摘要
结合胶合板缺陷检测分类,通过对比研究了几种有效、实用的模式分类方法,讨论了该方法的优势和局限性,提出了一种基于粗糙集和神经网络的模式分类方法.利用粗糙集处理图像特征中冗余的或较差的属性特征,有效地减小了网络规模,将该特性和神经网络的非线性映射能力和很强的抗干扰性相结合,能够进一步提高分类精度和收敛速度.在胶合板缺陷识别的实际应用中,其识别精度达到了90 93%,循环次数较粗糙集预处理前平均下降了1000余次.研究结果表明粗糙集神经网络模式识别方法适于胶合板缺陷分类.
Combined with wood veneer defect inspection,several pattern recognition methods,which are effective and useful,are researched.The advantage and disadvantage of each method are discussed in detail.According to their characteristics,a pattern recognition method based on neural network and rough set is put forward.The rough set theory is used to remove redundant and worse information,and network scale is reduced effectively.Moreover the nonlinearity and high antidisturbance ability of neural networks are introduced to further improve classification accuracy and convergence speed.With the application of defect inspection for wood veneer the accuracy is 9093% for the test set and the training time is reduced by 1116 epochs compared with nonpreprocessing.The experiment research shows that the pattern recognition method based on neural network and rough sets is suitable for defect inspection of wood veneer.
出处
《沈阳建筑工程学院学报(自然科学版)》
2003年第3期224-228,共5页
Journal of Shenyang Architectural and Civil Engineering University(Nature Science)
基金
建设部基金项目(01-4-045)
关键词
分类
神经网络
粗糙集
胶合板
classification
neural Networks
rough sets
wood veneer