"Low-Altitude Cross-View (LACV) dataset"

"The LACV dataset was collected using a UAV (DJI Mini 3 Pro) and a self-built UGV platform that was mounted with a RealSense D455 camera. The LACV dataset was collected on the Baoshan Campus of Shanghai University to support cross-view image matching and localization research. It contains 1200 UAV images and 2000 forward-view UGV images, and all of them are combined with high-precision GPS information. The UAV images are captured at about 100m altitude. The resolutions of UAV and UGV images are 1920 \\times 1080. The UGV images are captured sequentially, allowing the trajectory of the UGV to be inferred. The scenes include roads, buildings, and green spaces, offering high complexity and diversity. UAV\u2013UGV image pairs were first generated through GPS-based automatic pairing and then refined manually, yielding a high-quality benchmark for visual localization and scene understanding."

搜集汇总

背景与挑战

背景概述

该数据集由无人机和地面车辆在校园采集，包含1200张无人机图像和2000张地面车辆图像，均配备高精度GPS信息，支持跨视角图像匹配与定位研究。场景涵盖道路、建筑和绿地，具有高复杂性和多样性，图像对通过GPS自动配对并人工精炼，为视觉定位和场景理解提供高质量基准。