1
$\begingroup$

I've been reading how augmented reality works and how markers are detected. Anyway I can't find any resource that would help me in detecting an arbitrary complex image as, for instance, a book's cover.

I don't think that falls into the "template matching" field since I would also need its orientation to perform AR rendering on it.

What algorithms are used to detect an arbitrarily complex marker into an image and have its transformation extracted as well?

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ you're asking wildly for a really really huge topic – object recognition. I think it'd be good if you could list what you've been already researching and reading, and what your current level of experience in image processing is. As you can imagine, there's a whole world of different approaches to this, and it might be hard to answer your question without just pointing you to a bookshelf row in some library if potential answerers can't narrow things down. $\endgroup$
    – Marcus Müller
    Mar 27, 2017 at 8:25

2 Answers 2

Reset to default
0
$\begingroup$

In essence, you are looking for a way to do planar object detection.

Well, the most basic approach is to find a correspondence between your planar template and the image using a feature extractor, such as SIFT. You could then cast the detection problem as finding the subset of points, which have similar descriptors. OpenCV has that implemented already: http://opencv-python-tutroals.readthedocs.io/en/latest/py_tutorials/py_feature2d/py_feature_homography/py_feature_homography.html

http://docs.opencv.org/trunk/dc/dc3/tutorial_py_matcher.html

The next thing is to use template matching kind of approaches. For severe viewpoint this might be hard to achieve, but there are ways, such as this one.

If you can also design a marker in the image space, then there are approaches which let you to read a randomly generated marker. For example:

T. Birdal, I. Dobryden, S. Ilic X-Tag: A Fiducial Tag for Flexible and Accurate Bundle Adjustment International Conference on 3DVision (3DV), Stanford University, California, USA, October 2016 http://campar.in.tum.de/pub/tbirdal20163dv/tbirdal20163dv.pdf

Take a look at the related work section of the references I gave here. This could broaden your knowledge in the field. You might also like to read a survey in the field, such as: http://www.irisa.fr/lagadic/pdf/2012_fcv_uchiyama.pdf

Improve this answer
$\endgroup$
0
$\begingroup$

Did you hear about shape contexts descriptor where you use the contour of object and find the relative distribution of points in the plane relative to each point on the shape? There is also a Matlab demo in the link below.

https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/shape/sc_digits.html

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.