Fast 3D site reconstruction using multichannel dynamic and static object separation

Abstract

Three-dimensional (3D) models, characterized by their visualization, accuracy, and interactive information presentation, effectively facilitate collaboration and optimize management throughout the construction process. However, existing 3D reconstruction methods frequently fail to simultaneously satisfy the requirements for onsite applicability and fast performance. To address this challenge, this paper proposes a monocular camera-based 3D reconstruction method designed for onsite applicability and introduces dynamic–static separation to reduce the computational burden for faster processing. This approach enables the preestablishment of 3D models for static and dynamic objects. The positioning, pose, and orientation information of objects can be quickly integrated from multiple channels for fast 3D site reconstruction. Experimental results demonstrate that target objects can be identified across multiple channels and quickly integrated into 3D models. This paper offers both theoretical and practical contributions by enabling 3D reconstruction of construction sites using monocular cameras, which enhances project safety management and supports the implementation of digital twins.