The Method And Technique Of Particle Swarm Optimization Object Tracking

In this paper, the particle swarm optimization algorithm has the features of traversingthe search space and rapid convergence, which is the reason why the particle swarmoptimization algorithm is applied to object tracking to achieve the positioning in the case ofthe similarity function presenting "multimodal" when partial occlusion occurs in objecttracking. An improved particle swarm optimization object tracking algorithm is proposed toovercome the limitations of inertia weight adjustment mechanism when the particle swarmoptimization algorithm is applied to object tracking. The improved algorithm can ensureimproving the efficiency of arithmetic operations in the same tracking accuracy.The main research contents and achievements are as follows:1. The basic principles of object tracking and the particle swarm optimizationalgorithm are introduced. The advantage of the particle swarm optimization algorithm isapplied to object tracking are analyzed.2. The extraction of object feature, the similarity measure and particle swarmoptimization algorithm parameters selection are briefly introduced. Experimental resultsindicate that the method is well adapted to the situation when partial occlusion occurs inobject tracking. Especially by comparison with MAD algorithm, highlight the advantagesof particle swarm optimization object tracking algorithm.3. To overcome the limitations of inertia weight adjustment mechanism when theparticle swarm optimization algorithm is applied to object tracking, an improved particleswarm optimization object tracking algorithm is proposed. Firstly, the object and theparameters in particle swarm optimization algorithm are initialized. Secondly, the inertiaweight adjustment mechanism is improved by using the evolution rate of particle, and theinertia weight is achieved by taking the conditions of different particles in each generationinto consideration. Then the speed, the position, the individual optimum and the globaloptimum of the particles are updated simultaneously while the next iteration is proceeding.Finally, the area which has the largest similarity function value is defined as the object bycomparing the fitness value of each particle with the others. Experimental results indicatethat the method reduces the iterations to obtain the same fitness value, and improves theoperation efficiency by42.9%in comparison with the particle swarm optimization objecttracking method which uses self-adapted inertia weight adjustment mechanism. Theaccurate positioning of the object is achieved in the case of the similarity function presenting "multimodal", the method is well adapted to the situation when partial occlusionoccurs in object tracking.Finally, the major findings in this paper are summed up, while pointing out theorientation of further study.