基于YOLOv7的水工输水管道目标智能识别

Intelligent Target Recognition of Hydraulic Water Pipeline Based on YOLOv7

CCTV技术已广泛应用于水工输水管道病害检测,针对水工输水管道病害的识别仍然通过专业人员主观判断分析。文章利用广东省水工输水管道的实测图像构建水工输水管道病害数据集,并基于YOLOv7算法,实现对错口、破损、泄漏和障碍物的智能识别,4种识别的平均准确率分别为0.905、0.891、0.934和0.895。研究结果表明,模型识别的平均准确率可达90%以上,识别效果良好,可在实际工程中推广应用。

CCTV technology has been widely applied in the detection of water pipeline diseases in hydraulic engineering,and the identification of water pipeline diseases still relies on subjective judgment and analysis by professionals.The article first constructs a dataset of water transmission pipeline diseases based on measured images of water transmission pipelines in Guangdong Province.Based on the YOLOv7 algorithm,intelligent identification of wrong mouth,damage,leakage,and obstacles is achieved.The average accuracy rates for identifying wrong mouth,damage,leakage,and obstacles are 0.905,0.891,0.934,and 0.895,respectively.The research results show that the average accuracy of model recognition can reach over 90%,and it has good recognition effect for water pipeline diseases in hydraulic engineering,which can be promoted and applied in practical engineering.