An image-based automated pipeline for maize ear and silk detection in a highthroughput phenotyping platform
Résumé
Water deficit strongly impacts silk growth and silk emergence in
maize (Zea mays L.), which in turn determines the final number
of ovaries developing grains (Turc et al. 2016, Oury et al. 2016).
However, phenotyping silk growth and silk expansion is difficult
at throughput needed for genetic analyses. We have developed
an image-based automated pipeline for maize ear and silk detection
in a high-throughput phenotyping platform. The first step
consists of selecting the best whole plant side images containing
maximum information for each plant and day as that containing
the most leaves and whole stem, based on top view images. In
the second step, the best side images are segmented and skeletonized,
and potential ear positions are determined based on
changes in stem widths. The x, y, z ear position identified in this
way serves to pilot the movement of a mobile camera able to
take a detailed picture taken at 30 cm from the ear, with the final
aim of determining silk emergence and silk growth duration.
These methods were tested at the PhenoArch plant phenotyping
platform (www6.montpellier.inra.fr/lepse/M3P) in a panel of
300 maize hybrids. First results showed that in >80% of cases,
ears were successfully detected before silking and duration
of silk expansion significantly correlated with visual scores. The
image pipeline presented here opens up the way for large-scale
genetic analyses of control of reproductive growth to changes in
environmental conditions in reproductive structures.
Domaines
Biologie végétale
Origine : Fichiers produits par l'(les) auteur(s)