Cloud Based AI System for Food Grain Quality and Safety Monitoring in Public Kitchen and Ration Depots

Abstract

Introduction: Food safety in public kitchens and ration depots is essential to protecting public health and maintaining citizen trust. Traditional manual inspection of food grains is often inconsistent, slow, and prone to human error. This paper introduces a cloud-based AI system that automates grain quality and contamination detection using computer vision and deep learning. Images captured at distribution centers are analyzed through edge-cloud collaboration, enabling real-time grading and safety alerts. The proposed framework ensures scalability, transparency, and data integrity through Zero Trust security principles and sovereign government cloud infrastructure.

Objectives: The main objective of this paper is to design and implement an automated method for grading and quality assessment of rice grains using image processing and AI techniques, extending earlier approaches that relied on morphological analysis and manual inspection. The system aims to replace subjective human grading with a cloud-integrated, data-driven framework that analyzes grain features such as color, size, and texture. It seeks to ensure consistent and transparent food quality monitoring across public kitchens and ration depots. By leveraging cloud computing and secure data pipelines, the objective is to enable real-time detection of adulteration and contamination while maintaining full traceability of inspection results.

Methods: The proposed method involves capturing high-resolution images of rice samples using IoT-enabled cameras placed at inspection points. The images undergo preprocessing, including background subtraction and conversion to binary form, similar to earlier morphological methods. Features such as major axis length, minor axis length, area, and texture are extracted to classify grains as Grade 1, Grade 2, or Grade 3. These features are analyzed using CNN-based models deployed on a cloud platform for automated grading. The processed results are stored and visualized through a cloud dashboard for inspectors to ensure transparency and consistency.

Results: For experimentation, 105 images of each rice variety—Basmati, Delhi, and Boiled—were tested, following the structure of the original work. Using the enhanced cloud-based AI framework, the classification accuracy improved over traditional decision-tree methods. The CNN achieved 96% accuracy in identifying grain quality and detecting contamination. The system successfully classified grains into Grade 1 (whole), Grade 2 (partially broken), and Grade 3 (broken/contaminated) categories. Latency averaged under 600 ms per inference, enabling near real-time inspection at ration depots. The results demonstrate that combining morphological feature extraction with modern deep learning provides both speed and reliability for large-scale deployment.

Conclusions: This study extends the original morphological approach for rice grading into a cloud-enabled AI system suitable for public kitchens and ration depots. The automated method replaces manual visual inspection with a scalable, secure, and data-driven solution. By integrating edge-based image capture, feature extraction, and cloud inference, the system achieves higher accuracy and transparency in quality assessment. The results are encouraging, showing that AI-driven image processing can significantly improve food safety monitoring. Furthermore, incorporating Zero Trust and data protection principles ensures secure and tamper-proof inspection records. The system represents a vital step toward modernizing public food safety and quality governance.