Projective Point Matching Using Modified Particle Swarm Optimization

摘要

A projective point matching algorithm based on modified particle swarm optimization is presented. In the paper, the point matching problem turns into an optimization with two series of parameters, projective transform parameters and correspondent mapping parameters. Firstly, a modified particle swarm optimization (PSO) is introduced and a new rule searching for correspondences, closer point matching rule, is also proposed. We use PSO find the optimal solution. It updates the best geometric transform parameters constantly till find the global best, and in each iteration the closer point matching rule is applied to get the correspondent mapping parameters under the temporary fixed transform parameters. Experiments on both synthetic points and real images demonstrate the algorithm is reliable and validate.