The landmarks are often used in SLAM. What are the algorithms used to extract them, and how can a robot differentiate the landmarks, if they detect one in point A at Xt and another in Xt+1? How can the robot know if it's the same landmark or not?

  • $\begingroup$ You need to narrow your topic. There are several representations for map which is strongly related to what it is called a landmark. $\endgroup$
    – CroCo
    Jul 3, 2016 at 16:42

1 Answer 1


Landmarks or features are usually found using SURf, SIFT, or ORB. these algorithms extract corners from images and can be found in the opencv library.

Matching measured features to previously seen features is called the assignment problem. There are a couple ways to deal with this. The munkres algorithm solves this problem in O(n^4) time with a nxn cost matrix. The cost of assigning a measured feature to a previously seen feature is often the difference between the two scaled by the mahalonobis distance .

  • $\begingroup$ The proper term for "Matching measured features to previously seen features" called data association. Also, the algorithms you stated related to visual SLAM not SLAM in general since Laser sensors more popular in SLAM expect underwater environments. $\endgroup$
    – CroCo
    Jul 3, 2016 at 16:39
  • $\begingroup$ is data association a problem with slam when lasers are being used? It seems like if you get distance and direction data association would not be needed. $\endgroup$
    – holmeski
    Jul 3, 2016 at 17:41
  • $\begingroup$ Thank you for your answer , its been a time that im intressted in robotics domaine and i am very confused , i start with Stereo Vision and it was successful I would like to continue in 3D reconstitution , can I use only odometry And INS The project the RGB-D image ? why use the SLAM ? $\endgroup$
    – namster
    Jul 3, 2016 at 18:22
  • $\begingroup$ @namster, it might be time for another question. But i think the answer to your question is yes. Using only a single camera and an INS you will have the ability to do slam. $\endgroup$
    – holmeski
    Jul 3, 2016 at 19:10
  • $\begingroup$ @holmeski, I've read alot of papers when I was doing my project regarding SLAM and majority of these papers (i.e. most of them using lasers but some use cameras) using data association. $\endgroup$
    – CroCo
    Jul 9, 2016 at 4:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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