Banknote Fitness Classification Based on Convolutional Neural Network

The degree of defilement of the banknotes to some extent determines whether the banknotes can continue to circulate. How to accurately identify the dirtiness of banknotes is a major issue facing financial instruments. To solve this problem, we use a contact image sensor to collect double-sided reflection images of banknotes under red light, green light, blue light, and infrared light, and at the same time, collect images of banknotes under transmission of green light and infrared light. By using an image processing method to extract the banknote images, and then analysing the images formed by the various stain levels of the banknotes irradiated by various light sources, it is finally decided which kind of light source banknote images are input to the convolutional neural network. We process the classified training samples and test samples in the above manner, and we will get the training samples and test samples of the banknote images. Using the training sample to train our convolutional neural network, we will get the banknote dirt recognition classifier we need. Then use the test sample to test on this classifier, we will get the recognition effect of our trained classifier. The test results show that our designed convolutional neural network classifier is very accurate for identifying the dirtiness of banknotes.