Advancements in Motorcycle Safety: Automatic Helmet Detection Systems Through Machine Learning

Abstract

Motorcycle accidents are a huge concern today. The number of accidents occurring every day has increased rapidly which in turn results in loss of lives. One of the main factors in these fatal accidents is the lack of use of helmets. Even though the use of a helmet is mandatory people fail to follow it which causes these unfortunate events. The government has taken various initiatives regarding the increase in the use of helmets by creating awareness and also increasing the fines but people seem to be least bothered. To tackle these situations, we propose an automated solution for helmet detection on motorcyclists using a combination of techniques like YOLO and CNN. The combination of You Only Look Once (YOLO) and Convolutional Neural Network (CNN) helps to efficiently identify the helmets from the real-time datasets. The proposed system will help traffic police by reducing their work and automating the imposition of fines. As the techniques are used in combination, the accuracy and the degree of reliability of our system are higher compared to single techniques-based implementations. Key Words: Open CV-Open-Source Computer Vision Library ,CNN-Convolutional Neural Network ,OCR-Optical Character Recognition ,YOLO-You Only Look Once ,IOU-Intersection over Union ,DHCP-Dynamic Host Configuration Protocol ,DDNS-Dynamic Domain Name System ,PoE-Power over Ethernet ,API-Application Programming Interface.