Ask Your Question

calibration board tracking

asked 2017-08-15 14:38:37 -0500

HSV gravatar image

Hi all,

I know the x, y, z values of all the points on this calibration board very precisely. I want to calculate a homography between the found pixel locations of the large dots, and their corresponding x, y, values. Then I can map the image, so the dots are in nicely arranged rows and columns, so then I can determine what row/column each dot is on.

Does anyone have any tricks / tips on how to determine the correspondence between the found large dot pixel u,v location and their known x,y locations?

Thank you!

image description

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2017-08-15 22:56:45 -0500

Ziri gravatar image

updated 2017-08-15 23:02:58 -0500

I will answer your last question :

If you're able to detect all dots precisely then apply findHomography function to get transformation matrix .

Apply transformation matrix to convert your coordinates.

Hope it helps.

edit flag offensive delete link more


not all dots are detected in each image, like the example, there are 3 or 4 rows of dots at the top of the board that are not being detected.

HSV gravatar imageHSV ( 2017-08-17 16:32:58 -0500 )edit

You can achieve what you want using another board with less circles or squares (Chessboard) .

To get transformation matrix you need "world coordinates" and their correspondence in image.

Did you try findCirclesGrid function ?

Ziri gravatar imageZiri ( 2017-08-17 20:18:59 -0500 )edit

The large dots in the center are used to calculate the homography. What is good way to find only the large dots in the image and determine their correspondence to their known xyz values?

HSV gravatar imageHSV ( 2017-08-29 14:53:46 -0500 )edit

I think the easiest way is contour detection, then eliminate small circles using area or length threshold. you'll be able to find contours mass center. ( after detecting all large dots make sure they are sorted to matche their known values ).

Ziri gravatar imageZiri ( 2017-08-29 19:20:48 -0500 )edit

thank you.

HSV gravatar imageHSV ( 2017-08-30 09:37:02 -0500 )edit
Login/Signup to Answer

Question Tools

1 follower


Asked: 2017-08-15 14:38:37 -0500

Seen: 63 times

Last updated: Aug 15 '17