A low-cost and accurate computer vision system for rapidly estimating pomelo weight based on multi-view imaging technique
|
Full Text |
Pdf
|
|
Author |
Nhut-Thanh Tran, Huu-Nhon Nguyen, Thanh-Thien Ngo, Huu-Tinh Phan, and Chanh-Nghiem Nguyen
|
|
e-ISSN |
1819-6608 |
|
On Pages
|
813-818
|
|
Volume No. |
20
|
|
Issue No. |
12
|
|
Issue Date |
September 15, 2025
|
|
DOI |
https://doi.org/10.59018/062596
|
|
Keywords |
fruit weight estimation, multi-view image, irregular shape, vision-based system, global thresholding technique, simple linear regression.
|
Abstract
Weight is one of the key external quality indicators that directly affects the price of fruit. Although computer vision is considered an effective solution for rapidly and accurately estimating fruit weight, its high investment cost and the complexity of image processing remain major obstacles to its widespread adoption, particularly among small and medium-sized enterprises. This study proposes a low-cost and accurate computer vision system for rapid fruit weight estimation based on a multi-view imaging technique. A cost-effective imaging solution was developed to capture a multi-view image of the fruit within a single shot using one camera combined with commonly available flat mirrors. A simple data processing pipeline was implemented, which involves image segmentation using the global thresholding technique and weight estimation through a simple linear regression model. The proposed system was tested on Da Xanh pomelo, a species of fruit with a highly irregular shape. Regression analysis results indicate that the proposed system achieved highly accurate weight estimation, with a coefficient of determination of 0.98 and a mean absolute percentage error of 1.9%. Additionally, the paired t-test results demonstrate that the estimated weight obtained using the proposed system shows no significant difference from the actual weight measured with a commercial digital scale. These findings highlight the significant potential of the proposed system in providing a rapid, accurate, and cost-effective solution for fruit weight estimation.
Back