A novel approach using multispectral imaging for rapid development of seed pellet formulations to mitigate drought stress in alfalfa

Abstract

Seed pelleting is an advanced technology that can improve seed sowing and germination under abiotic stresses such as drought, promoting sustainable agriculture. However, the development of conventional seed pelleting formulations is often characterized by prolonged timelines and significant labor costs. This study aimed to accelerate the development of water retention agents (WRAs) pelleting formulations for alfalfa seed under drought stress using multispectral imaging technology. The efficacy of these formulations was evaluated under drought stress conditions by measuring seedling physiological characteristics and collecting multispectral images. Results showed that WRA-pelleting formulations with HS0.5, BS1, and BS2 outperformed other WRA-pelleting formulations in terms of catalase, seedling length, superoxide dismutase, proline, and shoot length by principal component analysis. Additionally, backpropagation neural network, support vector machine, and random forest models were employed to identify various WRA-pelleting formulations using multispectral data from seedlings and cotyledons. All three models had an accuracy above 0.9 based on multispectral data from cotyledons, with support vector machine achieving the highest accuracy of 100%. The SHapley Additive exPlanations method was applied to explain the prediction mechanism of the support vector machine model and identified important features such as 970, 850, 880, 940, and 780 nm. Significant correlations were found between spectra and physiological indicators such as antioxidant, chlorophyll, and drought stress. In conclusion, our new approach has great potential to reduce the time and cost of developing seed pelleting formulations for mitigating drought stress.