Rolling Shutter Camera calibration

The rolling shutter calibration tool provides full intrinsic calibration (projection, distortion and shutter parameters) of rolling shutter cameras [1]. The codebase currently only supports a single camera, and cannot calibrate an IMU alongside it.

How to use?

1) Collect rosbag / images

Create a ROS bag containing the raw image data either by directly recording a rosbag from a ROS sensor stream or by using the bagcreater script on a sequence of image files.

The camera system is fixed and the calibration target is moved in front of the cameras to obtain the calibration images.

2) Running the calibration

• --model pinhole-equi-rs
  he camera model to estimate. Currently supported models are pinhole-radtan-rs, pinhole-equi-rs, omni-radtan-rs
  
• --inverse-feature-variance 1
  Estimated inverse variance of the feature detector.
  
• --frame-rate [FRAMERATE]
  Approximate framerate of the camera.
  

The calibration can be run using:

rosrun kalibr kalibr_calibrate_rs_cameras 
    --bag [filename.bag] \
    --model [MODEL_0 ... MODEL_N] \
    --target [target.yaml] \
    --topic  [TOPIC_0 ... TOPIC_N]  \
    --inverse-feature-variance 1 \
    --frame-rate [FRAMERATE]

Example command of sample bag file (download it here):

rosrun kalibr kalibr_calibrate_rs_cameras \
	--model pinhole-radtan-rs \
	--target april_6x6_80x80cm.yaml \
	--topic /img_pub/camera/image/left \
	--bag ELP_rs_stereo.bag \
	--inverse-feature-variance 1 \
	--max-iter 10 \
	--frame-rate 30

3) The output

The calibration will be printed to the screen.

References

Please cite the appropriate papers when using this toolbox or parts of it in an academic publication.

1. L. Oth, P. Furgale, L. Kneip, R. Siegwart (2013). Rolling Shutter Camera Calibration, In Proc. of the IEEE Computer Vision and Pattern Recognition (CVPR)