Abstract | ||
---|---|---|
Multi-camera person tracking requires the combination of high performance computation and efficient communication approaches in order to satisfy both the required accuracy and real-time processing requirements. In this paper, we present the initial results of our ongoing research project for a multi-target multi-camera tracking system. We propose a modular image processing pipeline comprised of background estimation, person detection, feature extraction, feature matching, and position estimation to track people between video frames, conscious of maintaining lower computation times and efficient interaction between multiple cameras. We present a weighted sequential k-means clustering approach to address the key challenge of feature matching for identifying/re-identifying individuals in an indoor environment. This approach is a form of computationally efficient online unsupervised learning suitable for meeting real-world requirements. Our results show that our approach has comparable accuracy in terms of assigning labels for person tracking, while achieving real-time computational requirements in an unsupervised manner. |
Year | DOI | Venue |
---|---|---|
2017 | 10.1109/IVCNZ.2017.8402479 | 2017 International Conference on Image and Vision Computing New Zealand (IVCNZ) |
Keywords | Field | DocType |
Person Tracking,Person Re-identification,Motion Tracking,Unsupervised Learning,Camera Surveillance | Computer vision,Histogram,Pattern recognition,Computer science,Tracking system,Image processing,Feature extraction,Unsupervised learning,Artificial intelligence,Modular design,Cluster analysis,Computation | Conference |
ISSN | ISBN | Citations |
2151-2191 | 978-1-5386-4277-1 | 0 |
PageRank | References | Authors |
0.34 | 0 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Andrew Y. Chen | 1 | 25 | 7.15 |
Jerry Fan | 2 | 0 | 0.34 |
Morteza Biglari-Abhari | 3 | 100 | 19.47 |
Kevin I-Kai Wang | 4 | 167 | 29.65 |