lion-responses-to-aerial-monitoring-dataset
收藏资源简介:
本数据集是一个多模态无人机野生动物观测数据集,旨在研究非洲狮对空中监测的行为反应,并支持野生动物监测和生态感知研究。数据采集于肯尼亚中部奥佩杰塔保护区,在2024年10月至12月期间的15个工作日(08:00-19:00)完成。数据集包含同步记录的RGB视频(3840×2160分辨率,30 fps)和热成像视频(640×512分辨率,30 fps),由DJI Mavic 3 Enterprise T无人机在20至120米的不同固定高度进行受控实验性接近飞行时捕获。数据以原始视频文件(MP4格式)及其伴随的SRT遥测文件形式存储,后者包含每帧的GPS、高度、相机设置和UTC时间戳。此外,数据集提供了遵循达尔文核心(Darwin Core)标准的元数据表格(CSV格式),分别记录每个视频片段(occurrence)和每次完整飞行任务(event)的详细信息,如事件ID、日期时间、地理位置、海拔、物种分类信息(非洲狮,Panthera leo)、个体数量统计(成年雌性、成年雄性、幼崽)以及观察到的行为。该数据集适用于多种计算机视觉任务(如目标检测、实例分割、多目标跟踪、重识别、行为识别和RGB-热成像跨模态学习)、生态学研究(如种群丰度估计、行为分析、栖息地利用模式)以及机器人应用(如无人机感知系统基准测试和传感器融合研究)。数据集遵循CC BY 4.0许可协议发布。
This dataset is a multimodal drone-based wildlife observation dataset, designed to study the behavioral responses of African lions to aerial monitoring and to support wildlife monitoring and ecological perception research. Data were collected in the Ol Pejeta Conservancy in central Kenya over 15 working days (08:00-19:00) between October and December 2024. The dataset includes synchronously recorded RGB videos (3840×2160 resolution, 30 fps) and thermal imaging videos (640×512 resolution, 30 fps), captured by a DJI Mavic 3 Enterprise T drone during controlled experimental approach flights at varying fixed altitudes ranging from 20 to 120 meters. The data are stored as raw video files (MP4 format) along with accompanying SRT telemetry files, which contain GPS coordinates, altitude, camera settings, and UTC timestamps for each frame. Additionally, the dataset provides metadata tables (CSV format) following the Darwin Core standard, detailing each video clip (occurrence) and each complete flight mission (event) with information such as event ID, date-time, geographic location, elevation, species classification (African lion, Panthera leo), individual counts (adult females, adult males, cubs), and observed behaviors. The dataset is suitable for various computer vision tasks (e.g., object detection, instance segmentation, multi-object tracking, re-identification, behavior recognition, and RGB-thermal cross-modal learning), ecological research (e.g., population abundance estimation, behavior analysis, habitat use patterns), and robotics applications (e.g., benchmark testing for drone perception systems and sensor fusion studies). The dataset is released under the CC BY 4.0 license.




