基于AI视频辅助的上海延安东路隧道与东西通道火灾联动研究与实现  

Fire Linkage between Yan'an East Road Tunnel&East West Passage Based on AI Video Assistance

作　　者：辛奇峰 XIN Qi-feng(Shanghai Pujiang Bridge&Tunnel Operation Management Co.,Ltd.,Shanghai 200120,China)

机构地区：[1]上海浦江桥隧运营管理有限公司,上海200120

出　　处：《中国市政工程》2023年第4期73-77,111,共6页China Municipal Engineering

摘　　要：上海东西通道工程通车之后与延安东路隧道连接,形成1条东西走向的特长隧道。因2条设施建造于不同年代,存在各方面设计差异,无法有效实现火灾联动。根据“一体化”运营管理要求,梳理问题,提出优化解决方案。结合基于YOLOv8神经网络算法的火灾检测技术,应用在隧道雨喷淋视频联动,并精确定位。采用DehazeNet网络模型进行早期火灾的烟雾图像优化。通过知识蒸馏技术排除隧道环境干扰;通过试验测试,对比自动联动、人工操作及优化系统的效果,提高延安东路隧道与东西通道火灾联动准确率。After the opening of the Shanghai East West Passage project,it is connected to East Yan'an Road Tunnel,forming an extra-long tunnel running east-west.Due to the construction of the two facilities in different eras & differences in design, it is not possible to effectively achieve fire linkage. Based on the requirements of integrated operation management, sort out problems & propose optimized solutions. Combining fire detection technology based on YOLOv8 neural network algorithm, it is applied to tunnel rain spray video linkage & precise positioning. Using the DehazeNet network model to optimize smoke images for early fires. Eliminate tunnel environmental interference through knowledge distillation technology;through experimental testing, the effectiveness of automatic linkage, manual operation, and optimization systems is compared to improve the accuracy of fire linkage between East Yan'an Road Tunnel & East West Passage.

关 键 词：AI视频辅助 隧道火灾联动 延安东路隧道 东西通道 YOLOv8 

分 类 号：U453[建筑科学—桥梁与隧道工程]