基于YOLOv5s的CCTV排水管道缺陷识别方法研究

Research of CCTVDrainage Pipeline Defect Recognition Method Based on YOLOv5s

城市排水管道常出现堵塞、老化、腐蚀、破损等情况,严重影响城市的健康发展和市民的正常生活。通过分析当前排水管道缺陷识别方法的局限性,构建了基于YOLOv5s的排水管道缺陷识别模型,对10种常见的排水管道缺陷进行学习、训练和验证,通过数据收集、数据处理、模型训练和目标检测4个步骤,完成了排水管道缺陷的智能识别。实验结果表明,排水管道缺陷识别模型的多类别平均精确度达到了85.42%(其中错口、异物穿入、渗漏(附着型)和破裂4种缺陷识别的平均精确度分别达到了93.6%、91.7%、91.7%和91.3%),验证了该模型的有效性。另外,YOLOv5s模型相较于Faster R-CNN模型,多类别平均精确度更高,并且具有模型内存更小、检测速度更快的明显优势,大幅提升了模型的工程适用性。

Urban drainage pipelines are often blocked,aged,corroded and damaged,which seriously affects the healthy development of the city and the normal life of citizens.By analyzing limitations of the current drainage pipeline defect recognition method,a drainage pipeline defect recognition model was conducted based on YOLOv5s to learn,train and verifie 10 common drainage pipeline defects.The intelligent identification was finished by four steps of data collection,data processing,model training and target detection.The experimental results show that the mean average precision of the drainage pipeline defect recognition model reaches 85.42%(the average precision of the four defect recognitions of dislocation,foreign body penetration,leakage and rupture reaches 93.6%,91.7%,91.7%and 91.3%,respectively),which verifies the effectiveness of the model.In addition,compared with the Faster R-CNN model,the YOLOv5s model has higher mean average precision and the obvious advantages of smaller model memory and faster detection speed,which significantly improve the engineering applicability of the model.