Hyperspectral sensors and SAINT model for predicting soil boron content in smart agriculture
Year: 2026
Type: article
Abstract: With the development of smart agriculture technologies, real-time monitoring of farmland soil nutrients based on hyperspectral sensors, wireless communication, and intelligent decision-making is becoming an important part of precision agriculture. Hyperspectral sensors deployed in farmland, on drones, or on mobile platforms can rapidly acquire soil reflectance spectra, while the agricultural Internet of Things enables remote data transmission and analysis for precision fertilization and soil health management. This paper presents a novel application of the Self-Attention and Intersample Attention Transformer (SAINT) model for predicting soil boron content from hyperspectral data. Traditional soil nutrient analysis methods, although reliable, are often labor-intensive and time-consuming. The SAINT model, known for processing complex tabular data, is adapted here to analyze high-dimensional spectral data, which is challenging for conventional machine learning methods. By integrating self-attention and intersample attention, the model captures dependencies within and between soil samples, improving prediction accuracy. Results show that the SAINT model outperforms Random Forest and XGBoost in terms of Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R 2 ), while also identifying important spectral bands for boron detection. This study provides a tool for rapid soil analysis that can improve nutrient management and support sustainable agricultural productivity. Future work includes extending the model to other micronutrients and integrating real-time hyperspectral systems for on-site soil health assessment. Combining hyperspectral sensor data with attention-based deep learning models can provide an effective approach for rapid and accurate soil nutrient analysis in smart agriculture, supporting precision fertilization, resource conservation, and sustainable agricultural development.
Source: Discover Applied Sciences
Institution: Anqing Normal University
Language: English
Subfield: Analytical Chemistry
Field: Chemistry
Domain: Physical Sciences
SDG: Zero hunger
Open Access status: diamond
OpenAlex ID: W7204541982
FWCI:
Cites:
Cited by:
Related to: