ORB-SLAM2是一个完整基于特征点的SLAM系统,包含了地图复用、闭环检测和重定位的功能,输入可以是单目、双目或者RGBD相机的图像。
SL-SLAM则是基于线段特征的双目视觉SLAM,可以构建线段特征的场景地图、进行闭环检测等等,是目前较为完善的基于线段特征的SLAM系统。
SVO
代码 : https://github.com/uzh-rpg/rpg_svo
文章: SVO: Fast Semi-Direct Monocular Visual Odometry
PL-StVO代码:https://github.com/rubengooj/StVO-PL
文章:Robust Stereo Visual Odometry through a Probabilistic Combination of Points and Line Segments
为基于双目相机的点线视觉里程计,采用两端点参数化空间直线,在优化时分别计算点、线特征的重投影误差,通过统计这两类特征的误差分布特性调整点线特征的权重,与本文不同的是,其前端采用暴力匹配的方法进行特征匹配。
PL-SVO
代码 : https://github.com/rubengooj/pl-svo
文章: PL-SVO: Semi-Direct Monocular Visual Odometry by Combining Points and Line Segments
笔记:
PL-SLAM,代码 : https://github.com/rubengooj/pl-slam
文章:PL-SLAM: a Stereo SLAM System through the Combination of Points and Line Segments
以下总结的是PL-StVO:
I. Introduction
II. System Overview
1) Point Features:
2) Line Segment Features:
3) Motion Estimation:
4) Uncertainty Propagation:
III. Combined Stereo Visual Odometry
A. Problem Statement
B. On-Manifold Optimization关于流形优化
C. Fast Outlier Rejection快速离群点剔除
IV. Uncertainty of the Error Functions
A. Detecting ill-Pose Configurations
V. Experimental Validation
A. Video Sequences.视频序列
1) Tsukuba dataset:
2) KITTI dataset:
B. Comparison in the KITTI Vision Benchmark
C. Processing Time
VI. Conclusions