Multi-scale ROI-based SEM image analysis of processing time and magnification effects in lignocellulosic biomass.

Abstract

Quantitative characterization of treatment-induced microstructural evolution in lignocellulosic biomass using scanning electron microscopy (SEM) is challenging because image-derived descriptors are affected by both processing-induced structural changes and imaging magnification, limiting direct comparison of measurements. This study presents an ROI-based multi-scale SEM image analysis methodology for evaluating process-induced microstructural evolution and magnification-dependent descriptor behavior, demonstrated using ball-milled mustard seed husk as a representative case study. GLCM (gray-level co-occurrence matrix) texture features, including entropy (spatial randomness in gray-level relationships) and contrast, and pore characteristics were quantified from SEM images obtained at different milling stages and magnifications. Increasing GLCM entropy, contrast, and pore count, together with decreasing mean pore area, indicated progressive surface fragmentation. Entropy showed relatively low magnification sensitivity, whereas contrast and pore count were strongly scale-dependent. Pore area showed convergence at selected scales, indicating improved measurement consistency. Two-way analysis of variance (ANOVA) showed significant effects of milling time and magnification on all descriptors (p<0.001); the interaction was significant for contrast, entropy, and pore count, but not for pore area (p=0.150). The proposed framework provides a standardized approach for quantitative SEM characterization of heterogeneous lignocellulosic microstructures by considering both processing duration and imaging magnification.