Improved particle swarm optimization algorithm

Nowadays, one of the most fundamental processes for realization video of contents is the object tracking, in which the process of location the moving object is performed in each video frame. In tracking process, the target must be described by a feature. In this paper, for the purpose of describing the target and removing the appearance sensitivity, the weighted color histogram is used as a target feature in order to reduce the effect of edge pixels on the target feature. This reduces the sensitivity of the algorithm to change deformation, scale variation and rotation, as well as the occlusion on the description of target feature. In the proposed method, particle swarm optimization algorithm has been used for search process. Maximization of the similarity function and calculating the minimum Bhattacharyya distance are used to determine target location. The fuzzy control parameters are used for the particle swarm optimization algorithm, which provides a novel method, which can regulate each control parameter and update according to the different states of each particle in each generation. The improved particle swarm algorithm is evaluated with 11 benchmark functions. The obtained results by improved algorithm show that appropriate convergence in a low number of iterations. The proposed method compared to state-of-the-art methods provides high performance in the success and precision rate on the OTB50 dataset.

The particle swarm optimizer (PSO) is a population-based metaheuristic optimization method that can be applied to a wide range of problems but it has the drawbacks like it easily falls into local optima and suffers from slow convergence in the later stages. In order to solve these problems, improved PSO (IPSO) variants, have been proposed. To bring about a balance between the exploration and exploitation characteristics of PSO, this paper introduces computationally fast and efficient IPSO algorithms based on a novel class of exponential learning factors (ELF-PSO). This class contains time-varying exponential learning factors (TELF), random exponential learning factors (RELF), self-adjusting exponential learning factors (SELF) and linear-exponential learning factors (LELF) strategies. Experiment is performed and compared with a set of well-known constant, random, time-varying and adaptive learning factors strategies on a suite of nonlinear benchmark functions. The experimental results and statistical analysis prove that ELF-PSO algorithms are able to solve a wide range of difficult nonlinear optimization problems efficiently. Also these results show that the proposed methods outperform other algorithms in most cases.