摘要： 近些年，人工巡检的管理方式导致MODF端口状态的信息准确率较低，无法区分占用端口与虚占端口。针对MODF资源管理中的端口状态识别问题，提出了一种改进的YOLOv7目标检测模型。首先，鉴于数据集采集困难且类别不均衡,采用了多种数据增强方法来扩充数据集；此外，在骨干网络中使用共享权重的RFEM，扩大端口目标的感受野，减少训练过程中的过拟合风险；其次提出F-EMA注意力模块，以提高对空间上下文信息的利用率，减少因端口接近或被遮挡而导致的漏检、误检等情况；最后，使用NWD损失函数替代IoU度量，减轻对小目标位置偏差的敏感性，提升密集小物体检测准确率。实验显示，改进模型的mAP@.5值达98.8%，相比原Yolov7模型提升了2个百分点，mAP@.5:.95值达63.8%，提升了9.5个百分点，改进提高了MODF端口资源利用率,满足了智能巡检系统对于端口占用状态识别准确率的基本要求。

Abstract: In recent years, the manual inspection management method has led to the low accuracy in identifying the status of MODF ports, making it difficult to differentiate between occupied and unoccupied ports.. Aiming at the problem of status recognition in MODF port resource management, the paper proposes an improved YOLOv7 object detection model. Firstly, due to the difficulty of data collection and unbalanced categories, multiple data enhancement methods are used to expand the data set. In addition, Additionally, a shared-weight RFEM is used in the backbone network to enlarge the receptive field of the port targets and reduce the risk of overfitting during the training process; Then, the F-EMA attention module is proposed to improve the utilization of spatial context information and reduce missed detections and false alarms caused by ports being close or occluded; Finally, NWD loss function is used to replace the IoU measurement, which alleviates the sensitivity to the position deviation of small targets and improves the detection accuracy of dense small objects. The experimental show that the mAP@.5 value of the improved model reaches 98.8%, which is 2 percentage points higher than that of the original Yolov7 model, mAP@.5:.95 value reaches 63.8%, which is 9.5 percentage points higher, which improves the utilization rate of MODF port resources and meets the basic requirements of the intelligent inspection system for the accuracy of port occupancy status recognition.