Skip to Main Content
With the development of heterogeneous camera networks working at different wavelengths and frame rates and covering a large surface of a vacuum vessel, the visual observation of a large variety of plasma and thermal phenomena (e.g., hot spots, ELMs, MARFE, arcs, dusts, etc.) becomes possible. In the domain of machine protection, a phenomenological diagnostic is a key element toward plasma/thermal event dangerousness assessment during real-time operation. It is also of primary importance to automate the extraction and the storage of phenomena information for further offline event retrieval and analysis, thus leading to a better use of massive image databases for plasma physics studies. To this end, efforts have been devoted to the development of image processing algorithms dedicated to the recognition of specific events. However, a need arises now for the integration of techniques developed so far in both hardware and software directions. We present in this paper our latest results in the field of real-time phenomenon recognition and management through our image understanding software platform. This platform has been validated on Tore Supra during operation and is under evaluation for other present Tokamaks and for the foreseen imaging diagnostic of ITER.