Event detection in a video sequence
Sébastien Lefèvre, VINCENT, Nicole, PROUST, Christian · OpenGrey (Institut de l'Information Scientifique et Technique) · 2002
Here we are concerned by multimedia data indexing by use of pertinent sample searching. Our work focuses on video sequence analysis in order to detect some predefined events. This search being contextual, we propose an architecture and some rapid tools, both generic, for building specific indexing systems. We deal in particular with the following problems : data temporal segmentation, background and object separation, background structure extraction, object tracking (rigid or non-rigid objects, with or without learning) and audio data analysis. In order to solve these different problems, the generic tools we are proposing are based on semi-local analyses, multiresolution approaches, hidden Markov models, and the active contour method. The architecture and the tools introduced here have been validated through several applications.