3D model reconstruction with kinect

This article introduces 3D model reconstruction using Microsoft Kinect and OpenNI.
Kinect is an image and depth sensor, and outputs color images and depths images in 30 frames per second. Our reconstruction program detects characteristic points from color images, tracks the points in frames, and calculates where each point moved in camera coordinate by depth images. On the condition that things in the space are fixed, it can be assumed that all movements of characteristic points in frames are caused by the movement of the camera (kinect), so the variation of position and direction of the camera in each frame can be calculated. The program integrates the variations, calculates and correct the position and direction of the camera in absolute coordinate, and reconstructs 3D models by putting pixels of color images in absolute coordinate.

3d reconstruction
This video is a demo of 3D model reconstruction. this shows the process of the reconstruction in x16 speed.
3D model are represented as a set of 2cm grid voxels.
Left view is color images from kinect. Light green lines show tracking of characteristic points. Right view is reconstructed 3D model.

This entry was posted in 未分類. Bookmark the permalink.