The main goal of the master thesis is to develop a system for reconstruction of 3D objects from colour images. The main focus is on passive computer vision methods from which we select two, i.e., Stereo vision and Space carving. Both methods require information about camera poses. The camera pose for a given image is estimated from the information obtained by detecting a reference object, i.e., a standard A4 paper sheet. We develop an Android based mobile application to guide a user during image capturing process and a Matlab based system for 3D object reconstruction from captured images. The reconstructed 3D models of objects where compared to reference 3D models. Using an evaluation metodology we propose, we evaluated the implemented methods and analysed the results. We also present qualitative results and highlight strengths and shortcomings of the two selected passive computer vision methods for 3D reconstruction.
|
