Classification Method for Viability Screening of Naturally Aged Watermelon Seeds Using FT-NIR Spectroscopy

Area of Science:

• Agricultural Science
• Spectroscopy
• Plant Biology

Background:

• Seed viability decreases over time during storage, impacting crop yield.
• Accurate viability assessment is crucial for effective seed sowing and agricultural planning.
• Triploid watermelon seeds present challenges for analysis due to their thick seed coats.

Purpose of the Study:

• To investigate the efficacy of Fourier transform near infrared spectroscopy (FT-NIR) for assessing the viability of aged triploid watermelon seeds.
• To develop and validate a predictive model for discriminating between viable and non-viable seeds across multiple varieties.
• To confirm the penetration depth of FT-NIR light for effective spectral data collection from seed embryos.

Main Methods:

• Collected FT-NIR spectra from triploid watermelon seeds stored for four years under controlled conditions.
• Confirmed seed viability using standard germination tests.
• Developed a partial least discriminant analysis (PLS-DA) model to classify spectra.
• Validated the PLS-DA model using an independent dataset of 100 randomly selected seeds.

Main Results:

• The PLS-DA model achieved high classification accuracy for mixed varieties of watermelon seeds.
• External validation demonstrated good accuracy for viable seeds (87.7%) and non-viable seeds (82%).
• The developed model proved effective across different seed varieties and data collection times.

Conclusions:

• FT-NIR spectroscopy, coupled with PLS-DA, is a viable tool for non-destructive seed viability assessment.
• The developed "general model" shows applicability across multiple triploid watermelon varieties and varying storage durations.
• This technique offers a promising alternative to traditional germination tests for rapid seed quality evaluation.