应用深度卷积神经网络的色织物缺陷检测

发布时间：2018-03-01 00:32

本文关键词： 色织物图像库缺陷检测深度卷积神经网络映射函数　出处：《纺织学报》2017年02期 　论文类型：期刊论文

【摘要】：针对织物缺陷检测时传统人工的误检率、漏检率较高问题,提出一种应用深度卷积神经网络的色织物缺陷检测算法。因织物图像采集过程中含有较多噪声且信噪比较低,先对缺陷织物进行最优尺寸高斯滤波,有效滤除细节噪声;再根据织物图像特征建立深度卷积神经网络,利用径向基神经网络的非线性映射能力作用于卷积神经网络,并通过反向传播算法调整权值参数,获取无缺陷样本与训练样本之间的映射函数;最后,利用映射函数及特征字典重构图像并提取特征,根据Meanshift算法分割缺陷,确定缺陷位置。结果表明:应用深度卷积神经网络的缺陷检测算法对色织物图像库中的缺陷图像可实现提高检测效率、缩短检测时间,获取准确缺陷位置的目的。
[Abstract]:In order to solve the problem of high false detection rate and high missed detection rate in fabric defect detection, a color fabric defect detection algorithm based on deep convolution neural network is proposed, which contains more noise and lower signal-to-noise ratio (SNR) in fabric image acquisition. Firstly, Gao Si filter is used to filter the defect fabric, and then the depth convolution neural network is established according to the image features of the fabric, and the nonlinear mapping ability of the radial basis function neural network is applied to the convolution neural network. And through the back propagation algorithm to adjust the weight parameters to obtain the mapping function between the non-defect sample and the training sample. Finally, using the mapping function and feature dictionary to reconstruct the image and extract the feature, according to the Meanshift algorithm segmentation defects, The results show that the defect detection algorithm based on the deep convolution neural network can improve the detection efficiency, shorten the detection time and obtain the accurate defect position in the color fabric image library.
【作者单位】：西安工程大学电子信息学院;
【基金】：国家自然科学基金项目(61301276) 陕西省工业科技攻关项目(2015GY034) 西安工程大学研究生创新基金资助项目(CX201602) 陕西省教育厅专项科研计划项目(16JK1342) 陕西省自然科学基础研究计划项目(2015JQ6258)
【分类号】：TS106;TP183;TP391.41

【相似文献】