Sigmoid displacement profile and entropy-based computer vision analysis for shear-band in granular materials

Shear bands govern the stability of slopes, tunnels and foundations, yet there is no simple and noise-tolerant way to locate a band and quantify its thickness across tests and simulations. We combine triaxial DEM on dense and loose granular assemblies with a compact sigmoid profile of band-normal displacement that enforces boundary-consistent limits and returns two interpretable parameters: band center and thickness. The profile fits particle-scale fields with high fidelity (up to R 2 = 0.97) and enables stage-by-stage tracking of localization. To bridge to laboratory and field images, we develop an entropy-based image workflow (computer vision) that suppresses background noise and objectively delineates bands; linear features extracted from DEM and sigmoid-processed images show close agreement (R 2 ≈ 0.92–0.94). Energy analyses clarify how friction stabilizes force chains and shifts peak deviatoric strength, explaining contrasting localization paths in dense versus loose packings. The approach provides a practical toolbox for rapid, comparable characterization of shear-band geometry in granular materials. • Sigmoid displacement profile quantifies shear-band location and thickness. • Entropy-based computer vision robustly detects shear-band under noisy backgrounds. • DEM–CV agreement: R 2 ≈ 0.92–0.94 (linear fit); band geometry matches. • Displacement-field fit up to R 2 = 0.97; boundary displacement limits satisfied. • Higher friction stabilizes force chains; peak strength rises and peaks later.