VISION-BASED ROBOTIC BANANA SORTING SYSTEM FOR QUALITY CONTROL IN KETCHUP MANUFACTURING USING YOLOV11 (Preview)
Consistent raw material quality is a critical prerequisite for the reliable manufacturing of banana ketchup, particularly when utilizing the Saba banana variety. Conventional manual sorting workflows are labor-intensive, error-prone, and insufficiently scalable to meet industrial production demands. This paper proposes and evaluates an automated vision-based robotic sorting system designed to enforce quality control in compliance with the Philippine National Standards (PNS). The system integrates a six-degree-of-freedom (6DOF) robotic arm, an Arduino Mega 2560 microcontroller, and a computer vision module employing the YOLOv11 object detection algorithm. Bananas are classified across four freshness categories, ripe, unripe, overripe, and defective, enabling precise acceptance or rejection decisions prior to processing. Among the evaluated YOLOv11 variants, the medium-scale YOLOv11m model demonstrated optimal performance, achieving 99.8% precision, 99.4% recall, and a Mean Average Precision (mAP@0.5) of 99.4% after 157 training epochs. Physical sorting trials validated the seamless integration of the vision pipeline with the robotic actuation system, confirming accurate pick-and-place operations without mechanical damage to the fruit. The proposed framework significantly reduces dependence on manual inspection, mitigates human error, and establishes a scalable foundation for intelligent automation in agricultural food processing.
Keywords: Computer Vision, Robotic Arm, YOLOv11, Banana Sorting, Quality Control, Agricultural Automation.