Smartphone-Based Indoor Localization and Object Detection for Construction Site Asset Management and Digital Twins

Abstract

This research addresses the challenge of keeping building asset information accurate and up to date for facility management and digital twin applications, especially in indoor environments where GPS signals are not available. The asset management is still largely carried out manually, making it time-consuming, difficult to update, and susceptible to information fragmentation. These limitations reduce on-site productivity and limit the practical use of digital twins throughout the building lifecycle. One of the main technical challenges in indoor environments is determining the exact location where images are captured without GPS. To address this issue, we introduce a smartphone-based indoor positioning system that uses the smartphone’s built-in sensors and does not rely on expensive equipment such as beacons or LiDAR. The system can determine indoor locations with centimeter-level accuracy. This high level of accuracy allows photos to be placed correctly on indoor floor plans, making it easier to understand the current condition of buildings and the locations of newly installed assets. For asset identification, this study uses an image-based recognition approach based on a YOLOv8 object detection model. The model is trained on a mixed dataset that includes manually labeled images collected from real construction sites, along with selected asset categories from the COCO dataset. Once assets are detected, they are linked to their exact indoor locations and visualized within a digital twin environment. By combining accurate indoor localization with automated asset detection, the proposed approach can simplify asset surveys, reduce manual effort, and significantly improve productivity in construction and facility management.