WITHDRAWN: 3D recognition & automated access control reconstruction from a single image

Abstract

A novel algorithm has also been proposed for handling occlusion in 3D object recognition using monochrome images. Binocular Stereo vision system has been used to capture the left and right images of an image scene to infer 3D information. Surface model of the model object has been created using Computer Aided Design (CAD) software, as complex surfaces can easily be identified. A feature detector has been proposed to detect the local features of the object model and virtual camera has been used to create the comparator models for the computed 3D poses. The stereo images have been pre-processed using affine transformation to remove geometric distortions or deformations due to non-ideal camera angles. Bilinear interpolation technique has been used to fill the unknown points in the projected image during the image rectification process. During the process of recognition, the objects are recognized by matching the comparator models with the rectified image regions. A modified geometric mapping technique has been proposed for 3D reconstruction of the rectified mapped stereo images. The 3D reconstructed image has been compared with the object model for validation.