LISS

Simulator Of 3D Landmarks And Image Sequences Based On Line Features

17 Jun 2020: This is the 1st Edition, for Chinese version please click Introduction.

This is a generator tool used for 3D landmarks building based on line features which can be used to test VO/SLAM system. It can be easily changed into VIO-SLAM pattern.

Authors: iwander。

Videos: There is a introduction video to show how to use this tool.

1. Prerequisites

python

python 2 or python 3.
numpy.
pyplot.
openCV 3.

2. Build LISS on PC

2.1Clone the repository and run:

    cd ~/home/
    git clone https://github.com/iwander-all/LISS-SimulatorOf3DLandmarksAndImageSequences.git
    cd script

2.2 Format Please build folders in the way:

Simulator
|--support_files
|  |--EuRoC_traj_data
|  |  |--XX_0X_XXXX.tum         ->txt data of trajactory of EuRoC (TUM style)
|  |--txt_data
|  |  |--trajactory.txt         ->simulated trajactory and rotation data of robot
|  |  |--line.txt               ->simulated 3D coordinates of lines
|  |--series_data
|  |  |--images                 ->camera images simulated at different time and poses
|  |  |  |--difficult
|  |  |  |--medium
|  |  |  |--pure
|  |  |--landmarksGroundtruth   ->pixel coordinates of landmarks at different time and poses
|  |  |  |--cubes
|  |  |  |--lines
|  |  |  |--stars
|  |  |  |--triangles
|
|--script
|  |--traj_visual.py            ->viualize the raw trajactory data
|  |--trajactoryGenerator.py    ->simulate robot poses (20Hz)
|  |--lineGenerator.py          ->generate lines in 3D, before it runs, please move the previous data or be replaced.
|  |--utils.py                  ->all the supporting data and functions
|  |--imageGenerator.py         ->generate simulated camera vision (.png) and groundtruth (.txt)
|  |--visualize.py              ->visualize the trajactory and landmarks

3. Run with your device

3.1 Visually check the trajectories

    python traj_visual.py

3.2 Build trajectory

    python trajactoryGenerator.py

3.3 Build 3D landmarks

    python lineGenerator.py

The landmarks are randomly built, so if you want to check the built one, please use:

    python visualize.py

3.4 Build images

    python imageGenerator.py

4. Acknowledgements

I use EuRoC for trajectory and poses and the parameters for camera model.

5. Licence

The source code is released under Apache License 2.0 license.

6. Further Work

I am still working on improving the code reliability and further work for VIO-SLAM pattern is expected. For any technical issues, please contact me https://blog.csdn.net/iwanderu/ .