Session: 01-10-01: Machine Learning, Artificial Intelligence, and Deep Learning in Dynamics, Vibrations, and Control

Paper Number: 166300

Deep Learning-Based Object Detection and Analysis for Underground Water and Utility Line Localization 

Accurate detection of underground water and utility lines is essential for urban planning, infrastructure maintenance, and construction. Traditional methods like ground-penetrating radar (GPR) and electromagnetic induction, while effective, often require expert interpretation and can be time-consuming. Misinterpretation or incomplete data can lead to costly mistakes, such as accidental damage to critical infrastructure. To address these challenges, this paper introduces a deep learning-based approach to automate the identification and localization of subterranean water pipes, gas lines, and electrical conduits, offering a faster, more reliable, and scalable solution.

Our method leverages convolutional neural networks (CNNs), specifically YOLO (You Only Look Once) and Faster R-CNN, trained on a carefully curated dataset of subsurface images obtained from GPR scans and electromagnetic sensors. These models are optimized to recognize complex patterns and features that signify underground objects, ensuring precise detection and localization. Once an object is identified, geometric fitting algorithms, such as the Hough Circle Transform and RANSAC (Random Sample Consensus), are employed to estimate its diameter and cross-sectional shape. This additional layer of analysis helps infer the type of utility line—smaller diameters may indicate water pipes or fiber optic cables, while larger cross-sections could suggest main gas or sewer lines. The ability to extract both positional and physical attributes of underground objects adds a crucial dimension to utility mapping.

A key aspect of this research is the validation of the proposed model using real-world data from pre-determined areas, supported by a company like DEWA (Dubai Electricity and Water Authority). By cross-referencing the detected objects with known underground infrastructure layouts, the system’s accuracy and reliability are assessed and refined. This step not only strengthens the model's predictive capabilities but also highlights its practical applicability for fieldwork. The integration of deep learning with geometric analysis creates a robust framework capable of handling diverse subsurface environments and varying object sizes.

Furthermore, this approach is designed with scalability in mind. As the model processes more real-world data, its accuracy improves through continuous learning. In future work, we plan to enhance the system's responsiveness by deploying it on edge computing devices, enabling real-time detection and analysis directly at excavation sites. This advancement could significantly reduce downtime and operational risks associated with underground utility maintenance.

Ultimately, this research bridges the gap between AI and infrastructure management, offering a modern solution for underground utility detection. By automating both the identification and characterization of subsurface objects, our model empowers planners and engineers to make informed decisions, improving safety, efficiency, and project outcomes.

Presenting Author: Mahmoud Rezk Dubai Electricity & Water Authority

Presenting Author Biography: Mahmoud Rezk is a senior research engineer at DEWA (Dubai Electricity and Water Authority), specializing in the application of artificial intelligence and machine learning to infrastructure optimization and energy solutions. With a background in software engineering and a focus on AI-driven modeling, Mahmoud has worked on projects involving predictive maintenance, renewable energy systems, and intelligent inspection methods using drones and sensor technologies. His recent research explores the integration of deep learning algorithms with dynamical systems to enhance underground utility detection, contributing to smarter urban planning and infrastructure management.

Authors:

Mahmoud Rezk Dubai Electricity & Water Authority
Mariam Alnaqbi Dubai Electricity and Water Authority
Thani Althani Dubai Electricity and Water Authority
Mohit Vohra Dubai Electricity and Water Authority