Skip to content

Latest commit

 

History

History
69 lines (61 loc) · 3.5 KB

README.md

File metadata and controls

69 lines (61 loc) · 3.5 KB
Raw

surround-camera_calib

surround-camera_calib is a calibration toolbox for surround view cameras or surround view fisheye cameras, which contains four tools, as shown in the table below. For more calibration codes, please refer to the link SensorsCalibration

calibration param calibration type calibration method mannual calibration auto calibration usage documentation
surround_cameras (fisheye) extrinsic target-less ✔ manual_calib
surround_cameras (fisheye) extrinsic target-less ✔ auto_calib_fisheye
surround_cameras extrinsic target-less ✔ auto_calib
surround_cameras extrinsic target ✔ auto_calib_target

Prerequisites

  • Cmake
  • opencv 3.4.5
  • eigen 3

Compile

# mkdir build
mkdir -p build && cd build
# build
cmake .. && make

Run(our examples)

cd bin
#choose 1.Calibrate by fixing the front
./run_AVM_Calibration_F
#choose 2.Calibrate by fixing the back
./run_AVM_Calibration_B

Customize(your examples)

If you need to calibrate base on your data, you can follow as below steps:

Prepare your surround camera images and the extrinsics and the intrinsic parameters etc.(set in /src/optimizer.cpp)

  • set the extrinsics in class function---Optimizer::initializePose()
  • set the intrinsics in class function---Optimizer::initializeK()
  • set the fisheye distortion parameters in class function---Optimizer::initializeD() Ps:if pinhole camera, set 0
  • set the BEV camera intrinsic and height(or you keep same as ours) in class function---Optimizer::initializeKG() and Optimizer::initializeHeight()
  • set front,left,right,back BEV image tail size in class function---Optimizer::initializetailsize()

Set calibration model(set in /src/calibration_fixedF.cpp or /src/calibration_fixedB.cpp)

  • choose camera model(camera_model)：0-fisheye;1-Ocam;2-pinhole
  • if add extra disturbance on surround cameras(flag_add_disturbance)：1-add;0-not
  • choose phase solution model(solution_model_):
    • 1.pure gray pipeline in three phase of optimization: solution_model_="gray"
    • 2.(default)Adpative Threshold Binarization in first phase and pure gray in the 2nd&3rd phase of optimization:
      solution_model_="gray+atb"
    • 3.pure Adpative Threshold Binarization in all three phase of optimization: solution_model_="atb"
  • (Optional) if add road semantic segmentation when in texture extraction process to improve accuracy(add_semantic_segmentation_front/left/right/back):1-add 0-not
    • Ps: if you want to add road semantic segmentation mask you need to provide road semantic segmentation mask.
  • choose camera fixed(fixed):"front" or "back"

Citation

If you find this project useful in your research, please consider cite:

@misc{2305.16840,
Author = {Jixiang Li and Jiahao Pi and Guohang Yan and Yikang Li},
Title = {Automatic Surround Camera Calibration Method in Road Scene for Self-driving Car},
Year = {2023},
Eprint = {arXiv:2305.16840},
}

Contact

If you have questions about this repo, please contact Yan Guohang ([email protected]).