Image processing techniques to identify tomato quality under market conditions

• Published in: Smart agricultural technology Vol. 7; p. 100433
• Main Authors: Abekoon, Thilina, Sajindra, Hirushan, Jayakody, J．A．D．C．A．, Samarakoon, E．R．J, Rathnayake, Upaka
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01-03-2024; Elsevier
• Subjects: Classification; Convolutional neural network; Image processing; Machine learning; Post harvest technology; Tomato; CNN; Image processing; Post harvest technology; Convolutional neural network; Machine learning; Classification; ReLU; Tomato
• ISSN: 2772-3755; 2772-3755
• DOI: 10.1016/j.atech.2024.100433

•Cutting-edge image processing and AI techniques used for tomato quality assessment.•CNNs were used to accurately determine tomato maturity stages and post-harvest dates.•Established classification methods for better edible tomato identification were used.•Insights from this research optimize tomato production and quality management.
Tomatoes are essential in both agriculture and culinary spheres, demanding rigorous quality assessment. It is highly advantageous to discern the maturity level and the time range post-harvesting of tomatoes in the market through visual analysis of their images. This research endeavors to forecast tomato quality by accurately determining the maturity level and the duration post-harvest, specifically tailored to Sri Lankan market conditions, with a particular focus on Padma tomatoes. It identifies maturity stages (Green, Breakers, Turning, Pink, Light Red, Red) and post-harvest dates using image processing techniques. Greenhouse-grown Padma tomatoes mimic market conditions for image capture, and Convolutional Neural Networks facilitate this analysis. Model 1, using ReLU and sigmoid activation functions, accurately classifies tomatoes with 99 % training and validation accuracy. Model 2, with seven classes, achieves 99 % training and 98 % validation accuracy using ReLU and softmax activation functions. Integration of the IPGRI/IITA 1998 classification method enhances tomato categorization. Efficient tomato image screening optimizes resource use. This study successfully determines Padma tomato post-harvest dates based on maturity stages, a significant contribution to tomato quality assessment under market conditions.