This paper presents an innovative and highly effective method for vehicle detection, specifically tailored for operation during night-time conditions when visibility is limited. The core of our approach combines a sophisticated shape filter with a reflector removal scheme, allowing for precise identification of vehicles in various lighting scenarios.
To evaluate the performance of our method, we conducted thorough simulations, capturing over five hours of continuous on-road video footage. This extensive dataset includes a wide array of vehicle speeds, ranging from complete stops to fast-moving traffic at speeds of up to 100 km/h. To facilitate comprehensive analysis and testing, we categorized the captured video footage into three distinct environmental settings: urban areas, which often feature dense traffic and varying light conditions; rural landscapes, where vehicles may encounter less traffic and more open spaces; and high-speed highways, where vehicles generally travel at higher velocities.
The results of our proposed algorithm are notably impressive, demonstrating high vehicle detection rates across all tested environments. Specifically, we achieved a detection accuracy of 97.3% in urban settings, 98.4% in rural environments, and 97.6% on highways. These results underscore the robustness and reliability of our method, highlighting its ability to adapt to the challenges of nighttime vehicle detection. Our findings suggest significant potential for improving safety and enhancing traffic monitoring systems in low-visibility conditions.
Nighttime Vehicle Detection, Adaptive Threshold Method, Reflector Removal Method, Adaptive Headlamps Control