地铁施工安全信息价值密度研究

魏然

摘要

近年来，物联网、大数据、人工智能 、5G等新兴信息技术在我国地铁施工安全管理研究中得到大量应用，产生了来源丰富、结构复杂、规模巨大的信息数据集，如果对其进行无差别的管理必然导致低效和浪费。信息价值密度是衡量信息价值的重要因素之一，可用于指导信息管理。如何识别地铁施工中不同安全信息的价值密度等级如何提升信息价值密度以更好的支持安全管理决策，遵循何种信息管理策略来有序开展地铁安全信息的采集、传输、存储和处理分析，已成为现今地铁施工安全管理中亟待解决的重要课题。本文借鉴国内外相关研究，以信息管理学的相关理论为基础，综合运用模糊贝叶斯网络模型、计算机视觉技术、大数据存储技术等，系统研究了地铁施工安全信息价值密度定义、价值密度评价模型、价值密度提升方法及其信息管理框架等，实现了对现今具备大数据特点的地铁施工安全信息高效率、低成本的管理目标。主要研究工作主要包括：

1、系统研究并提出地铁施工安全信息模型与信息价值密度定义。通过分析安全系统论、安全控制论和安全信息论的各个要素，开展地铁施工安全信息的定义及信息模型构建；在深入分析其信息特征的基础上，定义相应价值密度并提出相关影响因素。

2、对地铁施工安全信息价值密度的质量特性、时效特性和业务特性这三项指标及分项指标进行定义，引入模糊综合评价法以实现对模糊指标的量化，然后构建基于模糊贝叶斯网络的信息价值密度评价模型，推理获得地铁施工安全相关数据集的信息价值密度等级。

3、针对地铁施工安全视频信息价值密度低、价值提取难的问题，提出了一种基于联合注意力时空池的递归卷积网络（Attentive Spatial Temporal Pooling Network ,ASTPN）的施工人员身份识别方法。该方法利用空间注意力网络提取空间特征图利用时态注意力网络提取时间信息，然后通过计算特征之间的距离来识别工人身份。最后，将上述视频数据加工分析技术融入到地铁施工作业安全管理场景中，提出了一种地铁施工安全视频信息价值密度提升方法 。

对地铁施工安全信息实行基于信息价值密度的低成本、高效率的管理策略是提高地铁施工安全管理水平，减少安全事故的重要工作。本文构建了地铁施工安全信息模型并定义了其信息价值密度，开展了信息价值密度评价模型及信息价值密度提升方法的研究，设计研发基于信息价值密度特点的地铁施工安全信息管理框架、存储方案及系统架构等，并在武汉地铁建设工程中进行了系统应用实例的验证，为实现地铁施工安全信息价值密度的落地应用提供了指导和参考。

关键词：地铁施工安全信息模型，信息价值密度评价，信息管理框架，模糊贝叶斯网络，计算机视觉

Abstract

In recent years, emerging information technologies such as the Internet of Things,big data, artificial intelligence, and 5G have been widely used for safety in subway construction. As a result, data with multiple sources, complex structures, and enormous scales are collected and stored. Ind ifferential management of these data inevitably leads to inefficiency and waste. Information value density is an essential factor in measuring information value and can be used in information management. In doing so, it is necessary to address the following research questions for safety management, including:(1) How to identify the levels of different safety information value density in subway construction (SIVDSC); (2) How to improve the SIVDSC to better support safetymanagement decisions; (3) What information management strategies to follow to carry out the collection, transmission, storage and analysis of subway safety information in an orderly manner.

To address the above research questions, this research systematically explores the definition of the SIVDSC the model of the SIVDSC, the enhancement method of the SIVDSC and its information management framework of the SIVDSC based on theoriesof information management, fuzzy Bayesian network model, computer vision technologies, and big data storage technology. This research can achieve the goal of high efficiency and low cost safety management in subway construction. The main workof this research includes:

First, the definition of the safety information model for subway construction(SIMSC) and SIVDSC are proposed systematically. Through the analysis of safetysystem theory，safety control theory and safety information theory, the definition ofsubway construction safety information and the architecture of SIMSC are carried out.Besides，the SIVDSC's characterisation and influential factors are carried out byanalysing subway construction safety information characteristics.

Second, this dissertation quantifies the subway construction data 's quality,timeliness, business characteristics, and corresponding sub characteristics. Also, the fuzzycomprehensive evaluation method is introduced to realise the quantification of fuzzyfactors. Then, the fuzzy comprehensive evaluation model of SIVDSC based on the fuzzyBayesian network is constructed to infer the SIVDSC's levels of various data sets.

Third, aiming at the problems of low-value density and difficult informationextraction of subway construction video data，,a construction personnel identificationmethod based on Attentive Spatial-Temporal Pooling Network (ASTPN) is proposed. Thedeveloped approach consists of (l) extracting spatial feature maps using the spatialattention network,(2) extracting temporal information using the temporal attentionnetworks, (3) recognising a person's identity by computing the distance between features.Finally, the video data processing and analysis technology is integrated into the subwayconstruction safety management scene,and a method to enhance SIVDSC of the videoinformation is proposed.

The low-cost and high-efficiency utilisation strategy of big data of subway construction safety information based on SIVDSC is essential to improve the safety management level of subway construction and reduce safety accidents. This researchconstructs SIMSC,defines SIVDSC，researches the evaluation model and promoticionmethod of SIVDSC, designs and develops the subway construction safety informatiion management framework,storage scheme and system architecture based on thecharacteristics of SIVDSC, and the application of the system is verified in Wuhan subwayconstruction. Furthermore, the developed management system can be a reference fordeploying SIVDSC's application.

Keywords: subway construction safety information model, information value density evaluation, information management framework, fuzzy bayesian network, computer vision