Evaluating defensive driving behaviour based on safe distance between vehicles: A case study using computer vision on UAV videos at urban roundabout

Abstract

While driving, maintaining a sufficient distance helps reduce collision risk. A time gap of two or three seconds on urban roads from a vehicle ahead is advised in defensive driving. The scenario becomes even more challenging in densely populated and developing countries because of limited road infrastructure, lane indiscipline, and heterogeneous traffic. The safe distance between vehicles and the driver’s reaction can be used as surrogate safety measures (SSMs) to evaluate defensive driving behaviour. This paper presents a case study evaluating defensive driving behaviour using the vision-based methodology and UAV video. This paper proposes two novel SSMs based on distance and acceleration and studies defensive driving behaviour, such as “for how long did a vehicle keep driving under another vehicle’s blind spots?” and “how is a vehicle driving (an interaction pattern) when another vehicle ahead is in its stopping distance range?.” Finally, each driver’s star rating depends on their interactions with other vehicles. We observed that around 48 % of the vehicles did not follow defensive driving practices. In our vehicle inter-class interaction analyses, we also found 16.6 % Rear-End, 6.3 % Side-Swipe, and 1.5 % Angled collision risks occurred between car-car, car-car, and 2Wheeler(2W)-car, respectively. Our methodology could help traffic law enforcement agencies and policy-makers elevate road traffic safety by taking counter-measures against the low-star vehicle categories in developing countries. Example videos of star rating are available on https://www.youtube.com/@YagnikBhavsar.