A fully automatic classification of bee species from wing images

Since bees are the main pollinators of natural and agricultural ecosystems, they play a fundamental role in the preservation of the environment and food production. However, species identification is one of the bottlenecks for bee conservation due to its complex taxonomy, a large number of existing species, and the scarcity of professional taxonomists. In this sense, the automatic identification of such species can present a good alternative for non-taxonomist scientists and to the general public. In this work, we propose a fully automatic bee identification system based on the patterns of forewing venation. Our system was based on a combination of image segmentation techniques followed by a simple classification method. We achieved an accuracy of 99% in the genus and 96% in the species in a dataset composed of 48 species and 23 genera. This result represents an advance compared to previous works in the literature and there are plans to make the system online available for the general public.