Hyperspectral imaging and machine learning for herbicide-resistant kochia identification in sugarbeet
Weed infestation in sugarbeet fields can substantially affect crop yield and quality, posing a major challenge to the global sugar industry. Herbicides are commonly used to control invasive weeds such as kochia; however, kochia has developed resistance to multiple herbicide sites of action, creating...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-10-01
|
| Series: | Journal of Agriculture and Food Research |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2666154325006556 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Weed infestation in sugarbeet fields can substantially affect crop yield and quality, posing a major challenge to the global sugar industry. Herbicides are commonly used to control invasive weeds such as kochia; however, kochia has developed resistance to multiple herbicide sites of action, creating additional management challenges. Effective detection and identification of herbicide-resistant kochia are therefore critical for site-specific management. In this study, hyperspectral imaging in the visible and near-infrared range was used to acquire spectral data from dicamba-resistant, glyphosate-resistant, glyphosate-susceptible kochia, and sugarbeet. Spectra data were acquired and preprocessed using standard normal variate transformation and Savitzky–Golay smoothing methods, followed by stepwise feature selection to identify wavelengths features for classification. The selected wavelengths were used to train supervised machine learning models, namely support vector machine (SVM), random forest (RF), linear discriminant analysis (LDA), and extreme gradient boost (XGBoost). SVM achieved the highest overall accuracy of 84.7 %, followed by RF, LDA, and XGBoost with accuracy of 81.1 %, 73.6 %, and 62.3 %, respectively. The results highlight the potential of hyperspectral imaging combined with machine learning for precise identification of herbicide-resistant weed for site-specific weed control and help reduce the evolution of resistance species in sugarbeet production. |
|---|---|
| ISSN: | 2666-1543 |