基于模拟退火与遗传算法结合的神经网络图像分割

Simulated Annealing and Genetic Algorithms Based on Image Segment with Partially Evolved Hopfield Neural Network

为解决运用Hopfield神经网络优化算法处理图像分割存在的收敛速度与局部最优的矛盾,采用模拟退火策略与遗传算法结合的优化方法来改进传统的优化算法,对迭代收敛后的Hopfield网络在局部范围内运用模拟退火遗传算法,以搜索阈值平面全局最优解,进行图像分割。实验证明,采用此方法可以得到较好的分割效果。

A combined optimization of genetic algorithms with simulated annealing has been applied to image segmentation with good results in this paper.The defect of Hopfield neural network is being captured by local optimal solutions,while the defect of genetic algorithms is the low speed of convergence.Both disadvantages mentioned above have been overcome here.Solutions obtained with the converged Hopfield neural network are applied to the genetic algorithm to search for the optimization on the plane of threshold value.