GPS注册的立体视觉数据集

本数据集由克罗地亚萨格勒布大学电气工程与计算学院创建，旨在评估立体视觉里程计的准确性。数据集包含2786帧图像，通过安装在移动车辆顶部的短基线立体装置获取，地面实况由消费级GPS设备提供。数据集创建过程中，通过后处理恢复了GPS读数与立体帧之间的同步和校准。该数据集主要应用于自动驾驶和驾驶员辅助系统，旨在解决城市环境中视觉里程计的精度和可靠性问题。

This dataset was developed by the Faculty of Electrical Engineering and Computing, University of Zagreb, Croatia, with the purpose of evaluating the accuracy of stereo visual odometry. It comprises 2786 image frames captured by a short-baseline stereo rig mounted on the roof of a moving vehicle, while its ground truth data is provided by consumer-grade GPS devices. During the dataset construction, synchronization and calibration between GPS readings and stereo image frames were restored through post-processing. This dataset is primarily utilized for autonomous driving and driver assistance systems, targeting the resolution of accuracy and reliability issues of visual odometry in urban environments.