Camera Calibration Software v1.

0
Copyright 2004

Engin Tola
<engin.tola{ at }epfl.ch> June 12, 2006 -This piece of program contained in Camera Calibration Software v1.0 can be used, copied, modified, merged, published, and/or have copies distributed for academic or research purposes only without restriction under the following conditions: 1. The above header and this permission notice shall be included in all copies or substantial portions of the program 2. The software is provided "as is", without warranty of any kind, express or implied, including but not limited to the warranties of merchantability, fitness for a particular purpose and noninfringement. In no event shall the author(s) be liable for any claim, damages or liability, whether in an action of contract, tort or otherwise, arising from, out of or in connection with Camera Calibration Software v1.0 or the use or other dealings in Camera Calibration Software v1.0 3. If you use this piece of code for research purposes, refer to Tola, Engin. 2006 June 12. Homepage. <http://cvlab.epfl.ch/~tola/ index.htm> 4. An acknowledgement note should be included as: "The software used here was originally created by Tola, Engin. 2006 June 12. Homepage. <http://cvlab.epfl.ch/~tola/ index.htm> **************************************************************************************

You need a calibration pattern to calibrate the camera. You can find a sample calibration pattern in the /Grid folder. Paste this image to a word file and print. There must be some white area around the pattern for the algorithm to work. The calibration images have to be taken in an upward fashion i.e, #of colums < #of rows

Algorithm Information: This program uses the OpenCVs cvFindChessBoardCornerGuesses function in order to find the location of the corners in the image and cvCalibrateCamera_64d function to find the A matrice of the camera. cvFindExtrinsicCameraParams_64d function is used to find R and T matrices. There are some bugs in the OpenCV implementation and I tried to correct or in a way avoid them. For example, cvFindChessBoardCornerGuesses function sometimes cannot find the time. all of the corners or returns them in a un-ordered way. To avoid wrong calculations you should check the Repojection Errors. However, algorithm is robust %95 of

This program is written in order to compare the automatic calibration results with manual calibration. - Program tips: - application example: Load at least 4 images and press Calibrate button. By pressing "..." button load the images you want. There must be minimum 4 images. there is no maximum limitation. For the algorithm to work correctly, a high contrast is needed between the white and black squares. "Column number" is the number of horizontal square corners and Row Number is the number of vertical square corners. If you are using the supplied grid do not change the preset values. Apply Ordering orders the resultant pattern. In order to do this Column Number <= Row Number-2 is necessary. ( If you use the given pattern there should not be any problem. ) However, code is not fuly debugged and therefore for some other pattern

this may fail. To test if ordering is succesful or not you should enable Display Found Corners option. If there is a problem you may disable this option. Compansate Distortion makes use of the calculated Distortion Parameters in the error calculation stage. Dafault option is enabled. Display Found Corners displays the corners found on the patterns. I recommend you to enable this option for the first times of your inspection. Print Rotation and Translation Matrices prints the estimated R and T matrices. View Results prints the found parameters. View Log File is the file the results are written. View Coordinates prints the ground truth corrdinates, estimated coordinates and the error between them. Finally, please send any comments or bug information to engin.tola { at } epfl.ch along with the input state of the variables ( input images, algorithm parameters... )

I hope you will find this program useful.

engintola@2006