Rice is a critical component of the world’s food supply chain. It is consumed after milling, but the milling process significantly impacts rice quality, with broken grains posing a major challenge for the industry. Traditional quality assessment methods, such as manual inspection, are inefficient and inconsistent, necessitating advanced solutions for accurate classification and morphological analysis. This paper presents an AI-driven machine vision system leveraging the You Only Look Once version 8 (YOLOv8) deep learning model to automate real-time rice quality assessment. A custom dataset of rice grain images was collected, pre-processed, and used to train the model for morphological feature extraction and kernel classification. Morphological features such as shape, size, and colour are extracted to evaluate milling yield, while classification assigns each grain to one of three categories: Good Rice, Broken Rice, or Brown Rice. Experimental results demonstrate the system’s robustness, achieving 98% classification accuracy and precise morphological analysis. The model also generates statistical data for condition monitoring and fault detection. By integrating deep learning with machine vision, this approach enables fast, consistent, and automated quality control, reducing human intervention and standardising rice milling processes. The system’s industrial applicability lies in its ability to enhance efficiency, minimise waste, and ensure compliance with global quality standards.