Honors Theses
Advisor
Dr. Theus Aspiras
Department
Electrical and Computer Engineering
Publication Date
4-22-2026
Document Type
Honors Thesis
Abstract
This work presents a computer vision-based approach for automated ripeness detection in tomato plants. The system leverages a YOLOv8 segmentation model trained on a publicly available dataset to detect and isolate individual tomatoes within complex visual environments. While segmentation provides an initial estimate of tomato regions, additional processing is required to remove non-fruit artifacts such as stems and leaves that may interfere with analysis. To address this, an edge-constrained region growing method is used to tighten segmentation masks by removing unwanted structures such as stems and leaves while preserving the underlying fruit. Ripeness is then estimated using a color-based analysis in the HSV color space. Pixels are grouped into green, yellow, and red categories, and a weighted scoring approach is used to produce a continuous measure of ripeness progression. The proposed system demonstrates reliable performance across a range of tomato appearances, capturing gradual transitions in ripeness while maintaining robustness to moderate variations in lighting and occlusion. Although limitations remain in artifact removal and sensitivity to environmental conditions, the results highlight the potential of combining deep learning and classical image processing techniques for automated agricultural monitoring. This work provides a foundation for future systems investigating harvest timing and enabling more autonomous food production.
Permission Statement
This item is protected by copyright law (Title 17, U.S. Code) and may only be used for noncommercial, educational, and scholarly purposes.
Keywords
Undergraduate research
eCommons Citation
Ebbing, Austin, "AI-Driven Computer Vision for Automated Ripeness Detection in Tomato Plants" (2026). Honors Theses. 504.
https://ecommons.udayton.edu/uhp_theses/504
COinS
