基于改进YOLOv5的航空器关键部件与工作人员目标检测算法OA

Enhancing the accuracy of target detection for critical aircraft components and ground personnel in apron scenes is crucial for aircraft flight safety and airport operational efficiency.In response to the scarcity of existing models for critical aircraft component detection and the low accuracy of airport ground personnel detection models in apron settings,a DT-YOLO object detec-tion algorithm based on Transformer improvements is designed.This algorithm builds upon the YOLOv5 model,incorporating the global attention mechanism of the Transformer with the local feature extraction capability of CNNs,and introduces a Drop layer to eliminate redundant and noise features,creating a new D-CTR module to boost the model's adaptability and recognition precision in complex backgrounds and for targets of various scales.An improvement in the loss function addresses the discontinuity and instabili-ty issues of GIoULoss under certain conditions,which can lead to sudden changes in loss values and gradient explosions,thereby further enhancing the stability and accuracy of model training.Experimental results demonstrate that,with a parameter count similar to the baseline model,the mAP of the DT-YOLO model has increased by 2.6%compared to the baseline model,and it exhibits su-perior detection performance when compared with other mainstream object detection models.