communication mediumMaster's Thesis
Industrial processes generate ever growing amounts of information. With some sensors reporting their state in sub-second intervals, data sets quickly reach large numbers. Computer
software is commonly used to analyze and visualize this information in a human-readable format. One specific problem area is the efficient representation of data in the form of 1D curves.
Drawing all curves on screen at once is not always optimal, because too much information not
only impacts render performance, but also makes it hard for users to find anomalies or trends
within a visually overloaded image. In the course of this thesis, problems and solutions regarding visual clutter are explored and possibilities to form clusters of curves are presented. After
an analysis of novel methods, k-means clustering and functional boxplots for visual clutter reduction are combined and integrated into the visual analytics software platform Visplore. Code
snippets, example figures and performance measurements are used to compare program behavior in diverse situations. Results show that visual clutter reduction speeds up render times,
at the cost of significantly increased computation times. However, if clustering is done in combination with clutter reduction, the heightened computation cost can again be mitigated.
visual analytics, visual clutter reduction, render performance, clustering