The UAV Target Detection Algorithm Based on Improved YOLO V8

UAV image detection has a wide range of applications in the fields of intelligent surveillance, disaster assessment and environmental protection. However, UAVs are non-cooperative targets that seriously threaten the safety of airspace. YOLO (You Only Look Once) series algorithms, as the classic algorithms for target detection, face the complex UAV image environment, and there is still room for improvement of the traditional YOLOv8 in special scenarios. Based on YOLOv8 algorithm, this paper proposes an improved UAV image detection method by combining the ParNet attention mechanism and the RepVGG reparameterisation module. The ParNet attention mechanism improves the model's ability to pay attention to different spatial positional features by introducing the position-sensitive weights, which enhances the detection capability of the target. RepVGG reparameterisation module The complex network structure is used to extract rich features in the training phase, while it is reparameterised to a simple convolutional network in the inference phase, which improves the inference speed and efficiency. Experimental results show that the improved YOLOv8 has a detection accuracy of 96.6% and a recall of 97.3% on the UAV image dataset. These improvements demonstrate the effectiveness and application potential of the proposed method in UAV image detection.