Detection of Urban Flood Inundation from Traffic Images Using Deep Learning Methods

• DOI: 10.1007/s11269-023-03669-9
• Publication Date: 2023-12-02T11:01:49Z
• Authors: Pengcheng Zhong; Yueyi Liu; Hang Zheng; Jianshi Zhao
• Keywords: 0207 environmental engineering; 02 engineering and technology

Abstract Urban hydrological monitoring is essential for analyzing urban hydrology and controlling storm floods. However, runoff in areas, including flood inundation depth, often inadequate. This inadequacy hampers the calibration of models limits their capacity early warning. To address this limitation, study established a method evaluating depth floods using image recognition deep learning. utilizes object model YOLOv4 to identify submerged objects images, such as legs pedestrians or exhaust pipes vehicles. In dataset 1,177 mean average precision water reached 89.29%. The also found that accuracy by influenced type reference flood; use vehicle yielded higher than person. Furthermore, augmentation with Mosaic technology effectively enhanced recognition. developed extracts on-site, real-time, continuous data from images video provided existing traffic cameras. system eliminates need installing additional gauges, offering cost-effective immediately deployable solution.