Articles
Evaluation of fruit growth using image AI in Satsuma mandarin and its relationship with water stress and fruit sugar content increase
Article number
1468_10
Pages
67 – 72
Language
English
Abstract
High-sugar Satsuma mandarin are favored by consumers and receive a high unit price.
Water stress during summer and autumn increases the sugar content of the fruit, however, excessive water stress leads to a decline in tree vigor and a reduction in fruit size.
Farmers use calipers to regularly measure fruit diameter and calculate the rate of fruit growth by hand, a laborious process, to maintain the appropriate water stress.
We developed a method to estimate fruit diameter using a smartphone as an easy-to-use field-based technology and used this method to investigate the relationship between fruit growth, water stress, and changes in sugar content.
To estimate fruit diameter from the monocular camera, a white square piece of plastic 15 mm per side was attached to the fruit apex and photographed with the wide-angle camera of an iPhone13 pro.
The area of the fruit and the piece of plastic was extracted from the image using a trained Mask R-CNN deep learning module.
The fruit diameter was estimated based on a regression model with the area measurement extracted from the image.
Measured data included water stress, sugar content, and fruit diameter data measured periodically by calipers.
The coefficient of determination between actual and estimated fruit diameter was 0.988, indicating that the method can estimate fruit diameter with high accuracy.
The relationship between the amount of fruit growth calculated using the fruit diameter estimation method and the amount of accumulated water stress and sugar increase were also higher than the amount of fruit size calculated using caliper measurements, with coefficients of determination of 0.78 and 0.59, respectively.
These results indicate that this method may be more effective than calipers in estimating water stress and the increase in sugar concentration.
Water stress during summer and autumn increases the sugar content of the fruit, however, excessive water stress leads to a decline in tree vigor and a reduction in fruit size.
Farmers use calipers to regularly measure fruit diameter and calculate the rate of fruit growth by hand, a laborious process, to maintain the appropriate water stress.
We developed a method to estimate fruit diameter using a smartphone as an easy-to-use field-based technology and used this method to investigate the relationship between fruit growth, water stress, and changes in sugar content.
To estimate fruit diameter from the monocular camera, a white square piece of plastic 15 mm per side was attached to the fruit apex and photographed with the wide-angle camera of an iPhone13 pro.
The area of the fruit and the piece of plastic was extracted from the image using a trained Mask R-CNN deep learning module.
The fruit diameter was estimated based on a regression model with the area measurement extracted from the image.
Measured data included water stress, sugar content, and fruit diameter data measured periodically by calipers.
The coefficient of determination between actual and estimated fruit diameter was 0.988, indicating that the method can estimate fruit diameter with high accuracy.
The relationship between the amount of fruit growth calculated using the fruit diameter estimation method and the amount of accumulated water stress and sugar increase were also higher than the amount of fruit size calculated using caliper measurements, with coefficients of determination of 0.78 and 0.59, respectively.
These results indicate that this method may be more effective than calipers in estimating water stress and the increase in sugar concentration.
Authors
R. Komiya, M. Iwasaki, H. Habaragamuwa, H. Fukamachi
Keywords
fruit size, image analysis, machine learning, water stress
Groups involved
Online Articles (70)