3D object model recovery from 2D images utilizing corner detection

Abstract

This research proposes a method to reconstruct the 3D object model from 2D images. The reconstruction theorem this research used is based on the stereo vision algorithm. Stereo vision is one type of the non-contact scanning measurement. The theorem of the stereo vision were simulates the human's eyes to calculate the object's depth information. There for, this research uses two CCD Cameras to capture the object's 2D images. Then find out the match points from the left and right two images. To reconstruct the object's 3D model by using the match points, parameters of the CCD Cameras and transform matrix that between the world coordinate and camera coordinate. The object this research reconstructed belongs to simple geometry. There for, we can reconstruct the object's 3D model by the feature points of the object. The feature points always appear on edges and vertices. In this way we can reduce the number of the points we should reconstruct, so the system can build up the object's 3D model fast. The CCD Cameras unable to capture the whole object's surface information in once time, so this research proposes a method that rotates the object to solve this problem. The system combines the rotation model with the reconstruction system to reconstruct the complete object's 3D model.