Sweet Corn Seed Germination and Vigor Prediction Using Single‐kernel Near‐infrared Spectroscopy

Sweet corn sh2 hybrids typically exhibit reduced seed germination and seedling vigor compared with field corn hybrids; therefore, accurate assessments of seed quality are important for predicting stand establishment and yield. Traditional germination and vigor tests are accurate but destructive, labor-intensive, and time-consuming, thus limiting their use for rapid decision-making. This study evaluated single-kernel near-infrared (skNIR) spectroscopy for predicting seed lot quality using samples from 181 sh2 sweet corn seed lots representing 13 commercial cultivars from five seed companies. Germination and vigor were assessed using seven laboratory tests and field trials across three environments. Partial least squares (PLS) and random forest models were developed from skNIR mean and standard deviation spectra. Both algorithms showed comparable predictive ability (R 2 = 0.43–0.44 averaged across all traits). Including seed company as a covariate improved PLS performance across all traits to R 2 = 0.54. Classification of seed lots into high-quality and low-quality groups achieved balanced accuracy of 64.0% to 79.8% and an area under the receiver-operating characteristic curve of 0.80–0.90. Consistent predictions were achievable with as few as 250 seeds per lot. To support practical adoption, an R Shiny application was developed providing a no-code interface for spectral preprocessing, model calibration, and trait prediction. These results demonstrate that skNIR offers a rapid, nondestructive, cost-effective complement to conventional seed quality testing.