eWand

This is the code and the build instructions for the paper eWand: A calibration framework for wide baseline frame-based and event-based camera systems

Citations

======= This paper contains the code for the paper eWand: A calibration framework for wide baseline frame-based and event-based camera systems.

Citing

If you use this code in an academic context, please cite the following publication:

Paper: eWand: A calibration framework for wide baseline frame-based and event-based camera systems

Video: YouTube

Project: See here

@inproceedings{Gossard2024icra,
  url = {https://arxiv.org/pdf/2309.12685.pdf},
  title={eWand: A calibration framework for wide baseline frame-based and event-based camera systems},
  author={Gossard, Thomas and Ziegler, Andreas and Kolmar, Levin and Tebbe, Jonas and Zell, Andreas},
  year={2024},
  booktitle = {2024 {International} {Conference} on {Robotics} and {Automation} ({ICRA})},
  publisher = {IEEE},
}

Using the code

The code of this project is split into several packages:

• Marker extrection from frame-based cameras
• Marker extrection from event-based cameras
• A calibration framework consuming the extracted marker positions and optimizing the cameras' intrinsics and extrinsics

Marker extraction from frame-based cameras

You can use the python script "script/extract_ball_frames.py". You just need to change the path of the directory were all your frames are saved in it.

Calibration framework

Software requirements

This code has been tested on Ubuntu 20.04.

Dependencies:

• OpenCV
• Ceres

How to build

1. Clone this repositories:

git clone https://github.com/cogsys-tuebingen/ewand

4. Build:

cd calibration_framework
mkdir build
cd build
cmake ..
make

5. Setup the intrinsic camera parameters

For every camera you want to calibrate, create a .yaml file with the following structure. You can get the intrinsic parameters from the intrinsic calibration tool of your choice, e.g. kalibr, OpenCV.

%YAML:1.0
camera_name: cam0
image_width: 1280
image_height: 1024
camera_matrix: !!opencv-matrix
  rows: 3
  cols: 3
  dt: d
  data: [1277.0348869354, 0, 637.71636909691, 0, 1276.78458195841, 488.458326168923, 0, 0, 1]
distortion_model: plumb_bob
distortion_coefficients: !!opencv-matrix
  rows: 1
  cols: 5
  dt: d
  data: [-0.220061443243291, 0.261537432483716, -0.0025783399940279, 0.00000095452629096082, 0]

6. Run:

cd buid
./src/wand_calibration_main \
  --cam1_config ../intrinsics/camera_0.yaml \
  --cam1_data ../data/camera_0.csv \
  --cam2_config ../intrinsics/camera_1.yaml \
  --cam2_data ../data/camera_1.csv \
  --cam3_config ../intrinsics/camera_2.yaml \
  --cam3_data ../data/camera_2.csv \
  --cam4_config ../intrinsics/camera_3.yaml \
  --cam4_data ../data/camera_3.csv \
  --cam5_config ../intrinsics/camera_4.yaml \
  --cam5_data ../data/camera_4.csv \
  --cam6_config ../intrinsics/camera_5.yaml \
  --cam6_data ../data/camera_5.csv \
  --wc_distance_scaling 1000 --wc_linearity_scaling 100000

where the yaml files contain the intrinsics of the cameras and the csv files the extracted markers from the frame- and event-based cameras.

Note: Use --help to see all the options.

License

This software is issued under the Apache License Version 2.0.