摘要
人体姿态估计是计算机视觉领域的重要研究方向,将其应用于施工管控领域对工作人员状态进行实时评估,能够有效提升现场施工安全管控的智能化水平。因此,基于OpenPose算法,实现施工现场工作人员人体骨架模型提取,并基于数据标准化及增强手段,构建施工人员姿态分类数据集,实现了对于6类施工人员典型姿态的精准识别;搭建了基于CGAN的人体姿态补全算法,因其具有较强的骨架模型样本生成能力,对于多处缺失等不利情形仍可实现高吻合度的骨架补全。
Human pose estimation is a significant research direction in the field of computer vision.Its application in construction management for real-time assessment of workers'conditions can effectively enhance the smart level of onsite construction safety management.Therefore,based on the OpenPose algorithm,human skeleton model extraction of construction site workers is achieved.Through data standardization and enhancement methods,a dataset for construction personnel posture classification is constructed,enabling precise recognition of six typical postures of construction personnel.A human posture completion algorithm based on CGAN is developed;due to its strong skeleton model sample generation capability,it can achieve high-degree skeleton completion even in adverse situations with multiple missing points.
作者
冯宇
刘正劼
房霆宸
龚剑
FENG Yu;LIU Zhengjie;FANG Tingchen;GONG Jian(Shanghai Construction Group Co.,Ltd.,Shanghai 200080,China;Shanghai Engineering Research Center of Super High Rise Building Intelligent Construction,Shanghai 200080,China;School of Civil Engineering,Tongji University,Shanghai 200092,China)
出处
《建筑施工》
2025年第2期234-239,245,共7页
Building Construction
基金
上海市科技创新行动计划启明星培育扬帆专项(22YF1418100)
上海市科技计划项目(22dz1201706)
上海人才发展资金(2021052)。
作者简介
冯宇(1990-),男,博士,工程师,fengyu@scg.cn。
