基于遗传算法思想的模拟退火算法应用于红外图像分割

IR Image Segmentation Based The 'SA Under GA Ideas' Algorithms

文中借鉴进化论的"适者生存"的思想和遗传算法中的概念,对传统的模拟退火算法进行了改进。特别是对Metropolis采样中的随机逼近,采用了仿生技术加以指导,加速随机逼近向全局最优方向进行,即文中所提出的基于遗传算法的模拟退火算法(‘SA Under GA Ideas’Algorithms)。仿真试验证明,该算法能够加速模拟退火过程,同时对噪声和孤立的小区域可以进行有效的抑制和合并。

Under the 'Survival of the fittest' ideas and Genetic Algorithms, this paper improves the Classical Simulated Annealing. Especially in the process of Metropolis Stochastic Approximation, the new algorithms called 'SA Under GA Ideas' Algorithms impulses the stochastic approximation to the optimization. The following experiments approved the advantages of the new algorithms, such as the noise immune and combination of petty independent regions.