Credit: Shivani Baldwa, Raghav Jethliya, under the guidance of Professor Chuang-Jan-Chang from Ming Chi University of Technology, New Taipei City
We used approach designed in MOIL-Lab , “MOIL SDK- MCUT Omnidirectional Imaging Lab” to extricate the fisheye picture into 6 diverse view as: Front, Left, Right respectively. By utilizing this technique, we made a trajectory of 3 view by the arrangement of fisheye pictures as a source document for odometry.
It means we use source fisheye images and integrate that with MOIL Library and generate 6-view odometry. Before we start this project, we have referred the AVI SINGH blogs and github https://github.com/avisingh599/mono-vo. So, this can be a referenced code before starting monocular visual odometry.
Watch this video for more understanding !!!! (Click on below image)
Calibration Parameter T265 CAMERA:
Focal Length: 618.8560
Center or Principle point: (427, 394)
MOIL LIBRARY PARAMETERS
md->Config("car", 3, 3,
427,394, 1, //center
848, 800, 1.68, // resolution
0, 0, -24.964, 38.2, -16.956, 183.42 //calibration
);
Download the MOIL dataset: https://bit.ly/3iq759W
If you want to know how to create dataset refer: https://github.com/Shivani1796/How-to-create-MOIL-Dataset
Step 1.Open Terminal
Step 2.
git clone https://github.com/Shivani1796/MonocularVisualOdometry-Using-3Views.git
MOIL SDK DEPENDENCIES
sudo apt update
sudo apt upgrade
sudo apt install build-essential cmake pkg-config
sudo apt install libjpeg-dev libpng-dev libtiff-dev
sudo apt install software-properties-common
sudo add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
sudo apt update
sudo apt install libjasper1 libjasper-dev
sudo apt update
sudo apt install libgtk-3-dev
sudo apt install libatlas-base-dev gfortran
sudo apt install libopencv-dev python-opencv
Step 3.Compile command
cd MonocularVisualOdometry-Using-3Views
mkdir build
cd build
cmake ..
make
Step 4.Run command
./vo /home/shivani/Documents/mydataset/D5_17-04/poses/D5.txt /home/shivani/Documents/fisheye-test-images/D5/%d.png 3.4 0,0 -10,0 10,0
DESCRIPTION:
Argument 1: Ground Truth Data (/home/shivani/Documents/mydataset/D5_17-04/poses/D5.txt)
Argument 2: Image sequence (/home/shivani/Documents/fisheye-test-images/D5/%d.png)
Argument 3: Zoom Factor (3.4)
Argument 4: Front View angle
Argument 5: Left View angle
Argument 6: Right View angle
NOTE
- Change img_path and pose_path to correct image sequences and pose file paths according to your system.
- Ensure focal length and principal point information is correct .
Please follow the below url to understand how to clone poject on Eclipse
---------------------------------------------------------------------------------------------------------------------------------------------------------------To use ORB matching for feature points switch branch- "ORB-Matching" or https://github.com/Shivani1796/MonocularVisualOdometry-Using-3Views/tree/ORB-Matching