how to improve detectCheckerboardPoints performance

8 views (last 30 days)
tic
[app.imgPts,~] = detectCheckerboardPoints(img);
toc
CommandWindow:
Elapsed time is 2.233638 seconds.
Elapsed time is 2.142986 seconds.
Elapsed time is 2.198445 seconds.
Elapsed time is 2.124059 seconds.
I want to at least halve the detection time so the camera stream is faster overall.
Is there a way to calculate this as a parallel process or through the GPU?
That is exactly what must be logically possible, but I can not see any approach.
Used hardware: Intel HD Graphics 5500, Intel Core i5-5300 U
I have tested it on another PC ( Geforce GTX1070,Intel Core i5) and its much faster
but for the later use of the script I can not ensure these hardware requirements.

Answers (1)

Sai Bhargav Avula
Sai Bhargav Avula on 21 Oct 2019
Hi,
Currently the detectCheckerPoints doesn’t support both parallel computing and GPU computing.
As detectCheckerboardPoints is meant to only detect the cross corners a possible workaround woiuld be increasing the contrast of your checkboard image. This might help in improving both performance and accuracy.
Hope this helps.
  1 Comment
Bradley Crouch
Bradley Crouch on 12 May 2021
Query, how accurate is cornerPoints (as this is abpe to use GPU)
If points could be arranged in grid as detectCheckerboardPoints, then this would be of benifit.
Currently we are using detectCheckerPoints (for camera calibration), but the performance is very slow (4000x3000 image)..
Also detectCheckerboardPoints, rotating MxN to match black square is somewhat annoying (as checkerboard in our case is always in one direction)

Sign in to comment.

Categories

Find more on MATLAB Support Package for USB Webcams in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!