基于单目视觉的三维重建算法综述（补充）

作者：SIGAI特邀作者陈泰红
PDF地址：http://sigai.cn/paper_97.html

个人观点：
1、尽管SfM在计算机视觉取得显著成果并应用，但是大多数SfM和基于周围环境是静止这一假设，既相机是运动的，但是目标是静止的。当面对移动物体时，整体系统重建效果显著降低。
2、传统SfM基于目标为刚体的假设。
3、个人对3D重建算法不是深入，SfM也许没有vSLAM技术热点，但是多视觉几何和SfM是进入三维世界的大门，基础应用永不过时。
以上仅为个人阅读论文后的理解、总结和思考。观点难免偏差，望读者以怀疑批判态度阅读，欢迎交流指正。
参考文献

[1] Seitz S M, Szeliski R, Snavely N. Photo Tourism:Exploring Photo Collections in 3D[J]. Acm Transactions on Graphics, 2006, 25(3):835-846.
[2] Agarwal S,Snavely N, Simon I, et al. Building Rome in a day[J]. Communications of the Acm, 2011, 54(10):105-112.
[3] Toldo R,Gherardi R, Farenzena M, et al. Hierarchical structure-and-motion recovery from uncalibrated images[J]. Computer Vision & Image Understanding, 2015,140©:127-143.
[4] Chen Y,Chan A B, Lin Z, et al. Efficient tree-structured SfM by RANSAC generalized Procrustes analysis[J]. Computer Vision & Image Understanding, 2017,157©:179-189.
[5] H Cui , X Gao,S Shen, Z Hu. HSfM: Hybrid Structure-from-Motion[C]// Computer Vision and Pattern Recognition. IEEE, 2017.
[6] D. Eigen, C. Puhrsch, and R. Fergus. Depth map prediction from a single image using a multi-scale deep network. NIPS,2014.
[7] D. Eigen,R. Fergus.Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-Scale Convolutional Architecture.ICCV2015
[8] Iro Laina , Christian Rupprecht .Deeper Depth Prediction with Fully Convolutional Residual Networks.2016
[9] V.Sudheendra, Susanna Ricco .SfM-Net: Learning of Structure and Motion from Video.arxiv2017
[10] Christopher B. Choy, Danfei Xu, JunYoung Gwak.3D-R2N2: A Unified Approach for Single and Multi-view 3D Object Reconstruction.ECCV2016.
[11] Jonathan Masciy∗ Davide Boscainiy∗ Michael M. Bronstein.Geodesic convolutional neural networks on Riemannian manifolds.AAAI2018.
[12] Charles R. Qi, Hao Su, Kaichun Mo, Leonidas J. Guibas.PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation.CVPR2017.