Table of Contents Author Guidelines Submit a Manuscript
Mathematical Problems in Engineering
Volume 2014, Article ID 820203, 8 pages
Research Article

Video Pedestrian Detection Based on Orthogonal Scene Motion Pattern

School of Computer Science and Technology, Xidian University, Xi’an 710071, China

Received 29 April 2014; Accepted 12 June 2014; Published 8 July 2014

Academic Editor: Yuping Wang

Copyright © 2014 Jianming Qu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.


In fixed video scenes, scene motion patterns can be a very useful prior knowledge for pedestrian detection which is still a challenge at present. A new approach of cascade pedestrian detection using an orthogonal scene motion pattern model in a general density video is developed in this paper. To statistically model the pedestrian motion pattern, a probability grid overlaying the whole scene is set up to partition the scene into paths and holding areas. Features extracted from different pattern areas are classified by a group of specific strategies. Instead of using a unitary classifier, the employed classifier is composed of two directional subclassifiers trained, respectively, with different samples which are selected by two orthogonal directions. Considering that the negative images from the detection window scanning are much more than the positive ones, the cascade AdaBoost technique is adopted by the subclassifiers to reduce the negative image computations. The proposed approach is proved effectively by static classification experiments and surveillance video experiments.