Fig. 1. Sample data collection
Fig. 2. Flowchart of tomato gray mold diagnosis and classification
Fig. 3. Hyperspectral images of inoculated pathogenic sample #1 in the first 8 days
Fig. 4. Hyperspectral mean reflectivity curves of tomato leaves in different periods
Fig. 5. Comparison of spectra with different pretreatment methods
Fig. 6. Feature wavebands extracted by DWT-CARS algorithm for three times
(a): The first extraction; (b): The second extraction; (c): The third extraction
Fig. 7. Feature bands extracted by DWT-CARS algorithm for three times
(a): The first extraction; (b): The second extraction; (c): The third extraction
Fig. 8. Confusion matrix of three model test sets
(a): DWT-FC-TLBO-ELM; (b): DWT-TLBO-ELM; (c): DWT-CARS-TLBO-ELM
DWT-CARS方法 | 第1次 | 第2次 | 第3次 |
---|
特征波段个数/个 | 5 | 5 | 7 | 特征波段/nm | 696, 765, 769, 838, 840 | 696, 767, 769, 772, 840 | 694, 696, 767, 769, 778, 838, 840 | RMSECV/% | 0.295 9 | 0.294 5 | 0.297 4 | 最终特征波段/nm | 694, 696, 765, 767, 769, 772, 778, 838, 840 |
|
Table 1. Feature bands extracted by DWT-CARS algorithm for three times
诊断模型 | 模型参数 | 测试集 准确率/% | 测试集 精确率/% | 召回率 /% | F1 /% |
---|
最优权值ω | 隐藏层最优阈值b |
---|
DWT-FC-TLBO-ELM | 1.044 8 | 155 | 95.71 | 96.47 | 95.71 | 96.09 | DWT-TLBO-ELM | 1.044 8 | 70 | 95.71 | 96.17 | 95.71 | 95.94 | DWT-CARS-TLBO-ELM | 1 | 107 | 100 | 100 | 100 | 100 |
|
Table 2. Model parameters and test set accuracy, accuracy, recall rate and F1 value