数学原理：
http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/ROSENHAHN1/PoseEstimation.pdf
http://openaccess.thecvf.com/content_cvpr_2018/papers/Grabner_3D_Pose_Estimation_CVPR_2018_paper.pdf
【泡泡点云时空】2D3D-MatchNet: 学习匹配2D图像和3D点云的关键点
http://www.sohu.com/a/342106564_715754
【泡泡一分钟】用于大规模3D地图中相机定位的高效全局2D-3D匹配方法
https://www.sohu.com/a/273507198_715754
【领域报告】2017以来的2D to 3D | VALSE2018
https://zhuanlan.zhihu.com/p/38611920
Are Large-Scale 3D Models Really Necessary for Accurate Visual Localization?
https://github.com/TerenceCYJ/VP-SLAM-SC-papers
3DMatch：从RGB-D重构中学习局部几何描述符
http://3dmatch.cs.princeton.edu/
https://github.com/weiweisun2018/awesome-point-clouds-registration
https://github.com/gwang-cv/Point-Set-Matching-Registration-Material
点云-特征匹配等论文集锦：
点云集锦：
https://github.com/Yochengliu/awesome-point-cloud-analysis
点云知识学习：
https://github.com/timzhang642/3D-Machine-Learning
其他知识点：
《视觉SLAM十四讲精品总结》7：VO—— 3D-2D：PnP+BA
https://blog.csdn.net/try_again_later/article/details/81813639
https://blog.csdn.net/try_again_later/article/category/7888541