公路隧道视频预处理和病害识别算法  

Video Preprocess and Defect Recognition Algorithm for Road Tunnel

作　　者：胡珉[1,2] 周显威 高新闻[2,3] HU Min;ZHOU Xianwei;GAO Xinwen(SHU-UTS SILC Business School,Shanghai University,Shanghai 201800,China;SHU-SUCG Research Center,Shanghai University,Shanghai 200072,China;School of Mechatronic Engineering and Automation,Shanghai University,Shanghai 200444,China)

机构地区：[1]上海大学悉尼工商学院,上海201800 [2]上海大学上海城建集团建筑产业化研究中心,上海200072 [3]上海大学机电工程与自动化学院,上海200444

出　　处：《华侨大学学报（自然科学版）》2020年第5期595-604,共10页Journal of Huaqiao University(Natural Science)

基　　金：上海市科委重点项目(18DZ1201204)。

摘　　要：基于计算机视觉技术,针对公路隧道病害进行检测与识别,提出视频数据的预处理方法.使用全卷积网络(FCN)模型识别病害的类别和位置,融合不同的上采样结果使最终结果更加精细,结合马尔可夫随机场(MRF)增强FCN模型的空间一致性.实验结果表明:该方法可解决数据冗余、镜头畸变及样本不均衡等问题;该方法在上海市虹梅南路隧道中的应用结果验证其准确度与可靠性.Based on computer vision,the defect detection and identification were conducted for road tunnel.A preprocess method for video data was proposed.Finally,fully convolutional networks(FCN)model was used to identify the category and location of defects.Different up-sampling results were integrated to make the final results more precisely,and the spatial consistency of FCN model was improved by combining Markov random field(MRF).The experimental results show that this method can solve the problems of data redundancy,lens distortion and sample imbalance.The application of this method in the Shanghai Hongmei South Road Tunnel in Shanghai validates its accuracy and reliability.

关 键 词：公路隧道 隧道病害 图像预处理 目标识别 深度学习 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]