A Practical and Automated Image-based Framework for TrackingPedestrian Movements from a Video

Video tracking of pedestrian movements can be used to gain better understanding of crowd features and behaviors as nowadays pedestrian safety is a very important consideration in many situations. The existing works on video tracking have some limitations. For example some works focus on a specific event and a specific place, and some require a lot of human intervention. In this paper, a practical and automated image-based framework for tracking pedestrian movements from a video is presented. The proposed framework consists of several steps such as detection, tracking and extracting characteristics of a pedestrian from a video. Consecutive frames in the video are processed to extract moving objects of interest which are assumed to be pedestrians. The extracted data obtained are filtered to remove unwanted objects and then objects are labelled for identification. These steps are automated and less human efforts are needed. The centroid positions of all objects are composed to obtain the movement vectors which are used to plot a graph and visualize the movement path of a pedestrian. Besides, it is also possible to estimate the speed of a pedestrian. The results of the preliminary experiment for the proposed framework using three videos with different scenarios are presented in the paper. The pedestrian movement is plotted accurately and the maximum and minimum numbers of pedestrians in the video are recorded correctly. However the speed of the pedestrian is slightly inaccurate.