpenCV library to preprocess banknote images, using the Keras framework to construct the nine-layer convolution neural network, compared with the seven-layer LeNet-5 network, the recognition accuracy of the nine-layer convolution neural network is better than that of the sevenlayer network, reaching 96.46%.