Channel-dependent polarization imaging for cultivar-specific apple freshness assessment

Abstract Amidst continuously improving living standards, there is a growing demand for non-destructive testing of the internal quality of food products. Techniques based on detecting the optical absorption or scattering properties of fruits represent a significant area of this research. This paper proposes a method for predicting apple freshness using scattered polarization information. Polarization images captured by a CMOS sensor were decomposed into three channels via RGB color space. Key parameters of the polarization images were calculated, including the degree of linear polarization, mean gradient, variance, mean value, information entropy, angular second moment, Contrast, and Correlation. The results revealed strong linear correlations (R 2 ) between these features and storage time, specifically in the green channel for Red Star and the red channel for Golden Delicious. Multivariate regression models were established for each cultivar, achieving coefficients of determination (R 2 ) of 0.9857 and 0.967 47, respectively, significantly outperforming single-parameter models. Validation experiments with new specimens confirmed the robustness of the approach, showing high reproducibility with Pearson correlation coefficients of 0.9759 for two Red Star apples and 0.9954 for two Golden Delicious apples. This research establishes a reliable, cultivar-specific framework for non-destructive apple freshness monitoring.