Multiple-shot People Re-identify based on Feature Selection with Sparsity

Dongping Zhang, Yanjie Li, Jiao Xu, Ye Shen
2015 International Journal of Hybrid Information Technology  
In a video surveillance network, it is always required to track and recognize people when they move through the environment. This paper presents a novel re-identification method for multiple-people using feature selection with sparsity. By using the multipleshot approach, each of appearance models is created in this method. The human body is divided into five parts form which the features of color, height, gradient were extracted respectively. Our appearance model is represented by linear
more » ... sion method. Experimental results show that our appearance model is robust and attain a high precision rate and processing performance. A Surveillance video is composed of a multi-frame image of each person. In order to achieve the purpose of re-identification, linear regression method is exploited to match each part of the body. Video preprocessing is necessary to detect the pedestrians, followed by further detailed analysis of the pedestrians.
doi:10.14257/ijhit.2015.8.1.03 fatcat:ojngbftu7rdj3iddjskwbjlbh4