3-D model-based multiple-object video tracking for treatment room supervision

We present a method to monitor a patient and the equipment in a radiotherapy treatment room, by exploiting the information in the treatment plan, enriched with other elements such as visual, geometric, and "semantic" information. Using all these information items, and a generic model, a virtual environment of the scene is created, with maximum precision. The images resulting from video sequences with several cameras are also used to confront the filmed information on the scene and its numerical representation. The method is based on the features of the scene elements, and on a fuzzy formalism. The feasibility of the method is being quantitatively evaluated in the absence of treatment, to be further exploited in a module for external control by video in real conditions.