"Flood Detection Dataset"

"Floods are among the most destructive and widespread natural disasters globally. They inundate vast areas, disrupt communities, and pose significant threats to both human and animal lives. Accurately predicting flood behavior is critical for improving emergency response, planning evacuation routes, and minimizing risks to first responders.Recent advancements in aerial imaging have shown considerable promise in enhancing flood monitoring and analysis. Among various aerial imaging technologies, Unmanned Aerial Vehicles (UAVs) and drones offer a practical and efficient means of capturing detailed flood-related data in real time.This study introduces an aerial imagery FLARE (Flood Level Aerial-based Remote Evaluation) dataset collected using drones during a controlled flood simulation in Southern Louisiana, USA. The dataset comprises multiple repositories, including raw aerial video captured by drone-mounted cameras, and thermal imaging data that reveals moisture distribution and water temperature anomalies.To support the research community, two primary flood-related tasks\u2014flood extent classification and flood region segmentation\u2014are defined based on the dataset. These applications aim to assist in the development of more effective flood detection and response systems."

洪涝灾害是全球范围内破坏性最强、分布最广的自然灾害之一。其会淹没广袤区域，扰乱社区秩序，对人畜生命构成严重威胁。精准预测洪涝态势，对于优化应急响应、规划疏散路线以及降低应急救援人员的风险至关重要。
近年来，航空成像技术的进步为提升洪涝监测与分析能力带来了显著前景。在各类航空成像技术中，无人驾驶航空器（Unmanned Aerial Vehicles, UAVs）与无人机可提供实用高效的手段，实时捕获详细的洪涝相关数据。
本研究介绍了一套航空影像FLARE（Flood Level Aerial-based Remote Evaluation，洪涝等级空中远程评估）数据集，该数据集采集自美国路易斯安那州南部的受控洪涝模拟实验，由无人机完成数据获取。
该数据集包含多个子仓库，涵盖无人机搭载相机拍摄的原始航空视频，以及可揭示湿度分布与水温异常的热成像数据。
为支持科研社区开展相关研究，研究者基于该数据集定义了两项核心洪涝相关任务：洪涝范围分类与洪涝区域分割。这些应用旨在助力开发更高效的洪涝检测与应急响应系统。