Abstract: A new method of cooperative evolution parallel particle swarm optimization algorithm is proposed in this paper. It aims at extended superquadric parametric fitting problem which is difficult to solve by traditional method. 3D representation characteristics of extended superquadrics were investigated firstly. The parallel particle swarm optimization algorithm was designed and practically implemented by an idea of island population model. Cooperative evolution parallel particle swarm optimization solved the constrained nonlinear optimization that was transformed to minimax problem for extended superquadric parametric fitting. The results show the enhanced representation capability and accurate and effective modeling by the proposed approach in extended superquadric 3D reconstruction.

Key words: 3D reconstruction, extended superquadrics, parallel particle swarm optimization algorithm, cooperative evolution, constrained nonlinear optimization

摘要： 针对用传统方法难以求解的扩展的超二次曲面三维模型参数拟合问题，提出了用协同演化的并行粒子群优化算法求解的新方法。通过对扩展的超二次曲面三维表示特性的研究，设计和实现了基于岛屿群体模型的并行粒子群优化算法，并用协同演化的思想，将约束非线性优化转化为极小极大问题进行求解。实验结果表明用协同演化的并行粒子群优化算法重构扩展的超二次曲面三维模型，扩大了模型表示能力，建模精确且效率高。

关键词: 三维重构, 扩展的超二次曲面, 并行粒子群优化算法, 协同演化, 约束非线性优化