Detecting Pedestrians and Vehicles on Urban Roads using Computer Vision

Abstract

This research explores the integration of advanced computer vision techniques with transportation studies for the detection of pedestrians and vehicles on urban roads in Cape Town. Using the YOLOv8 model, the study demonstrates commendable proficiency in object detection, emphasizing high recall and precision. Valuable insights into pedestrian road-crossing dynamics are unveiled, particularly the identification of double-gap strategies. Pedestrians' deliberate decisions to wait until all lanes are clear reveal a nuanced understanding of traffic flow and safety considerations. The findings underscore the potential of computer vision technologies to enhance pedestrian safety, contribute to traffic management, and identify anomalies, such as vehicles driving on the wrong side. The research signifies a notable advancement in leveraging technology for optimizing urban transportation systems and addressing critical challenges in contemporary urban landscapes.