为解决YOLOv8n算法在机场异物检测中存在计算复杂度高、计算资源消耗大的问题，通过在YOLOv8n算法中引入轻量化模块的方法研究了机场异物检测的问题，提出了Fast-BiYOLOv8n算法。首先，设计了C2f_FasterEMA模块并引入YOLOv8n算法的骨干网络中，该模块融合了FasterBlock模块和高效多尺度注意力（efficient multi-scale attention, EMA）注意力机制，增强了图像的特征提取能力，同时降低了算法计算量；其次，在路径聚合网络（path aggregation network, PANet，）网络架构中融合了骨干网络中的P2特征层并设计了双向特征金字塔网络（bidirectional feature pyramid network, BiFPN）网络架构，增加了跨尺度连接促进了不同特征图之间的信息融合，同时加入C2f_Faster模块提高了特征融合的效率并进一步降低了算法的计算量；最后，通过改进损失函数为Inner-CIoU（intersection over union, complete intersection over union loss）加快了算法的收敛速度，提高了检测准确率。结果表明，Fast-BiYOLOv8n算法的检测准确率达到99.0%，召回率为98.8%，平均精度均值（mean average precision, mAP）提升了3.5个百分点，达到99.3%，参数量比原模型降低了27%，模型的权重大小降低了21%，实现了在降低算法参数量的同时，提升检测准确率的目的。

In order to solve the problems of high computational complexity and large computational resource consumption of the YOLOv8n algorithm in airport foreign object detection, the problem of airport foreign object detection was studied by introducing a lightweight module into the YOLOv8n algorithm, and the Fast-BiYOLOv8n algorithm was proposed. Firstly, the C2f_FasterEMA module was designed and introduced into the backbone network of the YOLOv8n algorithm. This module integrated the FasterBlock module and the EMA(efficient multi-scale attention) attention mechanism, which enhanced the feature extraction ability of the image and reduces the algorithm's computational load. Secondly, the P2 feature layer in the backbone network was integrated into the PANet(path aggregation network) network architecture, and the BiFPN(bidirectional feature pyramid network) network architecture was designed. The cross-scale connection was added to promote the information fusion between different feature maps. At the same time, the C2f_Faster module was added to improve the efficiency of feature fusion and further reduce the algorithm's computational load. Finally, the convergence speed of the algorithm was accelerated by improving the loss function to Inner-CIoU(intersection over union, complete intersection over union loss), and the detection accuracy was improved. The results show that the detection accuracy of the Fast-BiYOLOv8n algorithm reaches 99.0%, the recall rate is 98.8%, and the mAP (mean average precision) increases by 3.5 percentage points to 99.3%. The number of parameters is reduced by 27% compared with the original model, and the weight of the model is reduced by 21%, achieving the goal of improving detection accuracy while reducing the number of algorithm parameters.