摘要
本文提出了一种基于视觉的方法,该方法使用卷积神经网络(CNN)的深层架构来检测隧道衬砌病害,而无需计算缺陷特征。由于CNN能够自动学习图像特征,所提出的方法在没有使用图像处理方法提取特征的情况下工作。设计的CNN在256×256像素分辨率的40k图像上进行训精度约为98%。进行比较研究以使用传统的Canny和Sobel边缘检测方法检查所提出的CNN的性能。结果表明,该方法具有较好的性能,在实际情况下确实可以找到具体的裂缝。
The paper proposes a vision-based approach that uses the deep architecture of the convolutional neural network(CNN)to detect tunnel lining diseases without the need to calculate defect features.Since CNN can automatically learn image features,the proposed method works without extracting features using image processing methods.The designed CNN has a training accuracy of about 98%on a 40K image with a resolution of 256 x 256 pixels.A comparative study is conducted to examine the performance of the proposed CNN using traditional Canny and Sobel edge detection methods.The results show that the method has good performance and can find specific cracks under actual conditions.
作者
胡利娜
Hu Lina(Shanxi Xat Automation Engineering Co.,Ltd.,Taiyuan Shanxi 030012,China)
出处
《山西电子技术》
2019年第5期38-40,共3页
Shanxi Electronic Technology
关键词
地铁隧道
病害检测
深度学习
机器视觉
subway tunnel
disease detection
deep learning
machine vision
