Accepted_test

Motivation and Aim: The morphometric characteristics of a wheat spike are among the most important for geneticists and breeders, since they are closely related to such economically valuable qualities as productivity, lack of fragility of the ear and ease of threshing. Statistical methods (QTL, GWAS) are used to identify the genes that control these traits, for the success of which it is necessary to collect and analyze a large amount of phenotypic data. Currently, in most studies, phenotyping is performed by experts based on visual analysis of the ear and manual measurements, which is labor-intensive. To increase the efficiency of ear phenotyping, methods based on automatic image analysis are needed.

Methods and Algorithms: For the phenotyping of wheat spike, we developed methods based on the analysis of two-dimensional images. These methods use both computer vision and deep machine learning methods and allow to determine the areas of the spike and awns in the image, determine the contour of the ear body, determine its geometric parameters, count the number of spikelets in the spike, evaluate the color of the spike and awns, classify spikes according to the glume pubescence and ploidy of the plant.

Results: The developed methods were used to evaluate the characteristics of ears of different types of wheat from the collection of N.P. Goncharov and the collection of bread wheat from SibNIIRS.

Conclusion: The use of methods for automatic analysis of digital images of wheat spikes allows for a mass analysis of their characteristics with high productivity and accuracy.