VINS-Stereo：基于VINS-Mono的双目VIO系统开发

Update on 19/11/2022

1. rewrite feature_manager.cpp and change trianglute() which use left and right to triangulate.
2. rewrite initial_sfm.cpp: (1)add triangulateLeftAndRight() which use left and right to triangulate. (2)use triangulateLeftAndRight() in construct(). (3)rewrite window BA in construct().

In VIO with two cameras, the input information of vision doubled. So it is easier to recover 3D coordinates of features. But it is more computational consuming in frontend tracking.

Earlier Info

这个系统是基于香港科技大学飞行机器人组的开源框架VINS-Mono开发的，原开源框架是针对单目SLAM。本双目SLAM系统是在原开源框架基础上的二次深度开发，外部接口与原框架一致。您可以依此对比原开源框架VINS-Mono或VINS-Fusion与本系统的区别。非商业用途，感谢HKUST的沈老师课题组提供的开源框架。

1.需要的基本配置

Ubuntu 16.04
ROS Kinetic
Ceres Solver
OpenCV 3.3.1
Eigen 3.3.3

2.在您的ROS上构造VINS-Dual

cd ~/catkin_ws/src
git clone https://github.com/Cai-RS/My_VINS-Stereo.git
cd ../
catkin_make

3. VINS-Dual的启动

请在EuRoC MAV Dataset下载您需要测试的数据集。

打开1个控制台，输入：

roscore

进入devel目录，分别打开2个控制台，输入：

source setup.bash

再分别输入：

roslaunch vins_estimator euroc.launch 
roslaunch vins_estimator vins_rviz.launch

如果不使用回环检测，请输入：

roslaunch vins_estimator euroc_no_posegraph.launch 

再打开1个控制台，请输入：

rosbag play YOUR_PATH_TO_DATASET/MH_01_easy.bag 

4.参考文献

T.Qin, P.L.Li, S.J.Shen, VINS-Mono: A Robust and Versatile Monocular Visual-Inertial State Estimator.