Classification of Lactuca Sativa Seed Varieties Using Spectral- Textural-Spatial Features and Decision Tree Classifier

Abstract

It is important to identify the variety of a crop before planting the seed. Identifying lettuce seed varieties manually can be a tough process. This study focuses on creating a system that can classify lettuce seed images using machine learning. Three Lettuce varieties were considered in this study, including the Romaine, Black Seeded Simpson, and Lollo Rosso. 100 seed images were captured for each lettuce seed variety in a controlled environment. The seed images were segmented using color thresholding in the HSV color space. Color, texture, and morphological features were used for variety classification including RGB color features, Solidity, Correlation, Compactness, Perimeter, Entropy. Decision Tree Classification algorithm was used to classify lettuce seeds. The system was able to successfully classify lettuce seed varieties with an average accuracy of 92%.