作者: Loris Bazzani , Marco Cristani , Vittorio Murino
DOI: 10.1007/978-1-4471-6296-4_3
关键词:
摘要: In video surveillance, person re-identification (re-id) is probably the open challenge, when dealing with a camera network non-overlapped fields of view. Re-id allows association different instances same across locations and time. A large number approaches have emerged in last 5 years, often proposing novel visual features specifically designed to highlight most discriminant aspects people, which are invariant pose, scale illumination. this chapter, we follow line, presenting strategy three important key-characteristics that differentiate it respect state art: (1) symmetry-driven method automatically segment salient body parts, (2) an accumulation making descriptor more robust appearance variations, (3) procedure casted as image retrieval problem, can be easily embedded into multi-person tracking scenario, observation model.