Automatic Flower Number Evaluation in Grapevine Inflorescences Using RGB Images

Abstract

A precise count of flower number per inflorescence is essential to characterize the reproductive behavior of a vine. Previous efforts to automatize this process by image-based technologies have failed in the development of a universal system that can be applied to multiple grapevine cultivars, or they have been tested in a set of inflorescences of narrow morphological diversity. Here, we have developed an alternative general method in the open-source platform Fiji for the nondestructive counting of visible flowers in red-green-blue (RGB) images, considering inflorescences from 45 different grapevine genotypes from three progenies segregating for inflorescence morphology. The algorithm, based on the segmentation of the image into regions of interest according to their color and morphology, provided counting results highly correlated to manual ones (R2 = 0.91). Similar results were obtained when validating this tool in an external data set of 400 images of four grapevine cultivars. Counting values were used for actual flower number estimation by linear modeling using a subset of 45 images, considering a flower density factor to reduce the adverse effect of the variable number of hidden flowers. Our approach allowed the estimation of flower number with satisfactory results, providing useful information for grapevine breeding and research.

Abstract from the paper’s OpenAlex record, shown for indexing and research. Copyright remains with the authors or the publisher. Read at the source ↗

The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.