SAAIL Lab Publications
Int. J. Food Properties · 2021 46+ Citations Prior Institutional Research · Pre-IITMZ (2021)

Weight and volume estimation of single and occluded tomatoes using machine vision

Innocent Nyalala 1 , Cedric Okinda 1 , Qi Chao 1 , Peter Mecha 1 , Tchalla Korohou 1 , Yi Zuo 1 , Samuel Nyalala 1 , Jiayu Zhang 1 , Chao Liu 1 , Kunjie Chen 1
1College of Engineering, Nanjing Agricultural University, Nanjing, China
International Journal of Food Properties (Taylor & Francis), Vol. 24, pp. 818–832 (Jun 2021)

Summary: Extends non-contact machine vision mass and volume estimation from single cherry tomatoes to partially occluded fruit clusters (46+ citations).

Abstract

Fruit occlusion severely degrades geometric volume and mass prediction in automated sorting lines. We combine depth imaging, morphological contour reconstruction, and machine learning regression to recover accurate mass and volume estimates for both isolated and partially occluded tomatoes.

Keywords: tomato volume estimationocclusion handlingmachine visiondepth imagingpostharvest

Key Findings & Contributions

R² = 0.982
Volume (RBF-SVM)
RMSE = 1.2683 cm³
R² = 0.971
Weight (Bayesian ANN)
RMSE = 1.468 g
7 Models
Regressors Tested
Single & occluded fruits
46+ Cites
Food Properties J.
Vol. 24, pp. 818-832
  1. Solved partial tomato occlusion on moving sorting belts via polygon approximation and concave/convex contour point extraction.
  2. Benchmarked seven regression algorithms across 2D and depth features, achieving R² = 0.971 (RMSE 1.468 g) for mass via Bayesian regularization ANN.
  3. Reached R² = 0.982 (RMSE 1.2683 cm³) for non-destructive volume prediction using RBF-SVM across isolated and clustered fruits.

Benchmark Results

Single & Partially Occluded Tomato Regression Results (2021)

Prediction TargetBest ModelR²RMSE
Fruit Volume (cm³)RBF Support Vector Machine (SVM)0.9821.2683 cm³
Fruit Weight (g)Bayesian Regularized ANN0.9711.468 g

BibTeX

@article{nyalala2021occludedtomatoes,
  title={Weight and volume estimation of single and occluded tomatoes using machine vision},
  author={Nyalala, Innocent and Okinda, Cedric and Chao, Qi and Mecha, Peter and Korohou, Tchalla and Zuo, Yi and Nyalala, Samuel and Jiayu, Zhang and Liu, Chao and Chen, Kunjie},
  journal={International Journal of Food Properties},
  volume={24},
  number={1},
  pages={818--832},
  year={2021},
  doi={10.1080/10942912.2021.1933024}
}
APA Citation

Nyalala, I., Okinda, C., Chao, Q., Mecha, P., Korohou, T., Zuo, Y., Nyalala, S., Jiayu, Z., Liu, C., & Chen, K. (2021). Weight and volume estimation of single and occluded tomatoes using machine vision. International Journal of Food Properties, 24(1), 818–832. https://doi.org/10.1080/10942912.2021.1933024