Ask Your Question

Keypoints, decriptors and matches research

asked 2013-10-16 01:37:18 -0500

Mariusz gravatar image


I have an Android project, where I have about 50 logos and I need to recognize one of them on the scene(a logo will take about 1/4 of the scene space).

I have no experience with computer vision, I went through all the tutorials about feature detection in OpenCV site, but still - I have like no idea what I'm doing.

As I assume, I have to figure out, which parameters(and classes) are the best for feature detection, decription and matching for my particular case, but how I should do the research?

If you cound give me any tips, where I could start, which steps I should take, I would be very grateful

Thanks in advance

edit retag flag offensive close merge delete



1 answer

Sort by ยป oldest newest most voted

answered 2013-10-16 05:15:56 -0500

Before digging into complex feature detection methods you should try simpler approaches.

Do the logos super-impose each other? Do they appear in different scales and rotations? Is it easy to remove background and leave only the logo in the image?

First check if the logo you want to find has any particular features that most do not, such as shape, color, and others. For example, if your object is round, you should be able to eliminate all non-round objects by a simple blob analysis approach.

If you have to detect the object in similar background scenes, you could use a trick based on color. Every logo probably has a unique signature that is determined by the mean of its RGB pixel values. Just check which is that value for your logo and then you'll know how to find it.

edit flag offensive delete link more

Question Tools


Asked: 2013-10-16 01:37:18 -0500

Seen: 254 times

Last updated: Oct 16 '13