Real-Time Multi-Class Paddy Grain Classification and Grading Using AI on Edge Computing

Abstract

Grading and classifying paddy grains manually is a slow process and can lead to human mistakes. To solve this problem, this study proposes an automated system to classify and grade paddy grains using Machine Learning (ML) and Deep Learning (DL) method. In this work, a dataset of paddy grain images was collected from BERNAS. Image processing techniques were used refine the images and extract important features such as size, shape, and color. These features were then used to train K-Nearest Neighbors (KNN) and Convolutional Neural Networks (CNN) models to determine the quality and grade of the paddy. The results show that the proposed system is effective, achieving a maximum classification the accuracy. By using this model, the percentage of KNN is 76.4% while CNN 96.5%. In conclusion, CNN based system provides a faster and more reliable method for paddy grain grading, which can help the rice industry improve the quality control.