| Evaluation of drought state of potato leaves based on Hyperspectral imaging |
|
View Fulltext View/Add Comment Download reader |
| 投稿时间：2024-03-26 修订日期：2024-05-07 |
| DOI： |
| Key Words: Potato Hyperspectral imaging Characteristic wavelengths Model construction Drought stress identification |
|
| Hits: 2 |
| Download times: 0 |
| Abstract: |
| Drought stress is one of the most common potato stresses. Rapid detection of potato drought state is very important for efficient management of potato production and drought-resistant breeding. To evaluate the drought state of potato leaves accurately and quickly, a hyperspectral imaging method was employed to assess the drought state of potato leaves. In 2022, the original seed of QingshuNo.9 was used as the material. Three types of drought states were obtained using hyperspectral imaging. A total of 420 spectral reflectance measurements from various samples were extracted and the influence of four spectral data preprocessing methods on modeling is discussed. A classification model for the drought status of potato leaves was constructed using the random forest cross-validation recursive feature elimination algorithm (RF-RFECV) and the feature wavelength selection method of competitive adaptive reweighted sampling (CARS) methods combined with Extremely Randomized Trees (EXT). The results showed that the three drought state classification models constructed in this study had accuracy higher than 85% in the test set, among which the RF-RFECV-EXT model performed the best with a prediction accuracy of 92.14% in the test set. Furthermore, in order to display the drought state of potato leaves objectively, the RF-RFECV-EXT model was used to visually represent the degree of drought in the leaves using different colors. This study offers novel insights into identifying the drought state of potato leaves, and provides a theoretical basis for efficient water management in potato production. |