塔吊回转视角下施工场景数字孪生高效建模方法  

An efficient digital twin modeling approach for construction sites via rotating tower crane perspective

作　　者：卢昱杰 仲涛 张修龙 赵宪忠[1] Lu Yujie;Zhong Tao;Zhang Xiulong;Zhao Xianzhong(College of Civil Engineering,Tongji University,Shanghai 200092,China;Key Laboratory of Performance Evolution and Control for Engineering Structures of the Ministry of Education,Tongji University,Shanghai 200092,China)

机构地区：[1]同济大学土木工程学院,上海200092 [2]同济大学工程结构性能演化与控制教育部重点实验室,上海200092

出　　处：《土木工程学报》2025年第4期124-136,共13页China Civil Engineering Journal

基　　金：国家重点研发计划(2022YFC3801700);中央高校基本科研业务费专项资金(2024-1-ZD-02);国家自然科学基金(52078374);上海市科委科技创新行动计划(22dz1207800,22dz1207100)。

摘　　要：数字孪生模型能够在信息空间表达施工场景实体特征,是高效的施工信息化管理工具。然而当前复杂施工场地数字孪生建模存在时效滞后、自动化程度低等问题,难以满足施工场地逐日精细化管理的要求。因此提出了一种基于塔吊高空环视图像的复杂施工场景高效建模方法,旨在提高数字孪生建模的时效性和自动化程度。该方法将摄像头布设于回转式塔吊吊臂以获取场地的多视角图像进行建模,首先针对受限的图像采集路径,提出基于点线特征的多视角图像集质量评价方法,以优化数据采集方式从而提高图像集质量,其次通过融合场景几何先验的建模方法,基于输入图像创建具备真实尺度的几何模型,并采用基于高斯采样的施工场地点云去噪方法提升几何模型语义赋予的准确性。最后,以上海某高层住宅项目为例,验证了方法的适用性。结果表明,方法能够自动化地实现施工场地的几何建模与建筑结构语义赋予,且模型几何精度达分米级,完全满足逐日进度管理需求。此外,提出的在塔吊作业场景下的高效数字孪生建模方法具备很强的拓展性,对桥梁、隧道等其他复杂施工场景的数字孪生建模均有一定参考意义。Digital twin models represent construction site entities in the information space,serving as efficient tools for construction information management.However,current digital twin modeling of complex construction sites suffers from time lag and low automation,hindering refined daily progress monitoring.An efficient modeling approach for construction scenarios via rotating tower crane perspective is proposed,aiming at enhancing the timeliness and automation of digital twin modeling.In this approach,cameras installed on rotating tower crane jibs are utilized to acquire multi-view site images for modeling.Firstly,by considering constrained camera trajectory,a multi-view image set quality evaluation method is developed based on point and line features to optimize data acquisition,thereby improving overall image quality.Secondly,by integrating scene geometric priors,a modeling approach is developed to create geometrically accurate and scale-preserving models from input images.Furthermore,a construction site point cloud denoising method based on Gaussian sampling is adopted to improve the accuracy of semantic recognition for building structures in the geometric models.Finally,the applicability of proposed method to a highrise residential project in Shanghai is validated.Results demonstrate that the method is capable of automatically achieving geometric modeling and semantic annotation of construction structures,with decimeter-level geometric accuracy,fully satisfying daily progress tracking requirements.Additionally,this efficient digital twin modeling approach under tower crane operation scenarios appears to be greatly extensible,providing valuable insights for digital twin modeling of other complex construction environments,such as bridges and tunnels.

关 键 词：施工过程 数字孪生 三维重建 多视角图像 图像质量评价 

分 类 号：TU17[建筑科学—建筑理论]