|
object recognition with Kinect
David Chik from Japan [31 posts] |
11 year
|
Kinect provides both depth information and rgb information.
(1) What is the best module to analyze depth information?
(2) Is it possible to provide a new Object Recognition module that utilizes both feature (from rgb) and something from depth?
(3) Is it possible to provide a new module that builds a 3D point cloud model and uses a 3D shape matching?
|
|
|
|
Steven Gentner from United States [1446 posts] |
11 year
|
1) Tough to say which method works best to analyze depth. Depends on what you want to do. Thresholding is a good way of removing background objects from one that may be in the foreground. Object Recognizition using the Haar technique would probably work well to identify valleys and peaks that may comprise an object.
2) That is possible. What kind of object are you thinking of needing to recognize using both depth and RGB?
3) Yes, but those results even with the Kinect are still noisy but do look cool. Very useful in generating localization maps, etc. Again, depends on what type of objects you plan to recognize using these enhanced techniques.
STeven.
|
|
This Thread