cumtchenchang / visual_odom

This repository is C++ OpenCV implementation of SOFT (Stereo Odometry based on careful Feature selection and Tracking)

Impression

1. No BA optimizatin in this project
2. Buckect and circle match is a good idea. Bucket can be seen as a type of features extraction such as image segmentated into blocks.
3. function distinguishNewPoints() will drag down the real-time nature of the system
   1. add new features, and check every feature is already exist.
4. Not use keyframe strategy

Stereo Visual Odometry

This repository is C++ OpenCV implementation of SOFT (Stereo Odometry based on careful Feature selection and Tracking)

Original Paper: https://lamor.fer.hr/images/50020776/Cvisic2017.pdf

Demo vedio: https://www.youtube.com/watch?v=Z3S5J_BHQVw&t=17s

![image](https://github.com/cumtchenchang/visual_odom/row/master/images/euroc_trajectory.png）

Requirements

OpenCV 3.0

Eigen 3.34

Dataset

Tested on KITTI odometry dataset

Compile & Run

git clone https://github.com/ZhenghaoFei/visual_odom.git

Change EIGENPATH in visual_odometry/CMakeLists.txt

The system use Camera Parameters in calibration/xx.yaml, put your own camera parameters in the same format and pass the path when you run.

mkdir build
cd build
cmake ..
make -j4
./run /PathtoKITTI/sequences/00/ ../calibration/kitti00.yaml

Reference code

1. Monocular visual odometry algorithm

2. Matlab implementation of SOFT