Constant-time monocular self-calibration
Constant-time monocular self-calibration
复制标题
恒定时间单目自校准
DOI：
10.1109/robio.2014.7090561
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Gabe Sibley
中科院分区：
文献类型：
--
作者：
Nima Keivan;Gabe Sibley
本文描述了一种可扩展的框架，用于在同步建图和定位（SLAM）设置中进行相机实时自校准。该系统被证明可以根据未知的初始参数校准针孔和鱼眼相机模型，同时无缝地实时解决最大似然在线 SLAM 问题。自校准是通过跟踪图像特征来进行的，不需要预先确定校准目标。通过自动识别和仅使用序列中包含用于校准目的的有用信息的那些部分，系统可以在恒定时间与图像数量之间增量地获得准确的结果。此外，不需要特殊的初始化动作。该框架估计的参数与批量解决方案以及离线校准值密切匹配，但是在恒定时间内实时计算的。通过不将信息滚动到假设的先验分布中，系统避免了早期线性化引起的不一致——这是限制过滤技术的问题。该系统通过实验数据进行评估，并显示与离线和批量校准估计相比都是准确的。
This paper describes an extensible framework for real-time self-calibration of cameras in the simultaneous mapping and localization (SLAM) setting. The system is demonstrated to calibrate both pinhole and fish-eye camera models from unknown initial parameters while seamlessly solving the maximum likelihood online SLAM problem in real-time. Self-calibration is performed by tracking image features, and requires no predetermined calibration target. By automatically identifying and using only those portions of the sequence that contain useful information for the purpose of calibration the system achieves accurate results incrementally and in constant-time vs. the number of images. Furthermore, no special initialization movements are necessary. Parameters estimated by the framework are shown to closely match the batch solution as well as offline calibration values, but are computed live in constant-time. By not rolling information into an assumed prior distribution, the system avoids inconsistencies caused by early linearization - a problem that limits filtering techniques. The system is evaluated with experimental data and shown to be accurate vs. both the offline and batch calibration estimates.