An End-to-End Recognition Method for IR-UWB Radar Dynamic Detection Mode for Detecting Targets in Fire Rescue Scenarios

In fire rescue operations characterized by dense smoke, water droplets, and thermal currents, impulse radio ultrawideband (IR-UWB) radar systems offer inherent advantages, such as high precision, interference resistance, and thermal resilience. However, a critical challenge persists in these scenarios: the inability to capture targets in motion, hindering their applicability in firefighting scenarios where the dynamic measurements are essential for the firefighter safety. Building upon a comprehensive mathematical model, this study validates the feasibility of target identification based on the waveform recognition through rigorous simulation. Leveraging this foundation, we propose DFCnet, a novel target recognition network comprising modules for the target detection, feature extraction, and classification. DFCnet integrates innovative methods, such as potential target extraction, trajectory-based feature fusion, and multidomain features fusion for improved target identification accuracy and reduced computational complexity. Through comparative experiments, DFCnet demonstrates superior classification accuracy and significantly reduced inference time compared to the mainstream image classification networks. Moreover, in overall target detection experiments, DFCnet outperforms the existing methods with a 6% miss detection rate. This research contributes to bridging the gap between IR-UWB radar technology and practical firefighting applications, offering enhanced capabilities for the firefighters to detect trapped individuals in dynamic firefighting environments.