On-demand Healthy and Chlorotic Lactuca sativa Leaf Classification Using Support Vector Machine in a Rotating Hydroponic System

Chlorosis is a manifestation of discoloration on leaves of plants that indicates an unideal cultivation environment. It can appear as yellowish, yellow-white, or white pigment on the leaf surface that might spread until the affected leaf dies. An essential first step in halting its spread is early detection of its presence. On the other hand, ineffective and erratic monitoring could compromise manual leaf examination. This research proposes a method for evaluating healthy and chlorotic lettuce leaves produced in rotating hydroponics setups using computational intelligence and computer vision. A total of 533 image sets (273 healthy and 260 chlorotic leaves) were utilized. For the healthy and chlorotic regions, phenotypic segmentation was used with lazy snapping and CIELab color space. Canopies of lettuce leaves were extracted for spectral (L, a*, b*) features. These features were then utilized for training the machine learning models. Regarding the health classification of of lettuce leaves, the support vector machine (SVM) performed best with 100% accuracy and took the least amount of time to process. K-nearest neighbors (KNN) also has the best performance of 100% accuracy; however, the inference time is second to the most extended (48.4 s). Naïve Bayes (NB) has a considerably high accuracy of 92.93% but has the longest inference time. Ultimately, a low-cost consumer-grade camera is used to classify leaves using the developed SVM model accurately.