About this role
University of Twente Jobid=637868401911922981 (0.05) In this EngD project, you will develop an AI model that automatically detects underground infrastructure in GPR radargrams and estimates its depth. The project builds on the growing availability of high-quality GPR data collected at the University of Twente's Utility Mapping Site (UMS), a unique test environment for utility mapping technologies. Current machine learning models and their training data are limited in size, comprehensiveness, and realism - resulting in partial automation with limited performance. This constrains their usefulness in real-world conditions. Your challenge is to develop and validate machine learning models using systematically collected and accurately annotated GPR datasets. By combining geospatial data, subsurface sensing, and AI, you will contribute to the next generation of utility mapping technologies and support safer excavation practices. Your environment This project is part of the ZoARG|ReDUCE programme, a collaborative initiative aimed at minimizing excavation damage to underground infrastructure in the Netherlands. You will work within a multidisciplinary environment that includes: - The University of Twente's Departments of Civil Engineering and Management (CEM) and Applied Earth Sciences (AES) - The Utility Mapping Site (UMS) at the UT FieldLab - Industry collaborators involved in the ZoARG programme What you will do - Analyse existing GPR interpretation methods, machine learning techniques, and relevant software tools - Explore and evaluate AI approaches for automated utility characterization - Prepare, preprocess, and manage large GPR datasets collected at the Utility Mapping Site - Design, develop, train, and validate machine learning models for interpreting GPR radargrams - Compare developed models with existing approaches reported in literature and commercial software solutions - Work in close partnership with infrastructure owners,...