Open Access

Sara Moridpour, Md. Shohel Parvez, Richard Tay, Amir Shtayat, Berthold Best

• Measuring pedestrian crossing time and speed using advanced techniques can enhance traditional road safety practices. This study aims to estimate the time and speed required for pedestrians to cross an intersection by employing computer vision applications to analyse video data and measure the deviation between the required crossing time and the allotted pedestrian signal cycle. The YOLOv11 framework with the COCO dataset, combined with Deep SORT, was implemented to accurately detect and track pedestrians from video data. Results show that the integration of the YOLOv11 and Deep SORT algorithms offers an effective approach for evaluating pedestrian speeds and crossing durations at intersections. The detection model demonstrates high accuracy in identifying pedestrians, while the integration of the Deep SORT tracker during the object tracking stage yields excellent results, with maximum confidence scores exceeding 0.9. The overall accuracy of the speed estimation is highly satisfactory, consistently yielding a percentage error of lessMeasuring pedestrian crossing time and speed using advanced techniques can enhance traditional road safety practices. This study aims to estimate the time and speed required for pedestrians to cross an intersection by employing computer vision applications to analyse video data and measure the deviation between the required crossing time and the allotted pedestrian signal cycle. The YOLOv11 framework with the COCO dataset, combined with Deep SORT, was implemented to accurately detect and track pedestrians from video data. Results show that the integration of the YOLOv11 and Deep SORT algorithms offers an effective approach for evaluating pedestrian speeds and crossing durations at intersections. The detection model demonstrates high accuracy in identifying pedestrians, while the integration of the Deep SORT tracker during the object tracking stage yields excellent results, with maximum confidence scores exceeding 0.9. The overall accuracy of the speed estimation is highly satisfactory, consistently yielding a percentage error of less than 2%. In addition, by comparing the required crossing time with the pedestrian signal cycle time, the study found that some pedestrians face a risk during each signal cycle while crossing the intersection. Finally, this study recommends an optimised pedestrian green time for intersection crossing, offering a new perspective on movement strategies in traffic management. The proposed method in this study significantly reduces the reliance on labour-intensive manual observations and data analysis in future research on pedestrian crossing behaviour.…