How to use kinect fusion to map a moving object?

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

How to use kinect fusion to map a moving object?

anson2004110
This post has NOT been accepted by the mailing list yet.
I want to capture the moving object and calculate its pose using Kinect fusion. However, Kinect fusion only use for a static scene, I copy the description from the fusion website as:

Conceptually, a moving sensor with a static scene (e.g. environment scanning), and a static sensor with a moving scene (e.g. object scanning) are equivalent. Kinect Fusion supports both, but in general, the moving sensor with static scene scenario is much more robust as there is a larger amount of varied data for camera tracking.
For object scanning with a static sensor, make sure the object has enough depth variation and covers a large enough area of the Kinect image without getting too close and having depth return 0 (this typically limits the minimum size of objects you can reconstruct). Try to isolate the object from the environment using the depth thresholds in the KinectFusionExplorer-D2D sample, or when calling NuiFusionDepthToDepthFloatFrame in your own code. Optionally mask the environment around the object from the reconstruction by setting pixels around the object in the input depth image to 0 depth. For scanning people, the playermask can potentially be used to perform a rough segmentation – see the GreenScreen Kinect For Windows sample applications.

dose anyone use kinect fusion to map the moving object before?
Reply | Threaded
Open this post in threaded view
|

Re: How to use kinect fusion to map a moving object?

anson2004110
This post has NOT been accepted by the mailing list yet.
how to use kinect for moving objects pose tracking?
dev
Reply | Threaded
Open this post in threaded view
|

Re: How to use kinect fusion to map a moving object?

dev
This post has NOT been accepted by the mailing list yet.
Hi,

Well it depends, what kind of dynamic scene that you wanna reconstruct:
1- If the moving objects move rigidly (rotation and translation)  and the kinect is steady, the algorithm can reconstruct the scene, since It computes the ICP matrix (rotation+translation)

2-If the moving objects move non rigidly (change of geometry, hidden parts) like moving a hand, kinect fusion cannot reconstruct it.

Cheers,
Y.Andam