Parallel algorithm implementation for multi-object tracking and surveillance

The objective of this paper is to propose a parallel implementation of dictionary learning based multiperson tracking on single camera. After its success of face recognition, dictionary learning is recently adapted to multi-person tracking and achieves very promising results [2]. However, by recognizing that the computational complexity of dictionary learning is a big factor preventing tracking algorithm to operate in real-time, especially when the size of dictionary increases over time, this paper proposes parallel algorithms, implemented on GPU, to try speed up the system.