Clustering Method for Counting Passengers Getting in a Bus with Single Camera

Tao Yang,Yanning Zhang,Dapei Shao,Ying Li

DOI: https://doi.org/10.1117/1.3374439

IF: 1.3

2010-01-01

Optical Engineering

Abstract:Automatic counting of passengers is very important for both business and security applications. We present a single-camera-based vision system that is able to count passengers in a highly crowded situation at the entrance of a traffic bus. The unique characteristics of the proposed system include, First, a novel feature-point-tracking- and online clustering-based passenger counting framework, which performs much better than those of background-modeling-and foreground-blob-tracking-based methods. Second, a simple and highly accurate clustering algorithm is developed that projects the high-dimensional feature point trajectories into a 2-D feature space by their appearance and disappearance times and counts the number of people through online clustering. Finally, all test video sequences in the experiment are captured from a real traffic bus in Shanghai, China. The results show that the system can process two 320 x 240 video sequences at a frame rate of 25 fps simultaneously, and can count passengers reliably in various difficult scenarios with complex interaction and occlusion among people. The method achieves high accuracy rates up to 96.5%. (C) 2010 Society of Photo-Optical Instrumentation Engineers. [DOI: 10.1117/1.3374439]