Person Re-identification in Appearance Impaired Scenarios

Mengran Gou, Xikang Zhang, Angels Rates-Borras, Sadjad Asghari-Esfeden, Octavia Camps and Mario Sznaier

Abstract

Person re-identification is critical in surveillance applications. Current approaches rely on appearance-based features extracted from a single or multiple shots of the target and candidate matches. These approaches are at a disadvantage when trying to distinguish between candidates dressed in similar colors or when targets change their clothing. In this paper we propose a dynamics-based feature to overcome this limitation. The main idea is to capture soft biometrics from gait and motion patterns by gathering dense short trajectories (tracklets) which are Fisher vector encoded. To illustrate the merits of the proposed features we introduce three new �appearance-impaired� datasets. Our experiments demonstrate the benefits of incorporating dynamics-based information into re-identification algorithms.

Session

Posters 1

Files

Extended Abstract (PDF, 763K)
Paper (PDF, 935K)
Supplemental Materials (ZIP, 20M)

DOI

10.5244/C.30.48
https://dx.doi.org/10.5244/C.30.48

Citation

Bibtex

        @inproceedings{BMVC2016_48,
        	title={Person Re-identification in Appearance Impaired Scenarios},
        	author={Mengran Gou, Xikang  Zhang, Angels Rates-Borras, Sadjad Asghari-Esfeden, Octavia Camps and Mario Sznaier},
        	year={2016},
        	month={September},
        	pages={48.1-48.14},
        	articleno={48},
        	numpages={14},
        	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
        	publisher={BMVA Press},
        	editor={Richard C. Wilson, Edwin R. Hancock and William A. P. Smith},
        	doi={10.5244/C.30.48},
        	isbn={1-901725-59-6},
        	url={https://dx.doi.org/10.5244/C.30.48}
        }