| Abstrakt: |
Plant diseases can lead to significant yield losses and economic damages, but these losses can be mitigated through early disease diagnosis. In recent times, remote sensing techniques have been widely used for early disease detection even before visible symptoms appear. This study focused on the potential of early detection of Fusarium Root and Crown Rot in Tomato Plants, which causes substantial yield losses in tomato plants, under controlled conditions using thermal images. In this research, thermal images were obtained from both disease-inoculated and disease-free control plants throughout the plant growth period under controlled conditions. These images underwent preprocessing in a computer environment, and various feature parameters related to temperature changes in both groups (such as minimum, maximum, standard deviation, and skewness) were extracted. These extracted features were then used as inputs for different machine learning techniques, including K-Nearest Neighbors (KNN), Logistic Regression (LR), and Naive Bayes (NB), to classify healthy and diseased plants. Overall, the disease-inoculated plants exhibited higher average temperatures compared to the healthy control plants. The performance of the compared machine learning techniques in distinguishing between healthy and diseased plants was found to be in the order of KNN, NB, and LR, with success rates of 72%, 68%, and 60%, respectively. This study demonstrated the potential of using combined thermal images with different machine learning techniques for early diagnosis of Fusarium Root and Crown Rot in Tomato Plants. The results show promising prospects for utilizing thermal imaging in the early detection of plant diseases, leading to better management and reduction of yield losses and economic impacts. [ABSTRACT FROM AUTHOR] |
