Phenotypic Analysis and Variety Classification of Waxy Corn Seeds Based on Machine Vision

This study aims to apply machine vision for phenotypic analysis and varietal classification of waxy corn seeds, addressing the challenges of varietal mixture that hinder industry development. High-quality imaging was used to capture seed images from three waxy corn varieties, and phenotypic features such as size, color, and texture were extracted by OpenCV. A comprehensive assessment of 15 phenotypic features for the three varieties has identified discernible variations in size, color, and texture features. To classify waxy corn seeds effectively, we applied a diverse set of machine learning algorithms, including K-Nearest Neighbors (KNN), Linear Discriminant Analysis (LDA), Decision Trees (DT), and Support Vector Machines (SVM), each contributing unique strengths to the classification process. The five models achieved high accuracy rates, exceeding 90 % with the following values: KNN at 0.947, LDA at 0.986, DT at 0.922, SVM at 0.977, and ANN at 0.981. These findings highlight the potential of these algorithms, particularly LDA, for high-accuracy seed classification based on phenotypic features. This research has successfully developed classification models that enable the fast, impartial, and economical identification of waxy corn varieties. By expanding this methodology to include a wider range of varieties, future research could play a significant role in preventing seed and product adulteration in agriculture.