WHU-Stereo
收藏资源简介:
WHU-Stereo是由武汉大学创建的一个用于高分辨率卫星图像立体匹配的挑战性数据集。该数据集包含超过1700对经过校正的图像对,覆盖中国六个地区,涵盖多种地形和景观。数据集的创建过程中,使用了航空LiDAR点云和GF-7卫星的高分辨率立体图像,通过提出的一种遮挡移除技术来生成高质量的地面真值视差图。WHU-Stereo数据集旨在解决立体匹配中的挑战,特别是在遮挡和精度方面的挑战,适用于深度学习模型的训练和测试,以及作为评估模型性能的基准。
WHU-Stereo is a challenging dataset developed by Wuhan University for high-resolution satellite image stereo matching. It contains over 1,700 rectified stereo image pairs, covering six regions across China and encompassing diverse terrains and landscapes. During the dataset's development, airborne LiDAR point clouds and high-resolution stereo images from the GF-7 satellite were used, and a proposed occlusion removal technique was applied to generate high-quality ground-truth disparity maps. The WHU-Stereo dataset is designed to address key challenges in stereo matching, particularly those associated with occlusion and accuracy, and is suitable for training and testing deep learning models, as well as serving as a benchmark for evaluating model performance.

- 1WHU-Stereo: A Challenging Benchmark for Stereo Matching of High-Resolution Satellite Images武汉大学 · 2022年



