Noisy QR Code Smart Identification System

Abstract

The resurrection of the quick-response (QR) code has been made possible by the expansion of mobile network coverage combined with a rise in smartphone online content over the years. They have become much more accessible by integrating a code reader in smart devices, thus removing several unpleasant procedures and providing faster access to crucial information. However, noise in the printed images is unavoidable owing to printer processes and restricted printing technology, thus may decrease the quality of a QR code image during digital image collection and transmission which may eventually cause failure while scanning and extracting actual information. As a result, this study proposes an intelligent image classification strategy to correctly identify noisy and original QR code types. For this, a new dataset is built, containing 20,000 images pertaining to the original QR code and noisy QR codes. Later, the study exploited three well-known machine learning algorithms (logistic regression (LG), support vector machine (SVM), and convolutional neural network (CNN)) to segregate noisy images among original QR code images. The experimental results show that SVM outperformed others by attaining an overall performance accuracy of 97.5%, precision of 97.50%, recall of 97.5%, and F1-score of 97.5%, while LG almost competes by achieving 97.25% accuracy, 97.31% precision, 97.22% recall, and 97.25% F1-score.