Abstract:

Estimating potato tuber mass and size grading with computer vision can help breeders, farmers, and potato processing units reduce manual labor for potato post-harvest handling through optimized technology. The objective of the study was to estimate potato tuber mass and size grades using 2D images. Physical data of potato tubers from 23 different cultivars were collected, and their 2D images were captured under controlled light conditions from top and side angles. The physically measured, volume-based, and image-extracted features were used to employ linear and stepwise regression models to estimate the tuber mass. These models were trained on 14 and tested on 9 different cultivars. In the second step, a random forest classification model was developed to grade the potato tubers based on image-extracted tuber width dimensions from the top, side, and both angles. Classification data was divided into 80% training and 20% test data, where the training process was conducted with 10-fold cross-validation with 5 replications, and the models were evaluated on test data. The tuber mass estimation was higher when combined with the image-extracted features from both angles (R2 = 0.99), followed by an volume based on the image-extracted including all geometric dimensions from both angles (R2 = 0.98) and top angle image-extracted features based stepwise regression (R2 = 0.98). The classification was 100% accurate when trained and tested using top and side widths. Future work is also required to train and test the model for the individual cultivar for higher model precision and robustness for popular potato cultivars.

Links:

Citation:

A. K. Sharma, L. Zotarelli, A. Zare, and L. K. Sharma, “Automated potato tuber mass estimation and grading with multiangle 2D images,” Smart Agricultural Technology, vol. 10, p. 100832, 2025. doi: 10.1016/j.atech.2025.100832.

@article{sharma2025automated,
title={Automated Potato Tuber Mass Estimation and Grading with Multiangle 2D Images},
author={Sharma, Ayush K and Zotarelli, Lincoln and Zare, Alina and Sharma, Lakesh K},
journal={Smart Agricultural Technology},
pages={100832},
year={2025},
publisher={Elsevier}
}