Quantitative prediction of post storage ‘Hayward’ kiwifruit attributes using at harvest Vis-NIR spectroscopy

• Published in: Journal of food engineering 2017-06, Vol.202, p.46-55
• Main Authors: Li, Mo, Pullanagari, Reddy R., Pranamornkith, Thamarath, Yule, Ian J., East, Andrew R.
• Format: Article
• Language: English
• Subjects: Actinidia deliciosa; Non-destructive; Quality prediction; Regression model
• ISSN: 0260-8774; 1873-5770
• DOI: 10.1016/j.jfoodeng.2017.01.002

Total soluble solids concentration (TSS) and flesh firmness (FF) are two important quality attributes indicating the eating quality and postharvest storability of kiwifruit. Prediction of TSS and FF using non-destructive techniques would allow strategic marketing of fruit. This paper investigates the ability of visible-near-infrared (Vis-NIR) spectroscopy utilised as a sole input at harvest, to quantitatively predict both TSS and FF after cool storage using a blackbox model. Four at-harvest Vis-NIR spectral and post-storage fruit quality data sets were collected during 2012–2013, in order to develop regression models using partial least squares (PLS) and support vector machines (SVM). The SVM models performed better than PLS. Predictive accuracy was fair to good for TSS (R2 = 0.58–0.83; RMSE = 0.66–1.02 °Brix) and was poor to moderate for FF (R2 = 0.30–0.60; RMSE = 2.65–4.32 N). The ratio of prediction deviation, SDR values (1.5–2.3 for TSS; 1.4–1.7 for FF) suggest the developed regression models are not as yet useful for online sorting purposes. •Quantitative prediction of total soluble solids and flesh firmness was achieved.•Predictive accuracy of regression model for total soluble solids was fair to good.•Predictive accuracy of regression model for flesh firmness was poor to moderate.•Reducing prediction error would be required for online sorting purposes.