Map detection in the scene using OpenCV

Question

Can someone direct me to some information on how to do the following:

I have a database of about 10,000 trading cards. Given the image from iPhone / Android / WebCam, I would like to find the nearest suitable trading card. I was looking for basic Haar-based image recognition, but this requires training with many images with common content. My program would have 1 perfect image for every possible match.

Now the input will not be scanned, so the routines will need to handle any type of scaling / rotation and cropping. But the whole image should be quite noticeable.

Any ideas or thoughts? This is my first foray into the world of OpenCV, and it is interesting to say the least.

Answer 1

How about SURF for detecting similarities in images?

Answer 2

For everyone who comes to this post, there are two broad approaches.

The first is a function mapping (as mentioned above). The general idea is that the unique points correspond to each other. OpenCV provides many ways to extract and map functions . See here . This approach is good for identifying the desired object in various conditions (affine transformations, lighting conditions, scale).

-, . . OpenCV . , , , , .