Just starting out: matching terminology with desired goal

asked 2015-09-06 15:24:25 -0500

WesS gravatar image

Hi folks,

I'm just starting out with OpenCV (and image recognition/computer vision in general). My goal right now is simply to hold up an image to my webcam and determine which image I'm holding up based on a database of images on my computer.

My problem is I'm not sure how to rephrase my goal into terminology that matches OpenCV's feature set. Is this an object recognition problem? Thresholding? Template matching? Any hints as to what libraries I should be focusing on?


edit retag flag offensive close merge delete


May be you should try examples given in opencv .

LBerger gravatar imageLBerger ( 2015-09-07 02:04:51 -0500 )edit

well, template matching might solve it, but:

"simply to hold up an image to my webcam" - this unfortunately make it anything but simple, because will introduce perspective distortion and scaling, which you have to remove before applying template matching, so a lot of preprocessing nessecary. (find your image in frame from webcam, get the outline, warp it back to straight horizontal, scale to original template size)

berak gravatar imageberak ( 2015-09-08 01:09:12 -0500 )edit

Thanks berak. That at least tells me what areas to start fiddling around with.

WesS gravatar imageWesS ( 2015-09-08 09:53:43 -0500 )edit

imho, you should start with template matching with 2 images (simple), then try to hold it into your cam, so it's about same size and orientation as your template, then improve. just do it step by step, and see, that you've always something to look at.

berak gravatar imageberak ( 2015-09-08 10:17:43 -0500 )edit