tum-phoenix / drive_ros_camera_homography

camera homography estimation

Camera Homography Estimator

estimates homography between camera and calibration pattern plane and save transformation parameters to a .yaml file.

usage

0. put your camera in front of a camera calibration pattern or use recorded *.bag file of pattern
1. adapt topic names to your environment
2. launch homography_estimator.launch file
3. fire up rqt and open dynamic_reconfigure plugin
4. set correct pattern parameters
5. adjust estimated points for homography until your pattern fits nicely into the green box
6. adjust parameters of blob detector until all pattern marks are recognized
7. parameter file will be saved to the directory specified in homography_estimator.launch file

parameters

• pattern type (supported: chessboard, circles, circles_asymmetric)
• estimated points for homography p_{bot|top}_{right|left}_{x|y} (border of green lines in homography_estimator_input image)
• size of pattern (points_per_col and points_per_row)
• delta distance between pattern elements (pattern_delta_x and pattern_delta_y)
• offset of lower right point (pattern_offset_x and pattern_offset_y)
• some scale factor for outline, just use 50 (outline_scale_factor)
• blob filter settings

detecting hints

• make the estimated points slightly bigger than the real edges of the pattern (see homography_estimator_input example)
• the keypoints detected by the blob filter should be the same as the size of the pattern (no missing or extra keypoints), use the blob filter parameters to tweak until all pattern marks are recognized by the blob filter and all noise blobs are gone

windows

homography_estimator_input

shows the input image and the estimated points for homography (green square)

homography_estimator_keypoints

shows the warped image and the blob filter detection results

homography_estimator_pattern

shows the pattern detection results

homography_estimator_preview

shows the finally calculated homography applied on the orignal image

example output file

%YAML:1.0
---
world2cam: !!opencv-matrix
   rows: 3
   cols: 3
   dt: d
   data: [ 1.6692432475121174e+02, -1.0473530728192594e+02,
       4.1899035694525743e+01, -5.3058063431231153e+00,
       -8.5485048630232763e-01, 3.0488767398247290e+01,
       2.5714915733718796e-01, 1.9354996942966881e-03,
       2.1857104647575391e-02 ]
cam2world: !!opencv-matrix
   rows: 3
   cols: 3
   dt: d
   data: [ 9.2773372432628046e-05, -2.8302943486492732e-03,
       3.7701732480210497e+00, -9.5001192175387123e-03,
       8.5086756092260935e-03, 6.3424138245340718e+00,
       -2.5022146305447759e-04, 3.2544990711281058e-02,
       8.3393416152725042e-01 ]

credits

• mherb
• Phibedy
• fabolhak

Original version