UAV-based Change Detection on Construction Sites for Rapid 3D Reconstruction

Three-dimensional (3D) reconstruction is a pivotal technology for generating digital representations of physical scenes. It plays an essential role in the timely and precise evaluation of site conditions and consequently facilitates construction management. Implicit representation-based methods such as neural radiance fields (NeRF), which have been recently proposed, demonstrate unprecedented accuracy and scene fidelity in 3D reconstruction. However, despite their potential, these implicit methods encounter significant challenges in rapidly updating representation for large-scale and complex scenes like construction sites due to their tremendous computational demands. This latency in reflecting the current state of scenes can hinder their applicability to tasks that require precise and real-time monitoring of site conditions. To overcome these challenges, this paper introduces a novel framework that employs a change detection technique, enabling the rapid identification and reflection of local changes into existing 3D models.