文献类型：期刊文献

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

作者：Sun, Jiwu[1]; Zhang, Cheng[1]; Xu, Cheng[1]; Wang, Pengfei[1,2]; Liu, Hongzhe[1]

第一作者：Sun, Jiwu

机构：[1] Beijing Key Laboratory of Information Service Engineering, Beijing Union University, Beijing, China; [2] Big Data Center, Ministry of Emergency Management, Beijing, China

第一机构：北京联合大学北京市信息服务工程重点实验室

通讯机构：[1]Beijing Key Laboratory of Information Service Engineering, Beijing Union University, Beijing, China|[11417103]北京联合大学北京市信息服务工程重点实验室;[11417]北京联合大学;

年份：2024

卷号：14868 LNCS

起止页码：402-413

外文期刊名：Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

收录：EI(收录号：20243316873825)

语种：英文

外文关键词：Floods - Hazards - Statistical tests - Storms - Vehicle performance

摘要：As global warming and urbanisation continue to accelerate, resulting in the increasing likelihood and uncertainty of extreme rainstorms and floods, how to detect things in flooding scenarios and implement relevant rescue measures has become an urgent problem to be solved. Flooding scenarios are difficult and dangerous to obtain data, which leads to relatively few data sets dedicated to the detection of the degree of danger of vehicles in flooding scenarios. To this end, a dataset for vehicle hazard detection in urban flooding is proposed and the YOLOv8s algorithm is improved to increase the detection accuracy. The proposed dataset aims to provide realistic, diverse and challenging vehicle images in flooding scenarios, including different flood hazard scenarios and time periods. The dataset contains a total of 20,152 images, which are divided into training, validation and test sets in the ratio of 8: 1: 1 and evaluated and validated on the existing target detection algorithms. The authenticity and accuracy of the dataset is ensured by collecting data from real flooding sites. ? The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.