Integration of Computer Vision and 4DOF SCARA Robot Arm for Tomato Sorting

Automating fruit recognition and classification is an important computer vision application that can improve efficiency and consistency in the fruit industry. Manual sorting methods based on visual inspection are tedious, time-consuming, and prone to inconsistencies. This work presents the design and deployment of an automated tomato sorting system that integrates computer vision and robotics. The system classifies tomatoes into three maturity stages (Green, Half-ripe, Ripe) based on color using the YOLOv5 deep learning algorithm. A SCARA robot is used for picking and placing the sorted tomatoes into respective bins. The system demonstrates a fast inference time of 22.31 milliseconds for tomato detection and can sort three tomatoes of different maturity classes in approximately 1.2 seconds on average. This consistent and efficient sorting is based on objective color criteria rather than manual visual inspection. The system can help improve product quality, reduce labor costs, and increase throughput in tomato processing facilities. The approach can be extended to sorting other fruits and vegetables based on color attributes.