The International Association for Automation and Robotics in Construction
Publications / CSCE/CRC 2025 - Montreal, Canada
Elevating Construction Progress Monitoring: A Comprehensive UAV-Based System for Long-Term Outdoor Construction Site Mapping
The construction industry is crucial to the global economy, yet project management and progress monitoring often suffer from delays and cost overruns. While many studies have examined RGB cameras, LiDAR scanners, and Computer Vision (CV) algorithms for tracking physical progress, most are project-specific, and a widely adopted data collection and processing system, along with a standard benchmark dataset, remains unavailable. Unmanned Aerial Vehicles (UAVs) have significantly advanced the methodology of construction site maps creation, activity recognition, and progress tracking, especially for indoor and static outdoor scenes. However, capturing dynamic outdoor environments over time with rich visual detail and processing large-scale data presents ongoing challenges, raising concerns about safety, quality, and consistency. To address these issues, this paper proposes a systematic and consistent data collection approach for outdoor construction sites, capturing RGB images, videos, and point clouds twice weekly starting September 27, 2024. To the best of our knowledge, we are the first to present a high-quality, safe, and repeatable workflow; our site mapping pipeline using DJI and CloudCompare; and alignment techniques using Umeyama and COLMAP. Key challenges include moving objects on site, occlusions causing incomplete point clouds, and acute weather conditions. Our dataset totals 20 successfully recorded and processed site maps by the time this paper was written. We also study modern 3D reconstruction baseline algorithms for building complementary digital site representations in addition to our processed point clouds and provide experimental setups details and benchmark results, demonstrating the effectiveness of 3D Gaussian Splatting (3DGS) models.