Advanced Automated Seed Sorting System Powered by IoT and AI

Abstract

Objectives: Seed quality is a crucial determinant of agricultural productivity and crop sustainability. This work aims to address the inefficiencies of manual seed sorting by developing an automated system capable of accurately classifying seeds based on their visual characteristics.
Methods: The proposed system utilizes Convolutional Neural Networks (CNNs) to classify seeds, specifically white beans and peanuts, as either good or bad. Images are captured using a Raspberry Pi 4 and Pi camera setup, enabling real-time classification. A servomotor, controlled by the Raspberry Pi, segregates the seeds into appropriate compartments. Furthermore, a web application developed using the ASP.NET Framework allows for remote system control and monitoring.
Findings: The integration of CNNs with automation significantly improved the accuracy and efficiency of the seed sorting process. The system achieved high classification precision under controlled conditions, demonstrating its potential to reduce labor requirements and human error. However, environmental factors and budget constraints posed challenges to performance consistency.
Novelty: This system represents a novel application of deep learning in agricultural automation, combining real-time image classification with mechanical sorting. The inclusion of a web-based interface enhances its usability and scalability for diverse agricultural settings.