On addressing driving inattentiveness: Robust rear light status classification using Hierarchical Matching Pursuit

摘要

Automatically recognizing rear light signals of front vehicles can significantly improve driving safety by automatic alarm and taking actions proactively to prevent collisions and accidents. Much previous research only focuses on the detection of brake signals at night. In this paper, we propose a novel and robust framework to detect rear lights of vehicles and estimate their signal states at daytime. Comparing with existing state-of-the-art works, our experimental results show the high detection rate and robustness of our system in complicated light conditions.