3D Shape Reconstruction using Single Camera and Two Mirrors

Abstract

This paper presents a novel approach to 3D shape reconstruction of objects using a single camera-based stereo vision system. The system is based on two convex mirrors attached and aligned properly in front of a low-cost camera. The camera captures the stereo images of a scene formed in the two mirrors, and the 3D shape of any object present in the scene is reconstructed. The 3D reconstruction is performed by extracting the target object from the stereo images and applying a proper 3D reconstruction model. In the present work, the 3D reconstruction of a target object has been performed by computing disparity map and using 3D point clouding technique. The depth of each point of the objects and the disparity is determined using featured-based algorithm. The depth image, which sometimes is called point clouds or grayscale image, has been used to generate the 3D shape and position of an object. This new system adds the advantages of using the principle of stereo vision for 3D reconstruction and reduces the shortcomings of using a pair of cameras in the conventional stereo vision systems. The performance of the system is verified by identifying different objects of different shapes and sizes and reconstructed 3D outputs from captured stereo images.