Brown Marmorated Stink Bug Detection Using Multiple Neural Networks

The identification and management of pest infestations in a timely manner have been an ever-pressing issue, more so than ever now, when new pests have arrived from around the world and governments are encouraging farmers to use fewer and fewer pesticides. The use of drones and artificial intelligence in the monitoring of orchards has taken a strong rise in recent years due to the advantages offered by efficiency, coverage, and price. A particularly dangerous insect in fruit production is the brown marmorated stink bug (BMSB). The paper seeks to look at a performance comparison in BMSB detection between three states of the art models of image detection and classification: MobileNetV2, Xception, and EfficientNet. Each network model was slightly modified being thus adapted for this application and the results obtained had high accuracy, between 97% and 99%.