New approach of simultaneous, multi‐perspective imaging for quantitative assessment of the compactness of grape bunches

• Published in: Australian journal of grape and wine research Vol. 24; no. 4; pp. 413 - 420
• Main Authors: Chen, X., Ding, H., Yuan, L.‐M., Cai, J.‐R., Lin, Y.
• Format: Journal Article
• Language: English
• Published: Melbourne John Wiley & Sons Australia, Ltd 01.10.2018; John Wiley & Sons, Inc
• Subjects: Automation; Bulk density; bunch compactness; Correlation coefficient; Correlation coefficients; Data analysis; Data processing; Digitization; Feasibility studies; feature extraction; Fruits; Grapes; image analysis; Image processing; Least squares; Mathematical models; Multivariate analysis; multi‐perspective imaging; Performance prediction; prediction; Regression analysis; tablegrape; Vineyards; Vitis labrusca; Wine; wine quality; Wines
• ISSN: 1322-7130; 1755-0238
• DOI: 10.1111/ajgw.12349

Background and Aim The compactness of a grape bunch can be a significant trait in determining tablegrape and wine quality. The Organisation Internationale de la Vigne et du Vin developed the most widely used visual method to assess bunch compactness from loose (1) to tight (9). This method, however, requires training and relies on manual measurements and is, thus, subject to bias and error. The aim of this study was to test the feasibility of multi‐perspective imaging analysis combined with multivariate modelling to predict grape bunch compactness. Such a method has the potential to be rapid, automated and objective. Methods and Results Vitis labruscana cv. Kyoto grape bunches were collected from three vineyards over two consecutive seasons and imaged with a multi‐perspective imaging system, which sensed mass and imaged the surface of the bunch from three perspectives using mirror reflection. Bulk density of the grape bunch was linearly related to compactness (correlation coefficient 0.679). The morphological features of grape bunches and their derivative variables were digitised using 23 image processing descriptors and were regressed with the measured compactness using multivariate data analysis, including partial least squares, multiple linear regression and principal component regression. The partial least squares model was the best performed, predicting bunch compactness with a correlation coefficient of prediction (rp) of 0.8481 as well as root mean squared error of prediction of 1.2287. Conclusions Multi‐perspective imaging combined with image processing and multivariate data analysis can assess the compactness of grape bunches. Significance of the Study The performance of this multi‐perspective imaging method could be developed to automate the postharvest assessment of the compactness of grape bunches.