摘要
针对冷轧带钢表面缺陷图像模式识别中出现的问题,引入模糊模式识别和反向传播神经网络识别方法。在研究比较两种识别方法的基础上,利用模糊模式识别在剔除噪音数据和反向传播神经网络在模型拟合和非线性识别上的优势,提出一种新的模糊神经网络方法,并详细讨论了算法的结构特点及其实现方法。对五种出现频率较高的典型缺陷图像进行计算机实验研究,结果表明,该方法能对缺陷图像进行有效的识别,具有良好的性能。
Aiming at the problems in pattern recognition of cold steel strip surface defect images, the methods of fuzzy pattern recognition and Back Propagation (BP) neural network recognition were introduced. Based on the comparison of these two methods, by making full use of advantages of the fuzzy pattern recognition in rejecting noise and of the BP neural network in model fitting and non-linear recognition, a new fuzzy-BP neural network recognition method was proposed, and the structure characteristics and realization approach of its algorithm were discussed in detail. Computer experiments were carried out on five typical and frequent surface defect images of cold steel strip, and its results showed that the method could effectively recognize the surface defect images of cold steel strip with good performance.
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2007年第9期1774-1779,1786,共7页
Computer Integrated Manufacturing Systems
基金
国家自然科学基金资助项目(50574019)
国家科技部重大基础研究前期研究专项资金资助项目(2003CCA03900)~~
关键词
冷轧带钢
表面缺陷
模糊模式识别
反向传播神经网络
模糊神经网络
cold steel strip
surface defect
fuzzy pattern recognition
back propagation neural network
fuzzy neural network