Data-Driven Real-Time Rice Milling Optimisation via YOLO26 Machine Vision and Adaptive Closed-Loop Motor Control

Area of Science:

• Agricultural Engineering
• Computer Vision
• Process Control

Background:

• Conventional rice milling quality assessment is post-process, preventing real-time adjustments and leading to increased breakage.
• Lack of immediate feedback hinders optimization of milling parameters to minimize broken rice kernels.

Purpose of the Study:

• To develop and validate a cloud-mediated, closed-loop control system for optimizing rice milling quality.
• To quantitatively assess the effectiveness of a machine vision pipeline coupled with actuator control in reducing rice breakage.

Main Methods:

• A YOLOv26 machine vision pipeline was integrated with Arduino-based actuator control on a laboratory rice mill.
• An image-acquisition node uploaded frames to the cloud for analysis, where a hybrid classifier estimated the broken-rice fraction.
• Control commands were sent to Arduino microcontrollers driving PWM-modulated motor and vibrator actuators.

Main Results:

• The YOLOv26 detector achieved a mean Average Precision of 0.951 on a test set.
• Closed-loop operation reduced the broken rice fraction from 21.04% to 6.19% compared to an open-loop baseline.
• Dynamic analysis revealed a near-linear plant gain of 1.0-1.5% breakage per 1% PWM.

Conclusions:

• The developed closed-loop system effectively improves rice milling quality, achieving Grade A-equivalent standards.
• The study provides a quantitative characterization of the PWM-to-breakage transfer relationship for rice milling actuators.
• A roadmap for scaling the technology from lab-prototype to pilot-scale deployment is outlined.