Towards a virtual environment for interactive analysis of cluster-based flow pattern abstraction
Suryatapa Roy, Yaoping Hu, Robert John Martinuzzi, Chris R. Morton · 2017
Recent research efforts show the benefits of using machine learning and interactive visualizations in data analytics. However, there is a void in the implementation of these techniques for the analysis of large and complex 4-dimentional (4D) unsteady flows. Hence, this paper presents an initial development of a virtual environment (VE) to fill this void. The VE has a two-layer architecture with different technologies running in the back and fore grounds. Using machine learning, the background layer implements clustering algorithms for abstracting spatiotemporal patterns of the flows like flow features, spatial regions and temporal phases. The outputs of the clustering algorithms are fed to the foreground layer for interactive selection, sorting and filtering of these patterns. The operations in the foreground make up our designed techniques of flow pattern analysis that aims to provide greater comprehension of 4D flow data. Running in the foreground, the virtual reality (VR) technologies of stereoscopic rendering and haptic feedback further enhance these analysis techniques. Thus, our work introduces a novel environment for the interactive analysis of cluster-based flow pattern abstractions.