Laboratory and UAV-Based Identification and Classification of Tomato Yellow Leaf Curl, Bacterial Spot, and Target Spot Diseases in Tomato Utilizing Hyperspectral Imaging and Machine Learning

Tomato crops are susceptible to multiple diseases, several of which may be present during the same season. Therefore, rapid disease identification could enhance crop management consequently increasing the yield. In this study, nondestructive methods were developed to detect diseases that affect tomato crops, such as bacterial spot (BS), target spot (TS), and tomato yellow leaf curl (TYLC) for two varieties of tomato (susceptible and tolerant to TYLC only) by using hyperspectral sensing in two conditions: a) laboratory (benchtop scanning), and b) in field using an unmanned aerial vehicle (UAV-based). The stepwise discriminant analysis (STDA) and the radial basis function were applied to classify the infected plants and distinguish them from noninfected or healthy (H) plants. Multiple vegetation indices (VIs) and the M statistic method were utilized to distinguish and classify the diseased plants. In general, the classification results between healthy and diseased plants were highly accurate for all diseases; for instance, when comparing H vs. BS, TS, and TYLC in the asymptomatic stage and laboratory conditions, the classification rates were 94%, 95%, and 100%, respectively. Similarly, in the symptomatic stage, the classification rates between healthy and infected plants were 98% for BS, and 99–100% for TS and TYLC diseases. The classification results in the field conditions also showed high values of 98%, 96%, and 100%, for BS, TS, and TYLC, respectively. The VIs that could best identify these diseases were the renormalized difference vegetation index (RDVI), and the modified triangular vegetation index 1 (MTVI 1) in both laboratory and field. The results were promising and suggest the possibility to identify these diseases using remote sensing.