OpenCV: cvFindChessboardCorners() 함수

OpenCV: cvFindChessboardCorners or cv::findChessboardCorners

source code: https://code.ros.org/trac/opencv/browser/tags/2.1/opencv/src/cv/cvcalibinit.cpp
file:///opencv/src/cv/cvcalibinit.cpp

ref.
Learning OpenCV: Chapter 11 Camera Models and Calibration

V.Vezhnevets, A.Velizhev (Graphics and Media Lab, CMC department, Moscow State University) "GML C++ Camera Calibration Toolbox", 2005
http://graphics.cs.msu.ru/en/science/research/calibration/cpp


http://www.vision.caltech.edu/bouguetj/calib_doc/htmls/example.html

IplImage* cvGetImage(const CvArr* arr, IplImage* imageHeader)

cvCheckChessboard
source file: //opencv/src/cv/cvcheckchessboard.cpp

Dilate or cv::dilate

Flood-fill algorithm
http://en.wikipedia.org/wiki/Flood_fill


quad_count = icvGenerateQuads( &quads, &corners, storage, thresh_img, flags );

CvMemStorage* cvCreateChildMemStorage(CvMemStorage* parent)

StartFindContours

cvContourArea

---

routine:
"cvcalibinit.cpp": cvFindChessboardCorners()'s routine

1. gray-scale input image (option: CV_CALIB_CB_NORMALIZE_IMAGE | CV_CALIB_CB_NORMALIZE_IMAGE ) -> save "imgGray"

2. binarize the gray-scaled image

adative thresholding (option: CV_CALIB_CB_ADAPTIVE_THRESH) or thresholding by "image mean-10" or "10"




3. dilate the binarized image -> save "imgThresh"

4. find rectangles to draw white lines around the image edge -> save "imgRect"

 

1. check if a chessboard is in the input image

 

1) erode and dilate ( cvErode() & cvDilate() )

2) find a threshold value to make contours ( "flag" )
3) select contours to make quadrangles
4) check if there are many hypotheses with similar sizes ( floodfill

-style algorithm )

 

2. (if pattern was no found using binarization) multi-level quads

extraction

3. draw white lines around the image edges ( "thresh_image" )

4. compute corners in clockwise order

5. find the quadrangle's neighbors

6. find connected quadrangles to order them and the corners

7. remove extra quadrangles to make a nice square pattern

8. check if each row and column of the chessboard is monotonous

9. refine corner locations

1. check if a chessboard is in the input image

1) erode and dilate ( cvErode() & cvDilate() )
2) find a threshold value to make contours ( "flag" )
3) select contours to make quadrangles
4) check if there are many hypotheses with similar sizes ( floodfill-style algorithm )

2. (if pattern was no found using binarization) multi-level quads extraction
3. draw white lines around the image edges ( "thresh_image" )
4. compute corners in clockwise order
5. find the quadrangle's neighbors
6. find connected quadrangles to order them and the corners
7. remove extra quadrangles to make a nice square pattern
8. check if each row and column of the chessboard is monotonous
9. refine corner locations