EUVIMEDEUVIMED entdecken
Crossref

Simultaneously prediction of multiple wheat leaf phenotypes using hyperspectral imaging with multi-task machine and deep learning

Xiaoyu Fu, Lili Yao, Xincheng Zhang, Yuxin Xiao, Kai Gao, Wenjie Liu, Qinfeng Zhang, Hengnian Qi, Chu Zhang, Yiying Zhao

Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy · 2026-12 · 363 · 128294

DOI: 10.1016/j.saa.2026.128294 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2026-12. Verfasst von Xiaoyu Fu und 9 weiteren Autor:innen. Beteiligte Einrichtung: School of Information Engineering, Huzhou Normal University, Huzhou 313000, China. Hauptthemen laut MEDLINE: Triticum, Hyperspectral Imaging, Plant Leaves, Deep Learning. Die Arbeit nennt 46 Literaturangaben.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Chlorophyll content (represented by the Soil and Plant Analyzer Development (SPAD) value) and leaf moisture content (LMC) are two key physiological phenotypic traits during wheat growth, and their variations among different wheat varieties reflect crop growth, stress response, and breeding evaluation. Simultaneous identification of wheat varieties and prediction of SPAD and LMC are therefore important for precision crop monitoring. In this study, hyperspectral imaging was employed to acquire leaf spectral information from four wheat varieties. After spectral preprocessing and outlier screening, 684 valid samples were retained for model development and evaluation. Single-task and multi-task models were constructed for wheat variety classification, SPAD prediction, and LMC prediction using support vector machine (SVM), partial least squares (PLS), convolutional neural network (CNN), and multi-task CNN (MLT-CNN) algorithms. In the MLT-CNN, a shared one-dimensional spectral feature extraction module and three task-specific branches were designed, and equal weight strategy (EWS), adjustable regularization weighted strategy (ARWS), and uncertainty-based weighted strategy (UWS) were compared. The best single-task models achieved a test-set classification accuracy of 0.75, with correlation coefficients (r) of 0.83 for SPAD and 0.84 for LMC. The MLT-CNN with EWS achieved balanced test-set performance across the three tasks, with a classification accuracy of 0.69, an r value of 0.84 for SPAD, and an r value of 0.81 for LMC. Shapley additive explanations (SHAP)-based visualization was further performed for both single-task CNNs and MLT-CNN task branches to identify important wavelengths and interpret shared and task-specific spectral contributions. These results indicate that hyperspectral imaging combined with multi-task learning provides a feasible and interpretable spectroscopic strategy for integrated wheat leaf phenotyping.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

TriticumHyperspectral ImagingPlant LeavesDeep LearningPhenotypeConvolutional Neural NetworksSupport Vector MachineLeast-Squares AnalysisChlorophyllClassification AlgorithmsPrediction Algorithms

Stichwörter der Autor:innen

Hyperspectral imagingLeaf moisture contentMulti-task learningSPADWheat varieties

Substanzen

Chlorophyll

Autor:innen und Institutionen

  • School of Information Engineering, Huzhou Normal University, Huzhou 313000, China.
  • Institute of Crop Science, Huzhou Academy of Agricultural Sciences, Huzhou 313000, China.
  • School of Information Engineering, Huzhou Normal University, Huzhou 313000, China. Electronic address: chuzh@zjhu.edu.cn.
  • Institute of Digital Agriculture, Zhejiang Academy of Agricultural Sciences, Hangzhou 310021, China. Electronic address: zhaoyy@zaas.ac.cn.

Zitieren

APA

Fu, X., Yao, L., Zhang, X., Xiao, Y., Gao, K., Liu, W., Zhang, Q., Qi, H., Zhang, C., & Zhao, Y. (2026). Simultaneously prediction of multiple wheat leaf phenotypes using hyperspectral imaging with multi-task machine and deep learning. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 363, 128294. https://doi.org/10.1016/j.saa.2026.128294

Vancouver

Fu X, Yao L, Zhang X, Xiao Y, Gao K, Liu W, et al. Simultaneously prediction of multiple wheat leaf phenotypes using hyperspectral imaging with multi-task machine and deep learning. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy. 2026;363:128294. doi:10.1016/j.saa.2026.128294

BibTeX
@article{Fu2026,
  author = {Fu, Xiaoyu and Yao, Lili and Zhang, Xincheng and Xiao, Yuxin and Gao, Kai and Liu, Wenjie and Zhang, Qinfeng and Qi, Hengnian and Zhang, Chu and Zhao, Yiying},
  title = {Simultaneously prediction of multiple wheat leaf phenotypes using hyperspectral imaging with multi-task machine and deep learning},
  journal = {Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy},
  year = {2026},
  volume = {363},
  pages = {128294},
  doi = {10.1016/j.saa.2026.128294}
}
Kostenlos starten →