-1

I need to calibrate a camera according to soccer field white lines. In order to do so I used Canny for edge detection and HoughLinesP to get the white lines vectors. The location of the camera is not fixed and the picture may contain also the crowed. In this case the crowed may be very noisy for HoughLinesP, thus i thought to extract the ROI of field from the image. Iv'e converted the image to HSV and used inRange on green color. Now, what is the best way to get the ROI?

Link to example for a noisy image - Source and after InRange

user1171777
  • 11
  • 2

1 Answers1

0

Well in this approach using a clustering method like DBSCAN you can get rid of the noise and select biggest region. Then you can use opencv to describe that region, with a contour, with lines or with a convex hull.

By the way, you may want to take a look at recent line detectors

We use field, line, circle informations together with some problem specific assumptions, also lens undistortion and model fitting for auto calibration of static cameras; however there will always be a noisy case that require minimal user input

baci
  • 2,528
  • 15
  • 28