Stereo_Odometry

---

Introduction

$\qquad$ this program complete a Stereo Vision odometry composed of Sparse feature based frontend and a Keyframe based sliding window backend,reference the ORB_SLAM2, PTAM and Visual SLAM 14 lectures

---

Reference paper

ORB-SLAM2: an Open-Source SLAM System for Monocular, Stereo and RGB-D Cameras

ORB-SLAM: a Versatile and Accurate Monocular SLAM System

Parallel Tracking and Mapping for Small AR Workspaces

Improving the Agility of Keyframe-Based SLAM

---

Dependency

• Eigen used for matrix calculate
• OpenCV used for feature extract and Optical flow tracking of feature points
• G2O used for nonlinear optimization in frontend and backend
• Sophus
• Glog used for log record
• Gflag used to define a global config param
• CSparse
• Gtest

---

Dataset

$\qquad$ program use Kitti gray img 00/01 dataset to test, change the dataset_dir in config/default.yaml file to your local dataset path. Befor compile the program, change the config_file in app/run_kitti_example.cpp to you local config yaml file path

---

Build program

$\qquad$ In project dir, execute following steps:

$\qquad$ mkdir build && cd build

$\qquad$ cmake ../

$\qquad$ make

$\qquad$ cd bin && ./run_kitti_example

a traj_result.txt file will generate after program run finished, use the plot_traj in bin dir to plot the traj

---

Test result

the following img may not show, the traj result is store in traj_result directory