Detection of Motorcycle Headlights Using YOLOv5 and HSV

Abstract

"Electronic Traffic Law Enforcement" (ETLE) denotes a mechanism that employs electronic technologies to implement traffic regulations. This commonly entails utilizing a range of electronic apparatuses like cameras, sensors, and automated setups to oversee and uphold traffic protocols, administer fines, and enhance road security. ETLE systems are frequently utilized for identifying and sanctioning infractions like exceeding speed limits, disregarding red lights, and turning off the headlights. In Indonesia, there is currently no dedicated system designed to detect traffic violation, especially regarding vehicle headlights. Therefore, this research was conducted to detect vehicle headlights using digital images. With the results of this study, it will be possible to develop a system capable of classifying whether vehicle headlights are on or off. This research employed the deep learning method in the form of the YOLOv5 model, which achieved an accuracy of 94.12% in detecting vehicle images. Furthermore, the white color extraction method was performed by projecting the RGB space to HSV to detect the Region of Interest (ROI) of the vehicle headlights, achieving an accuracy of 73.76%. The results of this vehicle headlight detection are influenced by factors such as lighting, image capture angle, and vehicle type.