Automated Image Analysis Framework for the High-Throughput Determination of Grapevine Berry Sizes Using Conditional Random Fields

摘要

The berry size is one of the most important fruit traits in grapevinebreeding. Non-invasive, image-based phenotyping promises a fast and precisemethod for the monitoring of the grapevine berry size. In the present study anautomated image analyzing framework was developed in order to estimate the sizeof grapevine berries from images in a high-throughput manner. The frameworkincludes (i) the detection of circular structures which are potentially berriesand (ii) the classification of these into the class 'berry' or 'non-berry' byutilizing a conditional random field. The approach used the concept of aone-class classification, since only the target class 'berry' is of interestand needs to be modeled. Moreover, the classification was carried out by usingan automated active learning approach, i.e no user interaction is requiredduring the classification process and in addition, the process adaptsautomatically to changing image conditions, e.g. illumination or berry color.The framework was tested on three datasets consisting in total of 139 images.The images were taken in an experimental vineyard at different stages ofgrapevine growth according to the BBCH scale. The mean berry size of a plantestimated by the framework correlates with the manually measured berry size by$0.88$.