Design and Development of Automatic Tomato Sorting System using Embedded Electronics and Machine Learning

The increasing demand for high-quality tomatoes and price fluctuation have necessitated the adoption of automated sorting systems that ensure accuracy, consistency, and efficiency in post-harvest operations. Thus, this research aimed to design and develop an automatic tomato sorting system that integrates embedded electronics, image processing, and machine learning. A prototype was built using the Arduino Uno microcontroller, a colour sensor, an infrared sensor, and servo motors, with a facilitating image capture and processing. Tomatoes were sorted by colour/ripeness and categorised into three groups: red (ripe), yellow (semi-ripe), and green (unripe) using a convolutional neural network trained on a custom dataset. Image preprocessing techniques, including pixel normalisation and histogram equalisation, were employed to enhance feature extraction. The developed system demonstrated classification accuracy of up to 97.9%, signifying robustness under varied lighting conditions. Besides, the system is suitable for practical implementation in small-scale agricultural operations. This scalable solution promises to reduce manual labor, improve sorting precision, and minimise post-harvest losses, making it a valuable tool for enhancing the tomato supply chain.

View article page

Download PDF