I-BALLAST: Computer vision solutions for ballast degradation analysis using deep-learning and data fusion methods

作者：Kelin Ding, Jiayi Luo, Haohang Huang, John M. Hart, Issam I. A. Qamhia, Erol Tutumluer, Hugh Thompson, Theodore R. Sussmann · 发表于：Geodata and AI. · 年份：2024 · DOI：10.1016/j.geoai.2025.100007 · 被引用次数：3 · 研究领域：Advanced Neural Network Applications、Robotics and Sensor-Based Localization、Remote Sensing and LiDAR Applications

Ballast plays a significant role in supporting tracks under repeated loading. Ballast Scanning Vehicle (BSV) is an automated platform recently developed to acquire high-quality data to assist with ballast inspection and detailed geotechnical analyses using computer vision technology. To process the image data collected by the BSV, this paper introduces an advanced image analysis software for ballast, i.e., I-BALLAST. I-BALLAST encompasses advanced data fusion techniques that integrate multiple data sources, thus producing more consistent, accurate, and comprehensive information. Field data collected by multiple image acquisition devices and other sensors like Global Positioning System (GPS) can be combined for conducting comprehensive analyses, such as segmented particle point clouds generation and varying-window-based Particle Size Distribution (PSD) and Fouling Index (FI) analyses along track sections with aligned GPS coordinates. I-BALLAST has been tested in multiple field inspections, proving it an efficient and innovative tool for ballast fouling condition evaluations.