Application of Inventive Design Methodology for a Sensor Fusion Approach in Structure-from-Motion (SfM) Applications

Generating three-dimensional models from real world objects based on images from a single camera is an important application in computer vision and numerous possible applications. For this Structure-from-Motion (SfM) algorithms provide an approach based on the successful tracking of 2D feature points throughout an image sequence and the subsequent camera egomotion estimation and a global 3D structure recovery. However, the performance of those approaches is still limited according their precision, denseness and completeness of the reconstructed scene model. This paper introduces an approach based on Inventive Design Methodology (IDM) for the systematic analysis of existing SfM approaches and a subsequent derivation of a novel sensor fusion approach in order to overcome current application-related restrictions.