Machine learning approach for plasma image processing: application to plasma-on-water characterization

Machine Learning (ML) is a set of computational tools particularly suited for the analysis and classification of large datasets. ML has been widely applied to image processing for application such as pattern recognition (e.g., face detection) and image segmentation (e.g., medical imaging, object detection). Digital cameras are one of the most versatile experimental diagnostics instrument. Optical images recorded during plasma experiments are often used to describe the shape and size of the discharge. A generic ML approach for plasma image processing has been developed using MATLAB Machine Learning Toolbox, and applied to image segmentation from plasma-on-water experiments on a pin-to-plate setup. The segmentation identifies and quantifies zones of interests in the discharge (plasma column, plasma-water interface, etc.). The segmentation data is then used for the three-dimensional reconstruction of the discharge. The ML approach can be extended to incorporate other types of data (e.g. voltage signals), which makes it a promising approach to current enhance plasma diagnostics approaches.