Using image processing and machine learning to distinguish between benign and dangerous insect species in photographs and videos

Invasive animal and plant species can be a serious threat to the health, well-being or even survival of indigenous local species. In much of Western Europe, the Asian 'yellow legged' hornet, Vespa velutina, is causing serious problems for vital pollinator species, including bees, whilst the Oriental hornet, Vespa orientalis, is causing similar problems in n the Southern Mediterranean. If left unchecked, these invasive hornets could post a major threat to native pollinators, and hence to the farming of many crops, to many wildflower plants and to the availability of bee products such as honey, beeswax and propolis. People are encouraged to report sightings of such invasive species, but non-experts may often misidentify superficially similar-looking harmless insects as the dangerous species, leading to the wasting expert time and effort of experts, and considerable expense to investigate such 'false alarms'. In this paper, we describe the use of image processing and machine learning to distinguish between the invasive Asian hornet from relatively harmless, but superficially similar in appearance, common wasp and European hornet in digital photographs and 'home videos'. The initial results are very encouraging and should help authorities locate and control such invasions of dangerous insects with fewer false alarms and less wastage of resources.