Ask Your Question

printed circuit board component image recognition

asked 2016-03-02 10:33:51 -0500

cbrake gravatar image


I would like to use opencv to detect printed circuit board components in a camera image. What technologies/approaches would you advise? Thus far I've tried simple template matching with CV_TM_SQDIFF_NORMED. It works somewhat for small components (like a surface mount capacitor), but for larger components with more detail it does not very well, and overall is very sensitive to threshold settings, etc.

Thanks, Cliff

edit retag flag offensive close merge delete


Are you trying to identify them, or just detect them?

Tetragramm gravatar imageTetragramm ( 2016-03-02 19:18:29 -0500 )edit

need to identify components

cbrake gravatar imagecbrake ( 2016-03-02 21:13:49 -0500 )edit

2 answers

Sort by ยป oldest newest most voted

answered 2016-03-03 01:55:16 -0500

Pulkit gravatar image

You will have to use classification approach towards the object detection in PCB.Either you can use Haar Cascade but i believe here is gonna be too much false positive in haar cascade detection. Another option is you will have to use some image processing techniques(do a little literature survey) and after use some state of the art classifiers like random forest or svm or any other.In my opinion template matching might not be very robust. :)

edit flag offensive delete link more


thanks for the thoughts -- template matching is proving to work somewhat, but requires a lot of threshold adjustments from one problem to the next, so is no real practical for real work use.

cbrake gravatar imagecbrake ( 2016-03-03 12:20:14 -0500 )edit

answered 2016-03-02 21:38:08 -0500

Tetragramm gravatar image

Alright, there are several options. The usual ones are creating a Cascade Classifier or a Bag Of Words categorizer. There are good tutorials on each of these available.

I would also suggest, since it's quick to test, segmenting the component then using huMoments to compare to your dictionary.

edit flag offensive delete link more


thanks for the ideas!

cbrake gravatar imagecbrake ( 2016-03-03 12:19:07 -0500 )edit

Question Tools

1 follower


Asked: 2016-03-02 10:33:51 -0500

Seen: 1,357 times

Last updated: Mar 03 '16