Performance of Support Vector Machines, Artificial Neural Network, and Random Forest for Identifying Banana Fusarium Wilt Using UAV-Based Multi-spectral Imagery

Abstract

The Fusarium wilt of bananas currently threatens to the banana production areas worldwide. Timely monitoring of Fusarium wilt disease is very important for the disease treatment and crop planting adjustment. The aim of this paper was to evaluate the performance of support vector machines (SVM), random forest (RF), and artificial neural network (ANN) with unmanned aerial vehicle (UAV)-based multi-spectral imagery to identify the locations that were infested or not infested with banana Fusarium wilt. A total of 139 ground samples were surveyed to assess the occurrence of banana Fusarium wilt. The results showed that the overall accuracies of SVM, RF, and ANN were higher than 90% for the pixel based. Among the classifiers, SVM had the best performance, followed by ANN and RF. The maps generated by SVM, RF, and ANN appeared a similar distribution trend with regard to the occurrence of Fusarium wilt. The areas of the occurrence of Fusarium wilt were between 5.21 and 5.75 ha, accounting for 36.3–40.1% of the total planting area of bananas in the study area. The results also showed that the inclusion of the red-edge band had 2.9–3.0% increases in overall accuracy. The results of this study indicate that the SVM, RF, and ANN with UAV-based remote sensing imagery have the potential to identify and map the banana Fusarium wilt.