Methods for sorting haploid maize kernels in haploid induction crosses are provided. A method of sorting haploid kernels can include capturing near-infrared (NIR) spectra and applying a general multivariate statistical model to the acquired NIR spectra to discriminate haploid kernels from diploid-hybrid kernels mixed therein. NIR spectra can be collected on a single kernel using a high-throughput apparatus such that the amount of time required to analyze individual kernels can be significantly reduced in comparison to existing NIR technology and in comparison to traditional manual sorting.