Estimating Speeds of Pedestrians in Real-World Using Computer Vision

This paper proposes a novel approach to a computer vision based automatic system for the estimation of pedestrian velocity in real world traffic systems in which a fixed camera is available. The paper will introduce the adopted framework, which includes a preprocessing phase, an identification and tracking phase, and a speed estimation final phase. Speed estimation, implying a conversion from image to real world coordinates, can be carried out with two different techniques that will be discussed in details and evaluated with reference to achieved results.