High Resolution Phenotyping: Semi-Automated Feature Extraction from RGB Images of Sorghum Panicles

Our long-term goal is to develop statistical models that will predict crop performance in diverse agronomic environments. Crop phenotypes such as yield and drought tolerance are controlled by genotype, environment (considered broadly) and their interaction (GxE). Our understanding of the molecular basis of GxE remains rudimentary. As a consequence of the next generation sequencing revolution genotyping data are now available for a wide diversity of accessions of each of the major crops, including sorghum. The volumes of phenotypic data necessary to construct predictive statistical models are, however, not available. We have develped and deployed a variety of image-based tools to collect such data. Here, we report on our development of a semi-automated feature extraction pipeline for use with RGB images and its application to sorghum panicles.