The structure of the movement, the restoration of a 3D point cloud when setting the correspondence of two image points

Question

Use case

The object rotates around its center with variable speed.
A fixed camera looks at an object
Given the correspondence of two-dimensional images, a 3D point cloud
As the object rotates, therefore, another part of it is visible on it, and thus different points and correspondences are detected.

Scene

and. N images
b. N-1 pairs of images
c. N-1 2D point correspondence (two arrays of two-dimensional points)

Implementation

For each of the (N-1) two-dimensional correspondence points

Calculate relative camera pose
Triangulate to get 3D glasses
For each 2 arrays of 3D points, derive a correspondence using the two-dimensional correspondence given in [c]
3D- @[3], , /

:

A (N-2) 3D-, , ( )

, :

, , .
, A. Camera:
( ). . , , , , ..
. [B], 3D- [A] Cameras Translations

D. ( @[4])

Pose), .

?

+4

NadavRub 12 '14 15:45

1

PeterNL · Accepted Answer · 2014-05-15T12:27:08+0000

. . , , , .

intereset. SURF SIFT - . 15 °, , USURF, , SURF ( SURF). Optical Flow OpenCV, , . , .

. , (, - - )? , . , , . , , .

. - . , . , . , , . , , , .

, . . , , , RANSAC . , , , Intereset .

, , - , . 2D 3D- . PerspectivenPoint Camera Pose Estimation (PnP).

. . , :