TransLIBS-CRS: Integrating LIBS with transfer learning for accurate cross-regional soil total nitrogen detection
Accurate detection of soil total nitrogen (TN) is crucial for enhancing crop growth and quality. However, soil variability across different regions limits the generalization ability of calibration models. To address this challenge, this study introduces a model named transfer-learning-assisted laser...
Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-12-01
|
| Series: | Smart Agricultural Technology |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S277237552500351X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate detection of soil total nitrogen (TN) is crucial for enhancing crop growth and quality. However, soil variability across different regions limits the generalization ability of calibration models. To address this challenge, this study introduces a model named transfer-learning-assisted laser-induced breakdown spectroscopy for cross-regional soil analysis (TransLIBS-CRS), specifically designed to mitigate the low cross-domain prediction accuracy caused by variations in regional soil properties. By fine-tuning with a limited number of target domain samples, this method significantly improves the cross-domain applicability of LIBS data, alleviating the challenges associated with the difficulty of obtaining soil samples from diverse regions. In the task of predicting TN in Guangzhou using the Beijing dataset, the TransLIBS-CRS model achieved optimal performance, with RV2 of 0.846 and RMSEV- of 0.814 g/kg. Further analysis through saliency map and chemometric methods revealed that spectral lines of carbon at 193.0 nm and 247.8 nm play a key role in the quantitative detection of TN. Notably, these spectral features also demonstrated stable predictive contributions when transferred to the Guangzhou soil dataset. This approach offers a feasible solution for large-scale and efficient soil TN detection. |
|---|---|
| ISSN: | 2772-3755 |