【作者】 李鑫；

【导师】 刘恒昌； 郑维；

【作者基本信息】 电子科技大学 ， 电子信息（专业学位）， 2023， 硕士

【摘要】 建筑行业作为典型的劳动密集型行业,对建筑工人效率的管理是整个行业实现降本增效的关键环节。然而,对于大型建筑项目而言,仅通过人工方式可能难以实现对项目人员活动的实时管理监控。与此同时,随着无线通信与物联网设备技术的飞速发展,基于传感器的人体活动识别技术逐步展现出了实现自动化人员管理的潜力。然而,考虑到传感器部署侵略性与肢体活动识别准确性之间的矛盾,目前已有的实施方案在实际应用中面临着一定程度的阻碍。因此,本文基于对传感器的实际应用,通过采用计算机方法,如基于卷积神经网络(CNN)的ROCKET和Deep Conv LSTM特征提取方法、循环神经网络(RNN)及线性分类器,对建筑人员行为进行识别分析,探讨了在现实场景下对工人效率管理的可行性方案,为建筑施工行业数字化转型提供了参考和借鉴。首先,本文采用案例研究的方式,探讨了基于传感器的施工活动信息采集方案对于传感器设置位置、个数以及采样频率的选择。接着,本文提出了基于工人工序与自动化活动识别结果的工人工作量统计方案。相比于其他研究计算工人时间利用效率进行统计的方案,该方案的优势在于能够建立起工人活动信息与项目施工进展的直接关联。最后,本文提出利用自动化采集得到的施工效率数据对建筑工地巡检员巡检路线进行优化的方法,并使用基于现实数据的模拟实验对该方法的有效性进行了验证。在规划巡检路线的实验中,作者提出了基于遗传算法的启发式路线搜索方案,并将该方案与使用LKH算法所得的最短路线方案进行对比,取得了更好的效果。结果表明,仅使用一至两个传感器就能够实现对工人主要施工活动进行较为准确的识别,而基于该识别结果的工作量统计的误差为8.64%;除此之外,使用工人工效数据进行优化后的巡检路线与最短巡检路线相比,可以提高14.88%的工人施工积极性。与人工采样的方法相比,自动化建筑效率监测可以更快速、客观地获取整个工地工人的工作效率数据。这对于建筑项目的降本增效以及管理水平的提升具有重要意义。更多还原

【Abstract】 The construction industry,as a typical labor-intensive industry,places great importance on managing the efficiency of construction workers to achieve cost reduction and increased effectiveness across the industry.However,for large-scale construction projects,real-time management and monitoring of project personnel activities may be difficult to achieve through manual methods alone.At the same time,with the rapid development of wireless communication and Internet of Things(Io T)device technologies,sensor-based human activity recognition technology has gradually demonstrated its potential for realizing automated personnel management.However,considering the contradiction between the invasiveness of sensor deployment and the accuracy of sensor-based human activity recognition,existing implementation plans face certain challenges in practical applications.Therefore,this thesis,based on the actual application of sensors and by adopting computer methods such as convolutional neural networks(CNNs)like ROCKET and Deep Conv LSTM for feature extraction,recurrent neural networks(RNNs),and linear classifiers,analyzes and recognizes the behavior of construction personnel to explore feasible solutions for worker efficiency management in real-world scenarios,providing reference and inspiration for the digital transformation of the construction industry.First,this thesis adopts a case study approach to explore the selection of sensor placement,quantity,and sampling frequency for construction activity information collection schemes based on sensors.Next,this thesis proposes a worker workload calculation scheme based on worker processes and automated activity recognition results.Compared to other research calculating worker time utilization efficiency,the advantage of this scheme lies in its ability to establish a direct connection between worker activity information and project construction progress.Finally,this thesis proposes a method for optimizing the inspection route of construction site inspectors using the automatically collected construction efficiency data and verifies the effectiveness of this method using simulation experiments based on real data.In the inspection route planning experiment,the author proposes a heuristic route search scheme based on genetic algorithms and compares it to the shortest route scheme obtained using the LKH algorithm,achieving better results.The results show that only one or two sensors can achieve relatively accurate recognition of workers’ primary construction activities,with an error of 8.64% for workload calculations based on these recognition results.In addition,compared to the shortest inspection route,the optimized inspection route using worker efficiency data can improve worker construction enthusiasm by 14.88%.Compared to manual sampling methods,automated construction efficiency monitoring can more quickly and objectively obtain work efficiency data for all workers on a construction site.This has significant implications for cost reduction and management level improvement in construction projects.更多还原