Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Moving Objects #81

Open
colek42 opened this issue Mar 20, 2015 · 4 comments
Open

Moving Objects #81

colek42 opened this issue Mar 20, 2015 · 4 comments

Comments

@colek42
Copy link

colek42 commented Mar 20, 2015

I am trying to localize moving objects. Is is possible with LSD slam, given a window, to extract depth information for objects that are moving from frame to frame, i.e. a walking person? I will be matching feature vectors in R^2 (HOG+SVM) and attempting to localize them in R^3. I noticed that the moving objects turn black in the visulizer. Much appreciated if you could point me in the right direction.

@amiltonwong
Copy link

Hi, colek42,
Could you present why it has significance to localize the moving objects? Or you want to do a dynamic reconstruction of the moving objects?
For localize moving objects, existing approaches like object tracking may be worth considering.

@colek42
Copy link
Author

colek42 commented Mar 20, 2015

I do not need to do reconstruction of the moving object. Right now our research is focused on tracking moving objects using low cost MAVs. Use case will be avoidance/following. Reconstruction using monocular slam is a difficult problem.

@amiltonwong
Copy link

So for your scenario, object tracking approach may meet your needs

@colek42
Copy link
Author

colek42 commented Mar 20, 2015

Matching SURF/SIFT features on humans does not work well. It looks like the depth information make be marked as invalid in somwhere in DepthEstimation/DepthMap.cpp.

Everytime I get a new key frame I am going to search for matching feature vectors (humans) in R^2. If I have a match I want to mark the 'box', and and segment the points in R^3 that reside in that box, in a seperate data structure. Right now it looks like those points get discarded. I want to use them, I have other aproaches as well if you don't think this will work well, we do want to use LSD SLAM for other reasons. I apologize if this does not make sense, I am still at the hypothesis stage.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants