基于工业互联网的隧道监控视频语义分析  

Semantic Analysis of Tunnel Surveillance Video Based on Industrial Internet

作　　者：刘文璇 巫世峰 夏红霞[1] 钟忺[1,3] LIU Wen-xuan;WU Shi-feng;XIA Hong-xia;ZHONG Xian(School of Computer Science and Artificial Intelligence,Wuhan University of Technology,Wuhan 430070,China;ZhongQianLiYuan Engineering Consulting Co.,Ltd.,Wuhan 430071,China;Hubei Key Laboratory of Transportation Internet of Things,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学计算机与人工智能学院,湖北武汉430070 [2]中乾立源工程咨询有限公司、交通物联网技术湖北省重点实验室,湖北武汉430071 [3]武汉理工大学交通物联网技术湖北省重点实验室,湖北武汉430070

出　　处：《软件导刊》2021年第10期19-25,共7页Software Guide

基　　金：国家自然科学基金项目(61303029);中央高校基本科研业务费专项资金项目(191010001);交通物联网技术湖北省重点实验室开放基金项目(2020Ⅲ026GX);湖北省自然科学基金重点项目(2017CFA012)。

摘　　要：近年来,移动终端、监控设备等多媒体数据的海量增长推动了互联网服务的巨量交互。隧道视频监控是城市交通的重要组成部分,其与工业互联网的结合是实现城市交通产业高质量发展的重要前提和保障,也是建设网络强国和制造强国的重要支撑。为此,提出基于工业互联网技术的城市隧道视频监控信息采集与语义分析框架,同时结合语义分析的城市隧道视频监控深度学习方法,解决隧道监控系统在实际应用场景下的安全隐患问题,减少因突发故障导致的损失。实验结果表明,框架中的视频图像场景分割算法在Cityscapes数据集上的均交并比(mIoU)达到82.9%,行人搜索算法在CUHK-SYSU数据集上的Top-1准确率达到95.7%,视频图像异常事件识别算法在WIDER数据集上的平均准确率(mAP)达到75.3%。所提框架使隧道视频监控系统实现了数据共享与信息网络化,提高了城市隧道的运行性能及在紧急情况下的事故处理能力,实现了人工智能与工业互联网的融合创新,有助于城市隧道的智能化管理。In recent years,massive growth of multimedia data from mobile terminals,surveillance equipment has promoted a huge amount of interaction with Internet services.As an important part of urban transportation,the combination of tunnel video surveillance and Industrial Internet is an important prerequisite and guarantee for achieving high-quality development of the urban transportation industry,and is also an important support for the strategy of building a network power and a manufacturing power.Proposes a framework for urban tunnel video surveillance information collection and semantic analysis based on Industrial Internet technology,and proposes a deep learning method for urban tunnel video surveillance combined with semantic analysis.The framework solves the hidden dangers of the tunnel monitoring system in actual application scenarios,reduces the output loss caused by sudden failures.The experimental results show that the proposed video image scene parsing algorithm achieves the performance of 82.9%on mIoU in the Cityscapes dataset,the pedestrian search algorithm achieves 95.7%on top-1 accuracy in the CUHK-SYSU dataset,and the video image abnormal recognition algorithm achieves 75.3%accuracy in WIDER dataset.The framework improves the operation function of urban tunnels and the ability to handle accidents in emergency situations,realizes the integration and innovation of artificial intelligence and Industrial Internet,enhances the safety of traffic in the tunnel,and fully realizes the intelligent management of urban tunnels.

关 键 词：工业互联网 城市隧道监控 人工智能 信息采集 语义分析 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]