Table 3.
BiLSTM maize-soybean diseased leaf identification model based on different spectral preprocessing methods.
| Model | Method | Training set classification recall/% | General accuracy |
Test set classification recall/% | General accuracy |
||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Maize normalcy |
Maize spot |
Maize rust |
Maize hybrid |
Soybean normalcy |
Soybean rust |
Maize normalcy |
Maize spot |
Maize rust |
Maize hybrid |
Soybean normalcy |
Soybean rust |
||||
| BiLSTM | SG | 98.6 | 88.8 | 59.2 | 83.5 | 100.0 | 84.9 | 86.0 | 100.0 | 75.0 | 58.6 | 76.2 | 100.0 | 81.5 | 82.0 |
| SNV | 100.0 | 87.8 | 86.3 | 86.7 | 100.0 | 94.6 | 92.7 | 100.0 | 96.2 | 63.0 | 74.0 | 100.0 | 84.6 | 84.3 | |
| MSC | 100.0 | 86.8 | 77.0 | 87.2 | 100.0 | 91.8 | 90.4 | 100.0 | 95.8 | 61.5 | 71.8 | 100.0 | 88.9 | 85.0 | |
| 1st Der | 100.0 | 90.3 | 92.5 | 67.1 | 100.0 | 96.1 | 91.1 | 100.0 | 85.7 | 90.0 | 63.0 | 100.0 | 75.0 | 85.3 | |