摘要
提出了一种基于BP神经网络的船舶焊缝缺陷图像识别的方法,通过对船舶焊缝图像进行预处理,提取出有用的目标缺陷,再进行缺陷特征参数计算,将特征参数和焊缝缺陷类型分别作为输入层和输出层,利用BP算法设计3层结构的神经网络,对样本进行训练和识别。实验结果表明,BP神经网络能较准确地识别出船舶焊缝缺陷。
A method of image recognition of ship weld flaw based on BP neural networks was proposed. The pretreatment of the ship weld image was conducted to extract the useful target disfigurement, calculate the characteristic parameters of the disfig- urement. And then the characteristic parameters and the weld disfigurement types were used as the input and output layers. The BP algorithm was used to design a three structure's neural networks ,training and recognizing the samples. The experimenta! result indicates that BP neural networks can more accurately identify the ship weld flaw.
出处
《武汉理工大学学报(信息与管理工程版)》
CAS
2012年第3期271-274,共4页
Journal of Wuhan University of Technology:Information & Management Engineering
关键词
BP神经网络
图像识别
图像预处理
特征参数
船舶焊缝缺陷
BP neural networks
image recognition
image pretreatment
characteristic parameters
ship weld flaw