A novel image processing and machine learning approach for wood pellet size classification using shadow analysis

The growing importance of wood pellets in renewable energy highlights the need for efficient, scalable methods of quality assessment. Current standards rely on manual caliper measurements, which are time-consuming and poorly capture variability across samples. This study introduces a patent-pending imaging approach that classifies pellet dimensions from shadow features. A prototype system was developed combining controlled lighting, a camera, and computational processing. Shadow characteristics were analyzed statistically and linked to pellet dimensions using machine learning models. Results show that illumination geometry strongly influences classification, with the best performance reaching 71 % accuracy. These findings demonstrate the feasibility of shadow-based imaging as a rapid, non-invasive alternative to manual measurement, with promising applications in pellet production, storage, and combustion systems.