AR-Driven High-Precision Mapping of Underground Pipelines Using MultiSet AI
Year: 2026
Type: conference-paper
Abstract: Immersive technologies such as Augmented Reality and artificial intelligence-based spatial computing may revolutionize the traditional pipeline mapping and visualization techniques. Most of the traditional methods for pipeline detection depend on GPS-based localization, a manual survey, and outdated maps. The accuracy is often poor, with minimal real-time visualization that may lead to pipeline strikes, safety hazards, and financial loss. AR-based systems can leverage enhanced spatial awareness and better decisionmaking during excavation and construction activities through highly interactive and intuitive visualization of subsurface utilities. The AR-driven framework for high-precision underground pipeline mapping presented in this paper allows users to visualize water, gas, and sewer pipelines directly over the real world using mobile devices. The Unity and AR Foundation combine with the MultiSet AI VPS to deliver centimeter-level spatial accuracy that significantly outperforms conventional GPS-based techniques. The system ensures correct alignment of virtual pipeline models with respect to the real environment using AI-driven visual localization even in challenging outdoor scenarios. The framework allows costeffective implementation with widely available smartphones, scalable deployment, and real-time interaction. Experimental data also revealed reduced dependence on specialized surveying equipment, enhanced user understanding, and more accurate visualization information. To sum up, a great opportunity exists for improving the dependability, efficiency, and safety of infrastructure management by adopting the proposed AR-based system for visualization along a pipeline. This method provides a feasible model for forthcoming smart cities and advanced spatial visualization systems for various applications.
Institution: National Engineering College
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203606214