Optimized Algorithm for Aircraft Surface Defect Detection Technology Based on YOLOV8n

To enhance the efficiency of detecting defects on aircraft surfaces, while achieving the goals of equipment lightweighting and real-time monitoring, this paper proposes an optimized YOLOv8n object detection algorithm. Firstly, add DyHead detection head in the model structure and improve the feature pyramid network (FPN) to enhance the model’s ability to detect defects of different scales. Secondly, the RGIOU loss function is used to replace the original model’s loss function. This not only significantly reduces the effort required for hyperparameter tuning but also enhances both the regression accuracy and convergence speed of the model. Finally, by replacing the backbone network of YOLOv8n with MobileNetV3, the computational complexity is significantly reduced while maintaining a high detection accuracy. The experimental results demonstrate that the optimized algorithm achieves outstanding performance on the aircraft surface defect data-set: compared to the original model, it has an average precision improvement of 4.86%, with reductions in parameters (12.13%), computational complexity (13.35%), and weight volume (10.12%). When benchmarked against mainstream object detection algorithms such as SSD, YOLOv4 to YOLOv8, the enhanced model demonstrates superior performance and accuracy, indicating promising application prospects in aircraft surface defect detection.