CAJ | 학술논문

提出了一种基于粒子群优化算法的BP神经网络图像复原方法。 BP神经网络具有很强的学习和泛化能力，可避免传统复原方法对先验知识的依赖性，粒子群算法的全局寻优能力弥补了BP算法对初始权值敏感、收敛速度慢和易陷入局部极小值等问题，将两者结合形成PSO-BP算法，使得图像复原的难度大大下降。实验表明，该方法对模糊图像的复原性能很好，收敛速度快，在视觉和定量分析上都获得了较好的效果。
제출료일충기우입자군우화산법적BP신경망락도상복원방법。 BP신경망락구유흔강적학습화범화능력，가피면전통복원방법대선험지식적의뢰성，입자군산법적전국심우능력미보료BP산법대초시권치민감、수렴속도만화역함입국부겁소치등문제，장량자결합형성PSO-BP산법，사득도상복원적난도대대하강。실험표명，해방법대모호도상적복원성능흔호，수렴속도쾌，재시각화정량분석상도획득료교호적효과。
A new method is proposed for image restoration of BP neural network based on particle swarm optimization. Different from traditional restoration methods,BP neural network has a strong learning and generalization ability to avoid the dependence on a priori knowledge.BP algorithm is sensitive to the initial weights,has slow convergence and is easy to fall into local minima,however,the PSO algorithm,as a global optimization algorithm,can make up for this issues and others.The two algorithms are combined to form a PSO-BP algorithm,making the difficulty of image restoration decline significantly. The experimental results show that the fuzzy image restoration method has better performance,fast convergence and better effects obtained in the visual and quantitative analysis.