Answers are grounded in this paper and cite their sources.
Tobacco mosaic virus and leaf curl virus pose serious threats to tobacco production, leading to substantial yield and quality losses. To improve disease detection accuracy, this study develops a robust uncrewed aerial vehicle-based hyperspectral remote sensing framework integrating multifeature fusion and systematic model evaluation. Spectral, wavelet, and texture features were extracted and refined through a hierarchical feature selection strategy combining analysis of variance, variable importance in projection, and principal component analysis (PCA). PCA was further employed for dimensionality reduction, resulting in 25 optimal features. Four models—support vector machine (SVM), random forest (RF), eXtreme gradient boosting (XGBoost), and multilayer perceptron —were constructed using single-variety (Yunyan 87, K326) as well as combined datasets. All models achieved accuracies of at least 0.91, demonstrating the effectiveness of the proposed multifeature fusion approach. For the Yunyan 87 dataset, XGBoost achieved the highest test accuracy (0.98), whereas SVM showed the highest cross-validation accuracy (0.98 ± 0.02), indicating more stable performance across different data partitions; for K326, both RF and XGBoost obtained the highest test accuracy (0.95), while XGBoost achieved the best cross-validation performance (0.98 ± 0.03); Notably, XGBoost exhibited relatively consistent and robust performance across all disease categories, particularly on the combined dataset. These results confirm that integrating complementary hyperspectral features significantly enhances tobacco disease detection and provides a reliable solution for precision disease monitoring.
No takes yet. Share an insight, caveat, or question.
Zhuang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: