Road safety in urban areas, and especially at pedestrian crossings, constitutes an important field of research, as these spaces are characterized by intense interaction between vehicles and vulnerable road users. This diploma thesis examines
inappropriate vehicle speed at signalized pedestrian crossings in Athens, using video data and computer vision techniques. The data were collected at a high-traffic intersection in Leoforos Vassileos Konstantinou through two parallel methods: on-site manual observation and a computer vision system combining YOLOv8, ResNet-50, homographic transformation, and Kalman filtering. Through data processing, motion variables and road safety indicators are extracted, such as Time to Collision (TTC) and Post-Encroachment Time (PET), in order to investigate their relationship with driver behavior. Statistical methods and regression models are then applied to identify the factors associated with the occurrence of inappropriate speed. The aim of the thesis is to achieve a better understanding of driver behavior near pedestrian crossings and to contribute to road safety assessment through modern analytical methods. The results may support the design of interventions to improve pedestrian safety in the urban environment.
inappropriate vehicle speed at signalized pedestrian crossings in Athens, using video data and computer vision techniques. The data were collected at a high-traffic intersection in Leoforos Vassileos Konstantinou through two parallel methods: on-site manual observation and a computer vision system combining YOLOv8, ResNet-50, homographic transformation, and Kalman filtering. Through data processing, motion variables and road safety indicators are extracted, such as Time to Collision (TTC) and Post-Encroachment Time (PET), in order to investigate their relationship with driver behavior. Statistical methods and regression models are then applied to identify the factors associated with the occurrence of inappropriate speed. The aim of the thesis is to achieve a better understanding of driver behavior near pedestrian crossings and to contribute to road safety assessment through modern analytical methods. The results may support the design of interventions to improve pedestrian safety in the urban environment.
| ID | ad185 |
| Presentation | |
| Full Text | |
| Tags |
