UAV Model Recognition Based on Multi-Station Collaborative Multi-Angle Attentional Feature Fusion

Traditional UAV detection and recognition systems rely primarily on a single radar, which can only detect local information from one side of the UAV. The measurement data has limited feature dimensions, and the target characteristics can change with the UAV's orientation, impacting recognition accuracy. This study performs research on UAV model recognition using multi-angle information fusion. It presents a UAV model recognition approach based on multi-station collaborative multi-angle attention feature fusion. An innovative multi-dimensional parallel residual convolutional attention neural network architecture is developed to extract critical multi-view properties of the target. In addition, a channel attention method is used for multidimensional feature fusion, allowing UAV recognition via multi-station collaboration. A comparison of single- and multi-station radar recognition was carried out. The results demonstrated that multi-station recognition improved accuracy significantly when compared to single-station recognition, demonstrating the efficacy of multi-station collaborative UAV model recognition.