An Experimental Comparison of Swarm Optimization Based Abrupt Motion Tracking Methods

In view of the problem that the traditional tracker does not adapt to the large displacement or abrupt motion well, the optimization method attracts more and more attention for a robust tracker. Considering population based has high feasibility for avoiding local optimal, the method of swarm optimization is introduced to visual tracking. These methods convert visual tracking to search the optimal solution in global. To show their merits, this paper reviews and evaluates three relatively classical swarm optimization-based tracking algorithms. These algorithms are ant lion optimization, cuckoo search, and particle swarm optimization. Their running results are compared with those of the probabilistic optimization algorithm, namely, simulated annealing. The experiments demonstrate the strength as well as the weakness of these methods. For illustrating their operational efficiency, run time is recorded and the convergence speed is analyzed. In addition, quantitative and qualitative analysis experiments are performed to interpret the accuracy of the tracking methods. In addition, the relation between parameters and tracking results is explained.