Multimodal-Fusion and Recognition of Flapping-Wing Targets Using Laser Micro-Doppler Combined with Polarization and Thermal Radiation Characteristics

In recent years, bionic flapping-wing drones, as an important branch of unmanned aerial vehicle technology, have achieved remarkable breakthroughs and have been widely used in both military and civilian fields. Due to their bird-like appearance, flapping-wing drones possess strong concealment, which poses significant challenges for detection and identification. Currently, micro-motion feature analysis and recognition algorithms for flapping-wing targets remain underdeveloped. Achieving accurate identification of flapping-wing targets has become a critical and urgent problem in the field of drone recognition. This paper proposes an integrated analytical approach combining micro-Doppler, polarization, and infrared imaging features to achieve precise identification of flapping-wing targets. By extracting discriminative features including dual-peak consistency, swept-wing frequency shift, and depolarization rate (threshold: 0.5), along with infrared texture parameters, the method effectively distinguishes between biological organisms and mechanical systems. The experimental results obtained through the proposed multidimensional feature fusion criterion, combined with Random Forest, Support Vector Machine, and Multilayer Perceptron algorithms for feature fusion and target identification, demonstrate that under noiseless conditions, classification accuracy using only micro-Doppler features reaches 66.66%, 66.67%, and 33.33% respectively, while fusion of 3D features significantly improves accuracy to 99.83%, 99.67%, and 95.33%. These results confirm the effectiveness and reliability of multidimensional feature fusion for flapping-wing target recognition, establishing a foundation for subsequent research.