A robust deep learning-based system for pedestrian-aware collision prevention in autonomous vehiclesOA
Pedestrian collisions remain a critical concern in road safety research due to pedestrians’heightened vulnerability and the substantial impacts on road networks and public healthcare systems.These incidents often lead to significant traffic disruptions,emphasizing the importance of pedestrian-focused road design.In dynamic urban environments,understanding street layouts and pedestrian behavior is essential for improving safety.In response to these challenges,this research presents an innovative and efficient automated collision avoidance system,leveraging an enhanced YOLOv9 architecture to achieve high accuracy and real-time performance.The proposed system is designed to detect and predict potential pedestrian collisions,while delivering proactive accident warnings and autonomously initiating the automatic emergency braking(AEB)mechanism when required,thereby effectively mitigating collision risks.By analyzing key behavioral and environmental factors including pedestrian exposure,jaywalking,intersection crowding,and hazardous turning locations,the system employs a proactive methodology aimed at enhancing pedestrian safety.Experimental evaluations using the KITTI and CityPersons datasets demonstrate the system’s efficacy,achieving precision rates of 98.79%and 96.62%,respectively,and mAP@0.5 scores of 96.21%and 94.83%.These results underscore the system’s capability to deliver high accuracy,fast inference,and proactive accident mitigation across diverse urban scenarios.
Wajdi Farhat;Marwa Guizani;Olfa Ben Rhaiem;Radhia Zaghdoud;Hassene Faiedh;Chokri Souani
Higher Institute of Applied Sciences and Technology,Sousse University,Sousse,Tunisia Electronics and Microelectronics Laboratory,Faculty of Sciences,University of Monastir,Monastir,TunisiaHigher Institute of Applied Sciences and Technology,Sousse University,Sousse,Tunisia Electronics and Microelectronics Laboratory,Faculty of Sciences,University of Monastir,Monastir,TunisiaDepartment of Computer Science,College of Science,Northern Border University,Arar,Saudi ArabiaDepartment of Computer Science,College of Science,Northern Border University,Arar,Saudi ArabiaHigher Institute of Applied Sciences and Technology,Sousse University,Sousse,Tunisia Electronics and Microelectronics Laboratory,Faculty of Sciences,University of Monastir,Monastir,TunisiaHigher Institute of Applied Sciences and Technology,Sousse University,Sousse,Tunisia Electronics and Microelectronics Laboratory,Faculty of Sciences,University of Monastir,Monastir,Tunisia
交通工程
Advanced driver-assistance systems(ADAS)Automatic emergency braking(AEB)Deep learningCollision avoidanceYOLOv9
《International Journal of Transportation Science and Technology》 2026 (2)
P.447-469,23
评论