Research On Distributed Characteristics Of Particle Swarm Optimization And Its Applications

Particle Swarm Optimization is a kind of swarm intelligence algorithm in which exist someproblems such as premature, oscillationμ and slow convergence rate. In order to overcome thesedeficiencies, this article studies PSO algorithm in-depth, proposes an improved Particle SwarmOptimization algorithms and applies the algorithm to image segmentation.The main tasks of this article are as follows:(1) In order to solve the oscillation problem where particles oscillate around the optimalposition, this thesis analyzes the location updating formula from the perspective of physics andproposes a modified algorithm of particle swarm optimization with random flying time. The newalgorithm improves the optimizing capability and reduces the possibility of particles flying overμthe optimal position which leads to oscillation problem while keeps a simple structure.(2) This thesis analyzes the randomness and fuzziness exists in PSO and introduces cloudmodel into the algorithm. This article modifies the original Multi-Colony Particle SwarmOptimization by dividing particles into three groups which evolve based on cloud adaptiveparticle swarm optimization, particle swarm optimization with random flying time and particleswarm Optimization with binarized inertia weight respectively. Experimental results show thatmodified Multi-Colony Particle Swarm Optimization can produce better result compared withoriginal Multi-Colony Particle Swarm Optimization and other modified Particle SwarmOptimization algorithms.(3) This thesis applies particle swarm optimization with random flying time into imagesegmentation by combining it with Otsu algorithm while proposes a new adjustment strategy ofinertia weight as well. Experimental results show that the new algorithm reduces the algorithm·siteration number and running time and has higher efficiency.