卫星图像立体基准数据集
收藏资源简介:
本数据集由普渡大学电气与计算机工程学院创建,旨在为多时相卫星图像的立体重建研究提供支持。数据集包含10个关注区域(AOI)的立体校正图像及其相应的地面实况差异,其中8个AOI来自IARPA的MVS挑战数据集,2个来自CORE3D-Public数据集。数据集主要使用WorldView-3(WV3)图像,UCSD区域则同时使用WV3和WorldView-2(WV2)图像。数据集不仅包括图像,还包含每幅图像的获取日期和时间、相机方位和仰角以及立体对之间的交角等元数据。此外,数据集还提供了建筑物掩码,以确保立体估计的差异在建筑物区域更为可靠。该数据集适用于解决由于季节变化导致的立体匹配难题,并可用于评估立体匹配算法的性能。
This dataset was created by the School of Electrical and Computer Engineering at Purdue University, and is purpose-built to support research on stereo reconstruction of multi-temporal satellite imagery. The dataset contains stereo-corrected images and their corresponding ground-truth disparities for 10 Areas of Interest (AOIs). Of these, 8 AOIs are sourced from the IARPA MVS Challenge Dataset, while the remaining 2 AOIs are taken from the CORE3D-Public Dataset. The dataset primarily utilizes WorldView-3 (WV3) imagery, with the UCSD region additionally employing both WorldView-3 (WV3) and WorldView-2 (WV2) imagery. Beyond the image data, the dataset includes rich metadata such as the acquisition date and time of each image, camera orientation and elevation angle, as well as the intersection angle between stereo pairs. Moreover, the dataset provides building masks to ensure that stereo-estimated disparities exhibit greater reliability within building-covered areas. This dataset can be used to address stereo matching challenges induced by seasonal variations, as well as to evaluate the performance of stereo matching algorithms.

- 1A New Stereo Benchmarking Dataset for Satellite Images电气与计算机工程学院,普渡大学 · 2019年



