Ask Your Question

Real Time detection of texture less objects?

asked 2013-07-30 17:59:24 -0500

yes123 gravatar image

updated 2013-07-31 16:52:51 -0500

Does OpenCV have some algorthms for the detection of texture less objects?

There are a number of paper on the argument:

But none of these seem to be implemented in OpenCV.
findContours and matchShape seem to be not sufficient for this task

edit retag flag offensive close merge delete


And to add a last comment, your remark about 'pretty bad documentation' can be countered. OpenCV is a open-source project that relies on the help of its community to devellop everything further. This means that if you have problems with documentation, you file a feature request on the dev forum and if you are really motivated, you add a pull request with the suggested adaptations.

StevenPuttemans gravatar imageStevenPuttemans ( 2013-07-31 02:34:44 -0500 )edit

3 answers

Sort by ยป oldest newest most voted

answered 2013-07-30 18:16:56 -0500

SR gravatar image

updated 2013-07-31 02:01:10 -0500

Search for "linemod" in the sources of OpenCV (an approach of Hinterstoisser).

See <opencv-src>/modules/objdetect/src/linemod.cpp

edit flag offensive delete link more

answered 2013-07-30 18:44:12 -0500

RaulPL gravatar image

Hi, I just find this video Real-time Learning and Detection of 3D Texture-less Objects and also the paper, it is quite recent and uses ROS.

edit flag offensive delete link more


Is there any documention in OpenCV about this? Otherwise it's useless

yes123 gravatar imageyes123 ( 2013-07-30 19:41:28 -0500 )edit

Since ROS does use OpenCV functionality inside, or very similar stuff, I think this is a very interesting link. Don't expect openCV to have ready made project solutions for you...

StevenPuttemans gravatar imageStevenPuttemans ( 2013-07-31 02:31:59 -0500 )edit

answered 2013-07-31 02:30:47 -0500

updated 2013-07-31 02:37:19 -0500

Personally I think that expecting OpenCV will contain a complete framework for finding textureless objects is rather rediculous. Looking at all the most recent computer vision conferences, still many PhD research goes into solving this exact problem. Basically what you do is implementing the algorithm suggested in a paper, by using openCV as a background.

This means that you will need to spit out every paper you are interested in, look for all subfunctionality and hope that it is implemented in OpenCV. Then use those functionality to combine them smartly into the desired algorithm.

The purpose of OpenCV as I see it is to provide computer vision building blocks, not complete ready served algorithms, because this would limit the options for devellopers. So remarks like, if it is not documented in OpenCV I can't use it, is a rather lame excuse. It just means that you are not prepared to expand existing functionality with new code added.


Looking for textureless objects can be done by using object categorization techniques, which do not depend on texture, like cascade classifiers, SVM algorithms, random forests, which are all available in OpenCV. They depend on shape features, rather then on texture features.

edit flag offensive delete link more
Login/Signup to Answer

Question Tools



Asked: 2013-07-30 17:59:24 -0500

Seen: 1,597 times

Last updated: Jul 31 '13