Background modeling and foreground extraction scheme for HD traffic bayonet
Yixin Zhao, Di Wu, Jian Chen, Jian Wang · 2014
In a HD traffic bayonet, HD camera captures a few images once the vehicles drive cross the underground sensing coils. Time interval between two adjacent images is usually uncertain. In this paper, a background modeling and foreground extraction scheme is proposed for these images captured in the bayonet. The proposed scheme contains three modules: initial background modeling, foreground extraction and background update. It can quickly extract the foregrounds appear in the scene, including people, vehicles or other objects, which benefits to the classification, identification analysis, and storage of the data. The experiments prove that the scheme is efficient, with good universality, and not need training in advance.