Reaserch on Aircraft Image Segmentation Algorithm Based on Improved YOLOv8

With the rapid development of target segmentation techniques, the YOLO family of algorithms has become popular due to its efficiency. In this paper, we propose an improved YOLOv8 model aimed at improving the performance of instance segmentation of aircraft images. We enhance the model's ability to capture the global dependencies of the aircraft in the image by introducing a Non-local attention mechanism, while integrating a bidirectional feature pyramid network (BiFPN) for finer feature fusion. Experimental results conducted on the publicly available COCO dataset aircraft category show that the improved YOLOv8 model outperforms the original model in several performance metrics, especially the significantly improved detection accuracy in complex backgrounds. These improvements provide effective technical support for real-time aircraft detection and segmentation, demonstrating the potential of attentional mechanisms and advanced feature fusion techniques for practical applications.