Assessment of kernel presence in winter wheat ears at spikelet scale using Near-Infrared Hyperspectral Imaging

Obtaining varieties with higher yield is one of the aims of crop breeding. For winter wheat (Triticum aestivum L.), yield is the result of three components which are the number of ears per unit area, the number of kernels per ear and the kernel weight. The occurrence of stresses during crop growth can have a negative impact on the yield components. Assessing the physiology behind the number of kernels and the impact of stresses is a complex process requiring observations at specific critical periods of the crop cycle.
The scope of this work is to propose a laboratory method based on Near Infrared Hyperspectral Imaging (NIR-HSI) to rapidly evaluate the presence or absence of kernels in winter wheat ears. This preliminary study has allowed evaluating the potential of the method using ears collected in a field trial in Belgium. The ears were scanned using a NIR-HSI camera in the spectral range 900-1700 nm. Hyperspectral images of the ears have been collected before and after manually removing the kernels from the ears. NIR-HSI combined with chemometric tools identified spectral bands potentially characteristic of the presence of kernels inside the ears. A chemometric model has been built using Partial Least Squares Discriminant Analysis (PLS-DA) to identify ears with and without kernels. The PLS-DA results obtained in this study allowed to track the presence or absence of kernels in winter wheat ears at spikelet scale. These results indicate that NIR-HSI might be a useful method to provide information assessing the status of kernels inside the ear which plays an important role in yield assessment.