Classification of Hydroponic Lettuce Leaves using Machine Learning

Hydroponic lettuce, known for its rich nutritional content and medicinal properties, holds immense potential as a staple vegetable. However, the emergence of yellow and rotten leaves due to nutrient deficiencies poses significant challenges in maintaining quality and efficiency during production. Manual sorting of these abnormal leaves before packing incurs high labor costs and inefficiencies, necessitating the development of automated sorting systems. Further, it is easy to classify the good leaves among the available stock. In this work, the four classes being considered are Fully Nutritious, Nitrogen, Potassium and Phosphorous Deficient leaves. Texture features help in identifying the physical condition and health of the leaves, capturing details related to the leaf surface which can indicate nutrient deficiencies. Color features, on the other hand, are crucial for detecting changes in leaf pigmentation, which often signal specific nutrient deficiencies. KNN, SVM, and CNN are used in this work. KNN relies on nearby data points for classification, while SVM constructs optimal decision boundaries in high-dimensional spaces. CNN leverages convolutional layers for deep feature extraction from image data. The performance of all the methods is assessed using confusion matrix. Experimental results specify that the proposed method gives better accuracy for four classes.