Congratulations to Mingjian Wu for successfully defending his PhD dissertation 

Mingjian Wu successfully defended his PhD dissertation on November 28th, 2024.

Dissertation Title:

Advancing Winter Road Surface Condition Monitoring Through Deep Learning and Geostatistics

Dissertation Description:

Monitoring road surface conditions (RSC) during winter is critical for road safety and effective winter road maintenance (WRM). Current methods, like Road Weather Information Systems (RWIS), are costly and lack detailed spatial coverage. This thesis addresses these gaps by combining deep learning (DL) and geostatistics to improve RSC monitoring. DL-based techniques, including convolutional neural networks (CNNs), pix2pix generative adversarial networks (GANs) and semantic segmentation, achieved high accuracy (up to 99.3%) in automating RSC recognition and estimating snow coverage. Geostatistical methods like regression kriging (RK) and nested indicator kriging (NIK) provided accurate interpolation of RSC data between RWIS stations. Tested on over 20,000 images and 5 million weather measurements from Iowa highways, these methods reduced the need for additional RWIS infrastructure, improved WRM operations, and were implemented in a real-time web application. The innovations enhance road safety, mobility, and sustainability during winter conditions.

A part of the research output was also integrated into LoRWIS (https://vstfl.github.io/mapbox-rsi/), a AI-driven application for real-time winter RSC monitoring and estimation (developed with Michael Urbiztondo and Dr. Tae J. Kwon).

Committee Members:

Dr. Tae. J. Kwon (Supervisor), Dr. Karim El-Basyouny (Supervisory Committee Member), Dr. Tony Qiu (Supervisory Committee Member), Dr. Luis Miranda-Moreno (Examiner), Dr. Qipei (Gavin) Mei (Examiner), and Dr. Juliana Leung (Chair)