摘要
现有的对焦评判规则一般都缺乏考虑人的主观因素,因此其成像系统的调焦效果难以有进一步的改善。采用基于遗传算法的优化BP网络完成图像质量模式识别的功能,即用神经网络去拟合图像经小波分析后提取的特征值到图像质量之间的映射关系,该映射有效的改善了调焦效果。同时由于遗传算法具有更好的鲁棒性,用遗传算法代替BP算法来搜索神经网络的连接权,可解决BP网络陷入局部极小问题。利用小波变换对目标进行多分辨率分析,从而模拟人眼的多频率通道分解现象,同时利用基于遗传算法的BP神经网络完成图像质量模式识别的功能,可以有效提高调焦效果。
The existing evaluation rules for auto-focus methods lack people＇s subjective factors. That＇s why the fo- cusing effect can hardly be improved. A genetic algorithm-based optimized BP network to complete the function of pattern recognition of figure quality features is proposed in the paper. In the method, the mapping relationship between eigenvalue which is extracted by wavelet transform and figure quality can be calculated. And the mapping relationships improve the focusing effect effectively. Meanwhile, because the genetic algorithm has better robustness, it can solve the local minimum problem of BP network by searching the connection weight of neural network. The paper uses the wavelet multi-resolution analysis on the target, which simulates the human eyes＇ phenomenon of multi-frequency channel decom- position. At the same time, by using genetic algorithm based on BP neural network to complete the function of pattern recognition of figure quality features, it can effectively improve the focusing effect.
出处
《电子测量与仪器学报》
CSCD
2012年第5期398-403,共6页
Journal of Electronic Measurement and Instrumentation
关键词
自动调焦
小波变换
遗传算法
神经网络
auto focus
wavelet transform
genetic algorithm
neural network
