Reducing mismatching under time-pressure by reasoning about visibility and occlusion

Abstract

Three methods are explored which help indicate whether feature points are potentially visible or occluded in the matching phase of the keyframe-based real-time visual SLAM system. The first derives a measure of potential visibility from the angular proximity to keyframes in which they were observed and globally adjusted, and preferentially selects those with high visibility when tracking the camera position between keyframes. It is found that sorting and selecting features within image bins spread over the image improves tracking stability. The second method automatically recognizes and locates 3D polyhedral objects alongside the point map, and uses them to determine occlusion. The third method uses the map points themselves to grow surfaces. The performance of each is tested on live and recorded sequences.

Estimating Planar Structure in Single Images by Learning From Examples

“…In general, planes have proven to be useful for 3D reconstructions (Bartoli, 2007), interactive modelling (Sturm and Maybank, 1999), and in visual mapping, for example to simplify the scene structure (Gee et al, 2008) and perform higher-level reasoning (Wangsiripitak and Murray, 2010). We describe a few important examples of single-image methods for finding planar structure, which use appearance information from the image as opposed to any 3D or multi-view information.…”

High-level scene structure using visibility and occlusion

“…Scenes with significant occlusion challenge the state of the art in online SFM and visual SLAM. Wangsiripitak and Murray [16] investigate the difficultly posed by occlusion and propose techniques for dealing with this problem including using higher order scene structure to anticipate occlusion. In this work we propose a model-based approach that exploits both visibility and occlusion information to recover high order scene structure.…”