Leveraging Diverse Handcrafted Features for Improved Rice Seed Recognition: A Step Toward Precision Agriculture

This paper introduces a novel approach to enhance rice seed purity recognition performance, contributing significantly to precision agriculture. By synergizing multiple feature extraction methods (GIST, Local Binary Pattern (LBP), Grey Level Co-occurrence Matrix (GLCM), and Basic features), we create a new comprehensive feature representation that improves rice seed variety classification. Experiments on a real-world dataset using various machine learning (ML) algorithms demonstrate the superiority of our method. Notably, the CatBoost classifier achieves remarkable classification accuracy when utilizing the combined features, highlighting the effectiveness of our proposed approach for precision agriculture applications. These findings have practical implications, enabling more precise and efficient seed selection within smart farming practices.