Experimental Validations of Insect Species Identification Based on Fully-Polarimetric Radar Measurements

Accurate identification of migratory insect species is essential for effective pest forecasting and control strategies. Radar entomology presents challenges in insect identification, prompting the exploration of novel approaches. This paper addresses the issue by leveraging machine learning techniques to classify insect species based on their body mass and length, fundamental parameters for characterization. Four insect species were collected via searchlight traps, and their body mass and length were promptly measured. An Artificial Neural Network (ANN) was tailored to these parameters to construct a species identification model. Using a fully-polarimetric radar, 40 insects-10 from each of the four species-were monitored in the air. The insects were affixed to their backs with thin PE line and lifted by drones to a fixed position. The results demonstrate the efficacy of the trained identification models, achieving an average identification rate of 87.5% through estimated body mass and length parameters. This study validates the feasibility of species identification based on insect size parameters, offering promising insights for radar entomology applications.