Modelling of 3D Objects Using Unconstrained and Uncalibrated Images Taken with a Handheld Camera

Abstract

3D models are an essential part of computer graphics applications such as games, movie special effects, urban and landscape design, architecture, virtual heritage, visual impact studies, and virtual environments such as Second Life. We have developed a novel technique which allows the construction of 3D models using image sequences acquired by a handheld low-cost digital camera. In contrast to alternative technologies, such as laser scanners, structured lighting, and sets of calibrated cameras, our approach can be used by everyone having access to a consumer-level camera. The user only has to create a set of images from different view directions, input them into our algorithm, and a 3D model is returned. We use a novel combination of advanced computer vision algorithms for feature detection, feature matching, and projection matrix estimation in order to reconstruct a 3D point cloud representing the location of geometric features estimated from input images. In a second step a full 3D model is reconstructed using the projection matrix and a triangulation process. We tested our algorithm using a variety of data sets of objects of different scales acquired under different weather and lighting conditions. The results show that our algorithm is stable and enables inexperienced users to easily create complex 3D content using a simple consumer level camera.KeywordsPoint CloudSurface ReconstructionScale Invariant Feature TransformFundamental MatrixBundle AdjustmentThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.