基于CNN的聚乙烯燃气管道缺陷分类  

Classification of Polyethylene Gas Pipeline Defects Based on CNN

作　　者：符前坤 王洋 李萌 Fu Qiankun;Wang Yang;Li Meng(School of Mechanical Engineering,Xinjiang University,Urumqi 830017)

机构地区：[1]新疆大学机械工程学院,乌鲁木齐830017

出　　处：《中国特种设备安全》2023年第S02期48-51,共4页China Special Equipment Safety

摘　　要：本文首先对采集到的图像进行初步筛选,提升了缺陷图像获取的速度。对于有缺陷的PE输气管道图像,先使用加权平均法对图像进行灰度处理,再使用自适应直方图法增强对比度,最后使用均值滤波法去除噪声。用CNN对获得的图像进行训练,通过选择合适的参数,完成了分类器的设计。实验结果表明,本文采用的方法平均准确率达到94.82%,能够实现对地下PE输气管道缺陷类型的识别。In this paper,the acquired images are initially screened to improve the speed of defect image acquisition.For the defective PE gas pipeline image,the weighted average method is used to process the gray level of the image,then the adaptive histogram method is used to enhance the contrast,and the mean filtering method is used to remove the noise.The obtained images are trained by CNN,and the classifier is designed by selecting appropriate parameters.The experimental results show that the average accuracy rate of the method adopted in this paper is 94.82%,which can identify the defect types of underground PE gas transmission pipeline.

关 键 词：管道缺陷 图像处理 图像增强 CNN网络 

分 类 号：X933.4[环境科学与工程—安全科学]