Evolutionary Algorithms for Warp Control Point Placement
alization requires the deformation and registration of templates with a target image. A warping and registration methodology is presented to facilitate this task, utilizing evolutionary optimization routines to automatically determine optimal control point placement between template and target image. The Local Weighted Mean warp used to deform the templates to fit these control points is presented, along with a discussion of three evolutionary algorithms and their application to the problem. The optimization routines of Genetic Algorithm, Particle Swarm Optimization and Simulated Annealing are compared in terms of accuracy, speed and computational requirements, with Particle Swarm Optimization being highlighted as the best method for this task.
article swarm optimization Evolutionary algorithm object localization
Jonathan Michael Spiller Tshilidzi Marwala
School of Electrical and Information Engineering, University of the Witwatersrand, Johannesburg, South Africa
国际会议
武汉
英文
2007-09-21（万方平台首次上网日期，不代表论文的发表时间）