Collaborative Research: Automatic Behavior Monitoring for In-depth Analysis of Construction Fatalities and Injuries

合作研究：自动行为监测，深入分析施工伤亡情况

• 批准号: 1200120
• 负责人: Feniosky Pena-Mora
• 金额: $ 6.83万
• 依托单位: Columbia University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2012
• 资助国家: 美国
• 起止时间: 2012-05-01 至 2015-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1200120&HistoricalAwards=false
• 关键词: Collaborative; Research; Automatic; Behavior; Monitoring

The objective of this research is to explore computer vision-based monitoring methods, enabling the automatic and constant observation of construction workers for in-depth safety analysis. Taking into account the fact that the number of fatalities in construction remains the highest among all industries, and that approximately 80 to 90 percent of accidents are strongly associated with workers' unsafe behavior and acts, the automatic capture and systematic understanding of unsafe behavior has great potential to contribute to the reduction and prevention of injuries and fatalities in construction. Specifically, using video and image sequences, the proposed system estimates the 2D location of a human skeleton, computes the 3D location of body joints, and identifies the worker?s unsafe motions using machine learning techniques. To investigate the feasibility and potential of the proposed methods for behavior monitoring, several representative motions in traumatic (e.g., falls) and ergonomic (e.g., overexertion and repetitive motions) injuries are tested as a case study. If successful, the findings of this research could lead to the prevention of injuries and fatalities in the construction industry by providing an in-depth understanding of human behavior and actions in terms of safety. Further, the motion analysis techniques developed in this project can be applied to diverse industries (e.g., manufacturing and shipbuilding) where labor is an important resource, providing a means to automatically collect data on human behavior and thus enabling its effective understanding. In addition, the education plan (e.g., the course that deals with understanding human behavior in safety, and the workshop planned for industry professionals and students) will provide effective education of future managers and engineers, both of whom will improve safety in the US workplace. Further, female and underrepresented students will be recruited and integrated into the planned research and education activities (e.g., safety seminars and workshops and interdisciplinary research participation opportunities).

本研究的目的是探索基于计算机视觉的监控方法，实现对建筑工人的自动持续观察，进行深入的安全分析。考虑到建筑业的死亡人数在所有行业中仍然是最高的，并且大约80%到90%的事故与工人的不安全行为和行为密切相关，对不安全行为的自动捕获和系统理解具有很大的潜力，有助于减少和预防建筑业的伤害和死亡。具体来说，使用视频和图像序列，该系统估计人体骨骼的2D位置，计算身体关节的3D位置，并识别工人？S使用机器学习技术的不安全动作。为了研究所提出的行为监测方法的可行性和潜力，我们对创伤性（如跌倒）和人体工程学（如过度运动和重复运动）损伤中的几种代表性动作进行了案例研究。如果成功，这项研究的结果可以通过深入了解人类在安全方面的行为和行动，从而预防建筑行业的伤害和死亡。此外，在这个项目中开发的运动分析技术可以应用于劳动力是重要资源的不同行业（例如制造业和造船业），提供了一种自动收集人类行为数据的方法，从而使其能够有效理解。此外，教育计划（例如，涉及理解人类安全行为的课程，以及为行业专业人士和学生计划的研讨会）将为未来的管理人员和工程师提供有效的教育，他们都将改善美国工作场所的安全。此外，将征聘女性和人数不足的学生，并将她们纳入计划的研究和教育活动（例如，安全讨论会和讲习班以及跨学科研究参与机会）。