The videos from three caeras, the foreground regions and the feet positions are extracted. The feet positions are correctly calculated despite the ambiguities of the projection.
