Abstract:Recent years, frequent floods have posed enormous challenges to urban systems worldwide. Real-time accurate urban flood information has a significant impact on emergency decisions. With the growth of citizen sensor science, video image is becoming a novel data source and emerging great potential in urban flood management. Recent studies have focused on detecting the flooded states of reference objects, like vehicles present in the video images to estimate urban flood. However, current detection models primarily focus on the accuracy testing while ignoring the lightweight and edge terminal execution, which limits their widespread deployment and applications. YOLOv8n, mainly oriented to edge terminals, has proven feasible in vehicle flooded state detection, but its performance in complex scenarios, such as multi-scale mixed vehicles and fuzzy targets, is yet to be enhanced. To address these challenges faced by YOLOv8n, an advanced approach is contributed in this research. Initially, Distributed Shifted Convolution is introduced to compensate for YOLOv8n's shortfall in flooding vehicle detection, and thirty improved options are implemented and trained with the flooded vehicle dataset. Subsequently, the optimal option is obtained as a new model, called DSC-YOLOv8n. Comparison with YOLOv8n and classic YOLO family models, DSC-YOLOv8n presents superior performance with the mAP50 of 75.4 %, representing a 3.6 % increase over baseline YOLOv8n, mAP50:95 of 58.6 %, marking a 3.1 % improvement, and the FLOPS of 7.4 G, achieving a decrease of 9.8 %. And then, the advancement and versatility of DSC-YOLOv8n in urban flood level detection are evidenced by its behavior in complex scenarios prediction and real case applications. Totally, this study successfully demonstrates the proposed method to combine both accuracy and lightweight in urban flood level detection and further promotes video images in urban flood management.

Estimating the spatial-temporal distribution of urban street ponding levels from surveillance videos based on computer vision

Toward Improved Real‐Time Rainfall Intensity Estimation Using Video Surveillance Cameras

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

A Bayesian Deep Learning Approach for Video-Based Estimation and Uncertainty Quantification of Urban Rainfall Intensity

Estimating Rainfall Intensity Using an Image-Based Deep Learning Model

Automatic detection of actual water depth of urban floods from social media images

Multi-Feature Fusion Based Outdoor Water Hazards Detection

Hazards Detection Multi-Feature Fusion Based Outdoor Water

Semi-supervised identification and mapping of surface water extent using street-level monitoring videos

A deep learning technique-based automatic monitoring method for experimental urban road inundation

A Real-Time Detecting Method for Continuous Urban Flood Scenarios Based on Computer Vision on Block Scale

DSC-YOLOv8n: An advanced automatic detection algorithm for urban flood levels

Improving the Accuracy of Urban Waterlogging Simulation: A Novel Computer Vision-Based Digital Elevation Model Refinement Approach for Roads and Densely Built-Up Areas

Towards Improved Real‐Time Rainfall Intensity Estimation Using Video Surveillance Cameras

ISE-UFDS: A Dataset for Detecting the Degree of Danger to Vehicles in Urban Flooding and Performance Assessment

Real‐time prediction and ponding process early warning method at urban flood points based on different deep learning methods

Automated Flood Depth Estimates from Online Traffic Sign Images: Explorations of a Convolutional Neural Network-Based Method

Adaptive iterative thresholding segmentation method guided by optical prior water body information in extracting urban flood inundation areas of synthetic aperture radar images

Identification of pedestrian submerged parts in urban flooding based on images and deep learning

Dynamic flood mapping by a normalized probabilistic classification method using satellite radar amplitude time series

Automatic extraction of flood inundation areas from SAR images: a case study of Jilin, China during the 2017 flood disaster