orbslam作者的ppt

转载声明：本文转载自  金木炎 的博客，仅供个人学习。感谢博主的无私分享，如有侵权，敬请告知。

         Should we still do sparse-feature based SLAM？

1) Monocular ORB-SLAM(Mur-Artal, Montiel, Tardos T-RO 2015)

2) Monocular Semi Dense Mapping(Mur-Artal, Tardos, RSS 2015)

3) Stereo/RGB-D ORB-SLAM(unpublished)

Automatic Map Initialization

Model SelectionHomography(Planar, Low Parallax)Fundamental Matrix(General)

Why should we still use features?

Robustness

Reliable two-view monocular initialization

Good invariance to viewpoint and illumination

Less affected by auto-gain and auto-exposure

Less affected by dynamic elements

Accuracy

Bundle adjustment (joint map-trajectory optimization)

Place Recogniton (loop detection, relocalization)

Bags of Words

But sparse reconstructions ...