Scale & Affine Invariant Interest Point Detectors
International Journal of Computer Vision
Distinctive Image Features from Scale-Invariant Keypoints
International Journal of Computer Vision
Keypoint-based detection of near-duplicate image fragments using image geometry and topology
ICCVG'10 Proceedings of the 2010 international conference on Computer vision and graphics: Part II
Real-time vehicle tracking for driving assistance
Machine Vision and Applications
Video-based lane estimation and tracking for driver assistance: survey, system, and evaluation
IEEE Transactions on Intelligent Transportation Systems
Hi-index | 0.00 |
A framework for the automatic detection of dangerously deteriorating visibility (e.g. due to bad weather and/or poor illumination conditions) is presented. The method employs image matching techniques for tracking similar fragments in video-frames captured by a forward-looking camera. The visibility is considered low when performances of visual tracking deteriorate and/or its continuity is lost either temporarily (i.e. a sudden burst of light, a splash of water) or more permanently. Two variants of the tracking algorithm are considered, i.e. the topological approach (more important) and the geometric one. Using the most difficult examples of DAGM2011 Challenge dataset (e.g. Snow, Rain and Light-sabre clips) it is demonstrated that the visibility quality can be numerically estimated, and the most severe cases (when even the human eye can hardly recognize the scene components) are represented by zero (or near-zero) values. The paper also briefly discusses the implementation issues (based on a previously developed similar real-time application) and directions of future works.