Visual Tracking using a Memory Guided Particle Filter with Annealed Weighted Quantum-behaved Particle Swarm Optimization

Abstract

Swarm Intelligence is a discipline motivating the design and analysis of new machine learning techniques and robotic systems and studies the collective behavior of populations of simple agents to solve problems that are far too complex for an individual. This thesis aims to develop an evolutionary Particle Filter with a memory guided proposal step size update and an improved Quantum-behaved Particle Swarm Optimization (QPSO) resampling scheme for visual tracking applications. The proposal update step uses importance weights proportional to velocities encountered in recent memory to limit the swarm movement within probable regions of interest. The QPSO resampling scheme uses a fitness weighted mean best update to bias the swarm towards the fittest section of particles while also employing a simulated annealing operator to avoid subpar fine tune during latter course of iterations. By moving particles closer to high likelihood landscapes of the posterior distribution using such constructs, the sample impoverishment problem that plagues the Particle Filter is mitigated to a large extent. Experimental results on benchmark sequences imply the proposed method albeit computationally intensive, outperforms competitive candidate trackers such as the traditional Particle Filter and the Particle Swarm Optimization based Particle Filter (PSO-PF) on a suite of tracker performance indices. This leads to the possibility of further work towards a parallelized implementation enabling multi-target tracking using multi-cue sensing models with niche formation for sub-swarms.