基于YOLOv8的铁路隐蔽工程影像目标检测方法及应用

Detection Method and Application of Railway Hidden Engineering Image Target Based on YOLOv8

铁路隐蔽工程影像作为留存查证的重要资料常包含海量工程数据,现有的影像资料管理系统为工程验收提供重要平台。为提高工程影像的审查效率,实现隐蔽工程审查少人化和规范影像验收管理,选择较为成熟的目标检测模型YOLOv8,建立可保障误差方向传播过程可执行的损失函数CIOU,用于精准识别验收人员、标识牌、量尺图像等验收目标。选取精确率、召回率,以及精确率与召回率的加权调和平均值作为衡量模型的评价指标。试验结果显示验收人员和标识牌的检测精度均在90%以上,量尺在80%以上,且相较于其他模型具有更好的目标检测效果。隐蔽工程自动识别处理软件可服务于工程影像检查要素的自动识别。研究成果可为规范铁路工程施工影像验收管理及隐蔽工程检查验收智能化管理提供参考。

As an important data to be retained and verified,railway hidden engineering images often contain massive engineering data.The existing image data management system provides an important platform for engineering acceptance.In order to improve the efficiency of engineering image review,realize less human hidden engineering review and standardize image acceptance management,a relatively mature target detection model YOLOv8 was selected,and an executable loss function CIOU was established to ensure the error direction propagation process,which was used to accurately identify acceptance targets such as acceptance personnel,identification plates and measuring scale images.The accuracy rate,recall rate,weighted harmonic average F1 of accuracy rate and recall rate were selected as the evaluation indexes of the model.The experimental results show that the acceptance personnel and the identification plate respectively reach more than 90%,and the detection accuracy of the measuring scale is more than 80%,and it has better target detection effect than other models.Hidden engineering automatic recognition processing software can serve the automatic recognition of engineering image inspection elements.The research results can provide reference for standardizing the inspection and acceptance management of railway construction image and the intelligent inspection and acceptance management of hidden engineering.