Reconstruciotn of 3D models from 2D images

The application enables the user to reconstruct simple block-shaped objects together with their position in the 3D world from 2D images of the scene. The user is required to draw 3 or 4 points specifying one base of the object while the volume and the position could then be easily derived as the user stretches the rendered object (to match the underlying image).

How it works?

Multiple images of the scene including a printed chessboard must be provided. It is then possible to calculate both the camera intrisic and extrinsic matrix. As the next step the world -> camera frame transformation matrix is computed. Since the chesboard serves the purpose of the marker here and defines an XY plane (Z = 0) it is possible to compute the homography between the world plane and the image. Using the inverse homography matrix the coordinates of the object vertices in the 3D scene are finally obtained.

Install and run

$ cmake
$ make
$ ./objectReconstruction