Automatic Surround Camera Calibration

Calibration Pipeline

The background of the project is that the position of the surround-view camera has shifted due to prolonged operation or subsequent maintenance work, and needs to be recalibrated. It needs to be re-calibrated, and the external reference of the camera needs to be corrected on-line during road driving. Specific responsibilities are listed below:

  • Realization of image preprocessing, keypoint selection based on pixel gradient magnitude and adaptive ROI selection;
  • Construct an optimization problem for the photometric residuals in the external parameters of the ring-view camera using the consistency of the photometry of the pixels in the projected common-view area of the ring-view camera. The problem was wrapped using the Ceres optimization library, and the algorithm was iteratively optimized using the L-M optimizer.
Hekui Guo
Hekui Guo
Master in Robotics