The paper presents a work on video sequence analysis. The possibility of extracting a backgrpund information from a series of frames allows analysing the spatial and temporal relationships among objects. The paper firstly describes the extraction of a mosaic from a series of images and introduces a robust mosaicing algorithm. Secondly introduces methods to analyse the movement of objects, independent from the background, and to detect the camera movement. Thirdly defines a linguistic interface, based on a fact-data-base that is automatically built by the system. These facts and their interpretation constitute an intelligent approach to the video analysis. In particular to video indexing and retrieval and to automatic knowledge extraction.
展开▼