摘要
将人工智能领域里的目标检测方法应用到城市内涝监测中.首先,选取YOLOv5(You Only Look Once version 5)算法用于城市洪水参照物模型的训练;其次,根据监控设备所拍摄的洪水期与非洪水期参照物的高度差来计算道路的内涝深度信息;最后,通过一个水池实验来验证该方法的有效性,将人工计算数据与YOLOv5自动监测数据作对比,实验结果显示RMSE值为0.007 m,MAPE值为2.738%.结果表明该方法能够监测城市内涝水深状况,具有检测速度快、准确性高、成本低等优势.
The paper applies target detection methods in the field of artificial intelligence to urban waterlogging monitoring.Firstly,the YOLOv5(You Only Look Once version 5)algorithm is selected for the training of urban flood reference model.Secondly,the waterlogging depth of the road is calculated based on the height difference of the reference objects measured by monitoring devices during flood and non-flood periods.Finally,the effectiveness of the method is verified by a pool experiment,in which the manually calculated data are compared with the YOLOv5 automatic monitoring calculated data,and the results show an RMSE value of 0.007 m and an MAPE value of 2.738%.Therefore,the method is able to monitor the water depth of urban waterlogging,with the advantages of fast detection,high accuracy and low cost.
作者
柳进元
张明锋
LIU Jinyuan;ZHANG Mingfeng(Key Laboratory for Humid Subtropical Ecogeographical Processes of the Ministry of Education,School of Geographical Sciences,Fujian Normal University,Fuzhou 350117,China)
出处
《福建师范大学学报（自然科学版）》
CAS
2023年第1期86-92,共7页
Journal of Fujian Normal University：Natural Science Edition
基金
国家自然科学基金资助项目(41871146)
福建省科技厅公益类科研项目(2021Rl002006)。
关键词
YOLOv5
图像识别
城市内涝
水深提取
YOLOv5
image recognition
urban waterlogging
water depth extraction
