Detection and Number Plate Recognition of Non-Helmeted Motorcyclists using YOLO

Abstract

Motorcycles are the most common mode of transport as they are affordable and low-maintenance vehicles. Motorcyclists were roughly 29 times more likely than passenger car passengers to die in an accident per vehicle mile travelled in 2019. One of the leading causes of fatal motorcycle accidents is the rider’s failure to wear a helmet. According to section 129 of the motorcycle vehicle act, the Government has made it mandatory for two-wheeler drivers to wear helmets while driving. Still, many traffic rule violators do not obey them. In most developing countries, traffic police manually monitor motorcyclists at road junctions. Still, this method is inefficient as it does not apply on highways where the probability of accidents is highest due to speeding. This paper presents an automatic surveillance system for detecting two-wheeler drivers without helmets and recognizes their License plate numbers in the system. Firstly, the system detects motorcycles in the image or live video using the You Only Look Once (YOLO) algorithm. It again applies this algorithm to detect whether the driver is helmeted or not for the detected motorcycles. Finally, the motorcycle’s number plate is detected for identified motorcyclists without a helmet, and the characters are extracted using Optical Character Recognition.

Publication
TechRxiv (IEEE)
Dishant Padalia
Dishant Padalia
Computer Science Master’s Student at UMass Amherst

My research interests include computer vision, AI in healthcare and natural language processing.