YOLOv8-RE4D: An enhanced UAV detection model for complex backgrounds with improved small-target recognition

Despite the widespread application of unmanned aerial vehicles (UAVs), these systems present numerous safety concerns. Conventional radar and optoelectronic detection methods exhibit certain constraints when identifying UAVs. This research introduces the YOLOv8-RE4D framework specifically designed for detecting low-altitude, slow-moving, and small-sized UAVs in intricate environments. Building upon the YOLOv8 architecture, the proposed model incorporates several enhancements: the Wise-IoU loss function for improved boundary regression precision, the integration of RepLKGELAN components and EMA attention mechanisms within the backbone network to strengthen feature extraction capabilities, and the addition of a specialized detection head for small objects to boost recognition performance. Experimental results demonstrate that the YOLOv8-RE4D model achieves a mean average precision of 93.5% on the Dut Anti-UAV benchmark dataset, outperforming both the baseline model and other state-of-the-art approaches while satisfying requirements for both real-time operation and detection accuracy in UAV monitoring applications.