Insect Species Identification Based on Polarimetric Scattering Feature Sequences Using MsRNN

This study introduces an original deep learning technique for the classification of insect species, which relies on the analysis of polarimetric scattering feature sequences. Utilizing a multi-scale Recurrent Neural Network (MsRNN) architecture, we achieve high-precision recognition of nine first-class migratory pest species using data collected from a high-resolution fully polarimetric radar. Our method achieves an average identification accuracy of 98.1%, demonstrating significant improvements compared to traditional classification techniques. This high accuracy makes our approach a valuable tool for pest management and agricultural protection, enabling more efficient and accurate monitoring of harmful migratory insect species.