Journal "Software Engineering"
a journal on theoretical and applied science and technology
ISSN 2220-3397

Issue N5 2014 year

Tracking Pedestrians for Video Surveillance for Non-Signalized Pedestrian Crossings
M. A. Kolosovskiy , e-mail: maxim.astu@gmail.com

This paper considers the task of video surveillance for pedestrian crossings. We analyzed peculiarities of the task, proposed the structure of the video surveillance system, and described the developed tracking subsystem. We evaluated accuracy, performance and robustness of the system. The precision of tracking is about 92 %, recall — 75...95% and its speed is close to real time mode.

Keywords: video surveillance, object tracking, object detection, computer vision
pp. 32–40