[Research on fast detecting tomato seedlings nitrogen content based on NIR characteristic spectrum selection]

Area of Science:

• Agricultural Science
• Spectroscopy
• Plant Physiology

Background:

• Accurate monitoring of tomato seedling nitrogen content is crucial for optimizing crop yield and quality.
• Near-infrared spectroscopy (NIR) offers a non-destructive method for analyzing plant nutritional status.
• Selecting relevant spectral features is essential for building robust NIR predictive models.

Purpose of the Study:

• To evaluate four characteristic spectrum selection methods for improving NIR-based detection of tomato seedling nitrogen content.
• To compare the performance of competitive adaptive reweighted sampling (CARS), Monte Carlo uninformative variables elimination (MCUVE), backward interval partial least squares (BiPLS), and synergy interval partial least squares (SiPLS).
• To determine the optimal method for robust and accurate nitrogen content prediction across varying nitrogen levels.

Main Methods:

• Sixty tomato seedlings were subjected to ten different nitrogen treatment levels (0-120 mg/L urea).
• Near-infrared spectra (12,500–3,600 cm⁻¹) were collected from leaf samples.
• Quantitative models were developed using CARS, MCUVE, BiPLS, and SiPLS for spectral data analysis.

Main Results:

• CARS and MCUVE showed better calibration model performance but lower prediction ability compared to BiPLS and SiPLS.
• The BiPLS method yielded the best prediction performance, with a correlation coefficient (r) of 0.9527, RMSEP of 0.1183, and RPD of 3.291.
• Single-wavelength selection models are sensitive to uniform objects, while interval selection models offer stronger anti-interference for uneven samples.

Conclusions:

• NIR technology combined with characteristic spectrum selection methods significantly enhances model performance for detecting tomato nitrogen content.
• The choice of spectral selection method (single wavelength vs. interval) should consider sample characteristics and model performance indicators.
• BiPLS is recommended for its superior prediction accuracy in diverse nitrogen status conditions.