Near infrared hyperspectral imaging for kiwifruit quality and maturity assessment during postharvest

This study employed near-infrared hyperspectral imaging (858–1700 nm) for the non-destructive assessment of soluble solids content (SSC) and maturity classification in 480 kiwifruits during postharvest storage at room temperature (20 ± 1 °C) over a 14-day period.

The effectiveness of six spectral preprocessing methods was evaluated, including multiplicative scatter correction (MSC), standard normal variate (SNV), Savitzky-Golay smoothing (SG), and their combinations MSC+SNV, MSC+SG, and SNV+SG. Additionally, three wavelength selection algorithms—CARS, SPA, and UVE—were compared.

Four predictive models were then developed and assessed: partial least squares regression (PLSR), backpropagation neural network (BP), linear regression (LR), and least squares support vector machine (LSSVM). The MSC+SG+CARS+LSSVM model achieved the best performance, with correlation coefficients Rc=0.881 and Rp=0.927, RMSEC=0.590 °Brix, RMSEP=0.597 °Brix, and RPD=2.61.

Furthermore, storage day classification models based on genetic algorithm-optimized BP (GA-BP) and radial basis function (RBF) neural networks achieved accuracies of 95.15% and 93.75%, respectively.

Overall, the findings confirm that near-infrared hyperspectral imaging is an effective tool for assessing internal quality and postharvest storage duration of kiwifruit, supporting optimal transport timing and offering strong potential for quality monitoring and maturity classification in the industry.

Source

Li, Y., Qiao, Y., Zhu, R., Wang, J., Sun, Z., Yang, S., Ai, Z., & Song, S. (n.d.).
Non-destructive prediction of SSC and storage days classification of kiwifruit during postharvest storage using near-infrared hyperspectral imaging. Springer.

https://link.springer.com/article/10.1007/s12161-026-03108-6