Assessing Ground Settlement Using Computer Vision and Artificial Intelligence in Geotechnical Engineering

Ground settlement is an important geotechnical concern because excessive or differential deformation can affect the serviceability and safety of foundations and structures. This paper outlines a proposed methodology for combining computer vision with machine-learning techniques to monitor and estimate settlement in laboratory soil tests. Sequential digital images are used to track surface markers, while camera calibration converts image displacement into physical units. Loading conditions and relevant soil properties may then be used as input variables for a regression model. Root mean square error is proposed as one measure for evaluating predictions against independent measurements. The work is methodological and does not report experimental results, because no test data were provided. The proposed workflow requires validation against conventional displacement measurements and careful control of camera movement, lighting, perspective and soil-surface conditions.