EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Wheat growth parameters prediction based on dual output Bayesian neural network using multi-modal information

Ke Xu, Dahui Cao, Qinyue Tai, Haiting Wang, Jun Ni, Yaocong Hu, Huicheng Yang

Frontiers in Plant Science · 26.06.2026 · 17

DOI: 10.3389/fpls.2026.1871483 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Frontiers in Plant Science, online erschienen am 26.06.2026. Verfasst von Ke Xu und 6 weiteren Autor:innen. Beteiligte Einrichtung: School of Integrated Circuits, Anhui Polytechnic University, Wuhu, China. Die Arbeit nennt 32 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Youth Science Fund Project.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction eaf area index (LAI) and leaf nitrogen accumulation (LNA) are key indicators of wheat growth and nitrogen nutritional status. However, existing prediction methods predominantly rely on single-modal information and single-output models, limiting their ability to characterize the complex structural and physiological traits of crops. This study aimed to develop a multimodal learning framework for the simultaneous and accurate prediction of wheat LAI and LNA.

Methods

Spectral, image, and canopy structural features were extracted from wheat canopies across different cultivars, nitrogen treatments, and growth stages. A canopy height correction-based preprocessing method was developed to improve the extraction of structural features. A Dual-Output Bayesian Neural Network (DO-BNN) was then constructed to simultaneously predict LAI and LNA. In addition, an Extreme Sample Mining (ESM) strategy and a joint loss function were introduced to strengthen the learning of complementary information across modalities and the intrinsic correlation between the two target variables.

Results

The DO-BNN achieved its best predictive performance when all feature modalities were fused. The coefficients of determination (R²) for LAI and LNA were 0.89 and 0.77, respectively, while the corresponding relative root mean square errors (RRMSEs) were 0.15 and 0.35. Compared with single-modal and conventional single-output approaches, the proposed method provided more accurate and robust predictions of both wheat growth parameters.

Discussion

The results demonstrate that integrating spectral, image, and structural information can improve the characterization of wheat canopy traits. By jointly modeling LAI and LNA, the DO-BNN effectively exploited the physiological relationship between crop growth and nitrogen accumulation. The proposed framework provides a promising approach for the high-accuracy, collaborative monitoring of wheat growth and nitrogen nutritional status.

Schlagwörter

Stichwörter der Autor:innen

Bayesian neural networkdual output predictionleaf area index (LAI)leaf nitrogen accumulation (LNA)multimodal informationwheat

Autor:innen und Institutionen

  • School of Integrated Circuits, Anhui Polytechnic University, Wuhu, China.
  • Anhui Provincial Key Laboratory of Advanced Detection and Intelligent Sensing, Wuhu, China.
  • School of Mechanical and Automotive Engineering, Shanghai University of Engineering Science, Shanghai, China.
  • School of Electrical Engineering, Anhui Polytechnic University, Wuhu, China.

Förderung

  • Youth Science Fund Project

Zitieren

APA

Xu, K., Cao, D., Tai, Q., Wang, H., Ni, J., Hu, Y., & Yang, H. (2026). Wheat growth parameters prediction based on dual output Bayesian neural network using multi-modal information. Frontiers in Plant Science, 17. https://doi.org/10.3389/fpls.2026.1871483

Vancouver

Xu K, Cao D, Tai Q, Wang H, Ni J, Hu Y, et al. Wheat growth parameters prediction based on dual output Bayesian neural network using multi-modal information. Frontiers in Plant Science. 2026;17. doi:10.3389/fpls.2026.1871483

BibTeX
@article{Xu2026,
  author = {Xu, Ke and Cao, Dahui and Tai, Qinyue and Wang, Haiting and Ni, Jun and Hu, Yaocong and Yang, Huicheng},
  title = {Wheat growth parameters prediction based on dual output Bayesian neural network using multi-modal information},
  journal = {Frontiers in Plant Science},
  year = {2026},
  volume = {17},
  doi = {10.3389/fpls.2026.1871483}
}
Kostenlos starten →