SwATrack: A Swarm Intelligence-based Abrupt Motion Tracker
Mei Kuan Lim, Chee Seng Chan, Dorothy Monekosso, Paolo Remagnino · 2013
Conventional tracking solutions are not feasi-ble in handling abrupt motion as they are based on smooth motion assumption or constrained motion model; where the motion is often governed by a fixed Gaussian distribution. Abrupt motion however, is not subjected to motion continuity and smoothness. To assuage this, we propose a novel abrupt motion tracker that is based on swarm intelligence- the SwA-Track. Unlike existing swarm-based filtering methods, we firstly introduce an optimised swarm-based sampling strategy to enrich the trade-o ↵ between the exploration and exploitation of the search space in search for the optimal proposal distribution. Secondly, we propose adaptive acceleration parameters to allow on the fly tuning of the best mean and variance of the distribution for sampling. The adaptive strategy requires no train-ing stage thus allowing flexibility in the motion model, while relaxing the number of particles deployed. Exper-imental results in both the quantitative and qualitative measures demonstrate the e↵ectiveness of the proposed method in tracking abrupt motions. 1