A Greedy Data Association Technique for Multiple Object Tracking
Gurinderbeer Singh, Sreeraman Rajan, Shikharesh Majumdar · 2017
With recent advancements in complex image analysis algorithms and global optimization techniques, the qualitative performance of Multiple Object Tracking (MOT) has improved significantly, at the cost of slow processing speed. With a focus on high-speed performance, in this paper, we propose a fast data association technique for tracking multiple objects by using Tracking-by-Detection paradigm. Followed by a pre-processing stage of creating reliable tracklets from given detection responses, we propose a threshold-based greedy algorithm that iteratively finds a locally optimum solution with significantly low computational overhead. Experiments conducted on two benchmark datasets show that our method is able to achieve qualitative results comparable to the existing state-of-the-art algorithms with an advantage of 50-600 times faster processing speed.