An AIoT-Based Approach to High-Throughput Grain Quality Control: Application to the Food Industry

This paper contributes to the Internet-of-Things (IoT) and Sensor Network Symposium track by presenting a novel IoT-based solution for quality control in the grain industry. We present an innovative mechanical setup integrated with a sensor network for high-throughput seed scanning. This setup collaborates with an edge computing model, showcasing a real-world IoT application in industry. Using a dataset of 36,479 seed images spanning 8 different classes-including healthy grains and various defects-we train and validate deep learning models. The edge computing model ensures real-time, low-latency image analysis, essential for swift and precise grain quality classification. We stream the analyzed data to both the cloud and a local manufacturing execution system (MES) using IoT protocols, enabling on-the-fly optimization of processing devices. This approach transforms quality control in the grain industry, highlighting potential gains in efficiency, waste reduction, and quality assurance.