Table 4.
The accuracy of estimation of different models for grain weight per plant.
| Type | No. of indices | Metrics | PLSR | RFR | Logistic | SVM | DNN |
|---|---|---|---|---|---|---|---|
| VIs | 25 | R2 | 0.43 | 0.42 | 0.45 | 0.48 | 0.48 |
| RMSE | 3.98 | 3.97 | 3.98 | 3.83 | 3.88 | ||
| rRMSE | 42.96% | 42.53% | 41.42% | 41.26% | 43.52% | ||
| VIs+TI | 60 | R2 | 0.45 | 0.45 | 0.51 | 0.58 | 0.58 |
| RMSE | 3.89 | 3.86 | 3.67 | 3.4 | 3.541 | ||
| rRMSE | 41.72% | 41.31% | 39.33% | 36.34% | 39.45% | ||
| VIs+TI+CC | 61 | R2 | 0.45 | 0.46 | 0.51 | 0.59 | 0.61 |
| RMSE | 3.9 | 3.85 | 3.67 | 3.40 | 3.28 | ||
| rRMSE | 41.75% | 41.26% | 39.32% | 36.47% | 35.29% | ||
| VIs+TI+CC+CHM | 62 | R2 | 0.48 | 0.49 | 0.52 | 0.61 | 0.62 |
| RMSE | 3.78 | 3.74 | 3.60 | 3.31 | 3.25 | ||
| rRMSE | 40.4% | 40.00% | 38.52% | 35.40% | 35.32% | ||
| VIs+TI+CC+CHM+Lodging | 63 | R2 | 0.48 | 0.49 | 0.53 | 0.62 | 0.64 |
| RMSE | 3.78 | 3.76 | 3.6 | 3.287 | 3.16 | ||
| rRMSE | 40.43% | 40.22% | 38.54% | 35.18% | 33.80% |
VIs, vegetation index; TI, texture information; CC, canopy cover; CHM, crop height; DNN, deep neural network; Logistic, logistic regression; PLSR, partial least squares regression; RFR, random forest regression; SVM, support vector machine regression. The best results in terms of R2, RMSE, and rRMSE values through different sensors with various modeling methods are shown in bold.