TY - GEN
T1 - Real-Time Multi-person Multi-camera Tracking Based on Improved Matching Cascade
AU - Guo, Yundong
AU - Wang, Xinjie
AU - Luo, Hao
AU - Pu, Huijie
AU - Liu, Zhenyu
AU - Tan, Jianrong
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2022
Y1 - 2022
N2 - In a small-scale distributed multi-camera system like video surveillance system of a museum, shopping mall, plaza, etc., or advanced driving assistance system, real-time multi-person tracking is essential for public and pedestrian safety consideration in the smart security system. In this paper, a real-time multi-person multi-camera tracking framework is presented, which is compatible with both overlapping and non-overlapping views. Since cameras have different orientations and exposures, false matching occurs frequently when people cross the camera boundaries or reenter the same camera. To deal with this challenge, an improved multi-person multi-camera matching cascade scheme is proposed, which can increase the accuracy of inter-camera person re-identification (Re-ID) by taking advantage of association priorities of targets and features. Besides, the proposed method can deal with the occlusion of people and variation of appearance features. Experiments are implemented with overlapping and non-overlapping videos, and results show that the proposed method has robust performance in different situations.
AB - In a small-scale distributed multi-camera system like video surveillance system of a museum, shopping mall, plaza, etc., or advanced driving assistance system, real-time multi-person tracking is essential for public and pedestrian safety consideration in the smart security system. In this paper, a real-time multi-person multi-camera tracking framework is presented, which is compatible with both overlapping and non-overlapping views. Since cameras have different orientations and exposures, false matching occurs frequently when people cross the camera boundaries or reenter the same camera. To deal with this challenge, an improved multi-person multi-camera matching cascade scheme is proposed, which can increase the accuracy of inter-camera person re-identification (Re-ID) by taking advantage of association priorities of targets and features. Besides, the proposed method can deal with the occlusion of people and variation of appearance features. Experiments are implemented with overlapping and non-overlapping videos, and results show that the proposed method has robust performance in different situations.
KW - Appearance Features
KW - Matching Cascade
KW - Multi-person Multi-camera Tracking
KW - Real-time
UR - https://www.scopus.com/pages/publications/85126184562
U2 - 10.1007/978-981-16-8048-9_19
DO - 10.1007/978-981-16-8048-9_19
M3 - 会议稿件
AN - SCOPUS:85126184562
SN - 9789811680472
T3 - Smart Innovation, Systems and Technologies
SP - 199
EP - 209
BT - Advances in Intelligent Systems and Computing - Proceedings of the 7th Euro-China Conference on Intelligent Data Analysis and Applications
A2 - Zhang, Jie-Fang
A2 - Chen, Chien-Ming
A2 - Chu, Shu-Chuan
A2 - Kountchev, Roumen
PB - Springer Science and Business Media Deutschland GmbH
T2 - 7th Euro-China Conference on Intelligent Data Analysis and Applications, ECC 2021
Y2 - 29 May 2021 through 31 May 2021
ER -