Integrating color classification and circularity metrics in wool fabric recycling

This work proposes a methodological framework for circularity-aware color classification in textile recycling. The approach combines machine vision data, spectral reflectance information, and process-related data within a unified architecture. The main aim is the definition of a circular color sorting index able to link classification outcomes to manufacturing feasibility and environmental impact, particularly in terms of avoided post classification operations. Unlike traditional performance metrics, the defined index incorporates classification confidence, chromatic proximity, environmental coefficients, and material reuse potential into a single indicator. The proposed framework is validated on a dataset of recycled wool fabrics representative of typical industrial conditions for companies working with such fabrics. Despite the experimental results are intended as a proof of concept, this work shows that models optimized for circularity-aware scoring can achieve substantially improved environmental performance at batch level, even when raw classification accuracy improvements are moderate.