AI in apiculture: A novel framework for recognition of invasive insects under unconstrained flying conditions for smart beehives. (March 2023)

• Record Type: Journal Article
• Authors: Nasir, Abdul; Ullah, Muhammad Obaid; Yousaf, Muhammad Haroon
• Is Part Of: Engineering applications of artificial intelligence. Volume 119(2023)
• Publication Date: 2023-03
• Subjects: Apiculture -- Deep learning -- Feature engineering -- Image recognition -- Invasive insects; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2022.105784
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: Uncompromised population growth of invasive insects endangers biodiversity, agribusinesses, and ecosystems. One such example is the invasion of Vespa hornets in honey harvesting areas which leads to a devastating impact on honeybee ecology and subsequently on the economic activities linked with it. To mitigate the imposed threat, a vision-based system capable to recognize Vespa hornets near beehives can serve the purpose. But the arbitrary position and pose of these small-scale fast-moving insects with respect to the camera viewpoint in natural daylight make it challenging to realize the task. Keeping in view the situational intricacies, a novel AI-based framework is proposed to recognize Vespa hornets near beehives under unconstrained flying conditions using a multi-modal data and multi-evidence approach. Multiple modalities include 3-D trajectories and IR imagery while multiplicity in evidence evolves through the retrieval of IR images from multiple spatial locations furnished by the insects' trajectories. The proposed framework exploits the information provided by a limited piece of evidence selected at random from multi-modal and multi-evidence observation sets through pre-evaluated deep learning/machine learning models. Individual inferences from selected recognition models are then fused using a weighted summation scheme to make the final decision. The recognition framework demonstrated a classification accuracy of 97.1% for two hornet types of the genus Vespa along with the honeybee Apis mellifera . The proposed strategy indicating promising results is a pioneering work of applying AI in the domain of entomology and apiculture to have the detection capability for invasive insects in the vicinity of beehives to make them safer and smarter.