Research Topics

Rain or Snow Detection by Camera

Deteriorated weather conditions may alter driver safety. A vision-based traffic monitoring systems is proposed to take hydrometeor into account and react accordingly. A probabilistic approach is introduced to make an orientation histogram of the moving streaks present in the image. The geometrical moments method is used to compute the orientation of each segment to build an histogram. An Expectation Maximization (EM) algorithm is used to model the observed distribution. A Kolmogorov-Smirnov test combined with a short-time sliding window allow us to take the decision on the presence of hydrometeors.

Publications Bossu, J., Hautière, N. and Tarel, J.-P. Rain or Snow Detection in Image Sequences Through Use of a Histogram of Orientation of Streaks. In International Journal of Computer Vision, 93 (3): 348-367, 2011.     Hautière, N., Bossu, J., Bigorgne, E. and Aubert, D. Détection de conditions de visibilité réduite par caméra bord de voies. In Prévention des Risques et Aides à la Conduite, Paris, France, 2010.   Bossu, J., Hautière, N. and Tarel, J.-P. Utilisation d'un modèle probabiliste d'orientation de segments pour détecter des hydrométéores dans des séquences vidéo. In XXIIe Colloque GRETSI (GRETSI'09), Dijon, France, 2009.   Hautière, N., Bossu, J., Bigorgne, E., Hiblot, N., Boubezoul, A., Lusetti, B. and Aubert, D. Sensing the visibility range at low cost in the SAFESPOT road-side unit. In ITS World Congress (ITS'09), Stockholm, Sweden, 2009.

Slides PRAC2010a presentation ITS2009 presentation

Posters GRETSI2009a poster

Videos

Related topics Physics-based Vision Detection of adverse weather conditions Roadside systems