Riding a motorcycle without a helmet can cause serious injury. Although a crackdown on traffic violations is a good way to stop this unsafe practice, it is not realistic to manually find and arrest riders who do not wear helmets where there are numerous motorcycle riders, as in Vietnam. In consideration of this situation, we developed an automatic detection system for riders who are not wearing a helmet using deep learning. The proposed method’s accuracy, precision, recall, and F-measure in classifying motorcyclists into helmeted and non-helmeted are 0.966, 0.957, 0.936, and 0.946, respectively. The quality of the classification was higher than in previous work which did not use deep learning. As with other image-processing systems using deep learning, our system achieved state-of-the-art performance. This system will reduce not only the number of motorcycle riders not wearing a helmet, but also the manual work of arresting illegal riders.
CITATION STYLE
Hirota, A., Tiep, N. H., Van Khanh, L., & Oka, N. (2017). Classifying helmeted and non-helmeted motorcyclists. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10261 LNCS, pp. 81–86). Springer Verlag. https://doi.org/10.1007/978-3-319-59072-1_10
Mendeley helps you to discover research relevant for your work.