Recognition Method for Unmanned Aerial Vehicles Based on a Lightweight Dual-Channel Multiscale Convolutional Neural Network

With the increasing prevalence of unmanned aerial vehicles (UAVs) in daily applications, the accurate and timely detection of UAV targets is of paramount importance. As artificial intelligence continues to advance, deep learning technology has emerged as a key solution to this challenge. In light of this, we have developed a recognition model for UAV using a dual-channel multiscale convolutional neural network (CNN). This model initially enhances UAV recognition accuracy by integrating the range-Doppler (RD) diagrams and time-frequency distribution (TFD) diagrams of radar echo signals and then further refining it by fusing features of different scales. The accuracy of this model reaches 98.67% in the actual datasets, confirming its effectiveness in UAV surveillance applications.