Three-dimensional reconstruction system with dual cameras using artificial neural network

A three-dimensional (3-D) reconstruction system using an artificial neural network (ANN) technique is presented. A mathematical model and an ANN framework are used to construct the 3-D reconstruction system implemented with twin pinhole cameras, which are located side by side. The coordinates of an object are calculated by the mathematical model of the system. An ANN model is trained to compute the coordinates of an object in a 3-D space from the image coordinates of the mathematical model and real image data. The precision of the ANN model is investigated. The experimental platform of the 3-D reconstruction system contains three kinds of components, including dual cameras, an optical linear encoder, and object points. The experimental results of 3-D point objects demonstrate that the high interpolation precision of the coordinates is achieved by the ANN. The accomplished 3-D reconstruction system exhibits fast and accurate advantages for 3-D computer vision applications.