ACM / Crossroads / Xrds14-2 / Visualizing Flow Data Using Assorted Glyphs

Visualizing Flow Data Using Assorted Glyphs

by Amit Prakash Sawant and Christopher G. Healey

Abstract

This project visualizes a scientific dataset containing two-dimensional flow data from a simulated supernova collapse provided by astrophysics researchers. We started our project by designing visualizations using multiple hand drawings representing the flow data without taking into consideration the implementation constraints of our designs. We implemented a few of our hand drawn designs. We used an assortment of simple geometric graphical objects, called glyphs, such as, dots, lines, arrows, and triangles to represent the flow at each sample point. We also incorporated transparency in our visualizations. We identified two important goals for our project: (1) design different types of graphical glyphs to support flexibility in their placement and in their ability to represent multidimensional data elements, and (2) build an effective visualization technique that uses glyphs to represent the two-dimensional flow field.

Introduction

Visualization is an area of computer graphics that manages and presents information in a visual form to facilitate rapid, effective, and meaningful analysis and interpretation. Visualization is used in areas such as geographic information systems, land and satellite weather information, scientific simulations, aerospace research, molecular biology, defense, and medicine. Visualization also supports more abstract domains, for example, program visualization, data mining, and network security. In situations where user collaboration is required or time is a critical factor, visualization enables people to analyze and interpret vast amounts of information and make important decisions. The desire to extract knowledge rapidly and efficiently from large, complex datasets motivates the need for effective visualization systems [6].

Visualization presents information in a pictorial form [1]. Formally, a dataset D contains n data elements, e_i, such that D = {e₁,..., e_n}. A dataset represents a set of data attributes, A = {A₁,..., A_m}, where m > 1. Data elements encode values for each attribute: e_i = {a_i,1, ...,a_i,m}, a_i,j belongs to A_j. A data-feature mapping converts raw data into visual information. Such a mapping is denoted by M(V, θ), where V = {V₁, ..., V_m} is a set of m visual features with V_j selected to represent each attribute A_j, and θ_j : A_j -> V_j maps the domain of A_j to the range of displayable values in V_j. Thus, visualization is the process of selecting an appropriate M. An effective M produces images that support rapid, accurate, and effortless exploration and analysis [4].

Figure 1: Visualization of a weather dataset using perceptual texture elements with temperature mapped to color, wind speed mapped to coverage, pressure mapped to density, and precipitation mapped to orientation.

Figure 1 demonstrates the visualization of a weather dataset. Individual weather readings (or data elements, D) are visualized using small glyphs that vary their color and texture properties (V). The mapping M used in Figure 1 is as follows:

• Color represents temperature; bright pink and red strokes for hot temperatures to dark green and blue strokes for cold temperatures.
• Coverage represents wind speed; tightly packed strokes with little or no background showing through for strong winds to sparsely packed areas for weak winds.
• Density represents pressure; more strokes displayed in a fixed area of screen space for increasing pressure.
• Orientation represents precipitation; vertical strokes for little or no rainfall to horizontal strokes for high rainfall.

The need to visualize vector fields is common in many scientific and engineering disciplines. It is difficult to analyze and interpret this collection of flow field data using traditional nonvisual techniques. One possible solution is to use visualization techniques to convert large amounts of data into an image that domain experts can use for exploring, discovering, comparing, and validating. In this article, we present a technique to visualize two-dimensional flow fields using individual glyphs.

The remainder of this article proceeds as follows. In the next section, we describe our hand designs. Section Flow Visualization describes the design of our flow visualization tool. Section Types of Assorted Glyphs provides details about our different glyph representations. Finally, the last section discusses conclusions and future work.

Hand Designs

We proposed three hand designs as shown in Figure 2 to represent the two-dimensional flow data (flow vectors) on a regular grid, taken from a simulated supernova collapse. We used colored direction arrows, color, and curves to represent the flow as shown in Figure 2.

Figure 2: Hand designs - colored direction arrows, color, and curves.

Flow Visualization

Flow Data

Data points within each slice form a 500 x 500 regular grid of flow sample points composed of the attributes du, dv, magnitude, pressure, and density. du and dv were used to determine the direction of the flow at each sample point.

Data-Feature Mappings

A variety of visual features have been used in visualization [3]. The use of color and texture has a long history in the graphics, vision, and visualization literature. Our data-feature mappings are constructed from psychophysical studies of how the visual system "sees" fundamental visual properties in an image. Color is widely used in many visualization techniques. Scientists have used fundamental perceptual properties of texture such as regularity, directionality, contrast, size, and coarseness to model human vision, to perform texture segmentation and classification, and more recently to visualize multiple data attributes [2, 5]. In visualization, an individual attribute's values control a corresponding texture dimension. This results in a display that changes its visual appearance based on the underlying dataset.

Figure 3: Data-feature mapping: u mapped to x position, v mapped to y position, pressure mapped to hue and size, density mapped to luminance, and magnitude mapped to transparency.

When we design a visualization, properties of the dataset and the visual features used to represent its data elements must be carefully controlled to produce an effective result. Important characteristics that must be considered include [8]: (1) dimensionality (number of attributes in the dataset), (2) number of elements, (3) visual-feature salience (strengths and limitations that make it suitable for certain types of data attributes and analysis tasks), and (4) visual interference (different visual features can interact with one another, producing visual interference; this must be controlled or eliminated to guarantee effective exploration and analysis).

Using the knowledge of the above perceptual guidelines, we choose visual features that are highly salient, both in isolation and in combination. We map features to individual data attributes in ways that draw a viewer's focus of attention to important areas.

Our glyphs support variation of spatial position, color, and texture properties, including x position, y position, hue, luminance, height, size, width, orientation, and transparency. A glyph uses the attribute values of the data element it represents to select specific values of the visual features to display. The most important attributes should be mapped to the most salient features, and secondary data should never be visualized in a way that would lead to visual interference. The order of importance for visual features is luminance, hue, and then various texture properties. We are particularly interested in representing the density and pressure of the flow data along with the direction of the flow field. Currently, we only represent steady flow. Figure 3 shows an example of a data-feature mapping used to visualize flow data. Figure 4 walks through the steps for generating flow visualizations.

Figure 4: Flow visualization system showing the input dialog, the data-feature mapping dialog, and the visualization window.

Types of Assorted Glyphs

We used different types of glyphs to represent each flow sample, such as dots, lines, arrows, and triangles. We also used transparency in our visualizations. In the following subsections we briefly describe each assorted glyph.

Dots

Figure 5: Dot representation: u mapped to x position, v mapped to y position, pressure mapped to hue, density mapped to luminance, magnitude mapped to size.

Figure 5 shows the dot representation of the flow data. [7] presents a new technique to visualize 2-D flow fields with a collection of variable size dots.

Interpretation

1. The size of the dots gradually increases from left to right indicating the increase in the magnitude of the flow field.
2. Pressure, represented by hue, decreases from left to right.
3. Density, mapped to luminance, increases from left to right.
4. The variation in the size of the dots (small dots in certain areas) combined with the luminance helps us identify vortices.

Lines

Figure 6: Line representation: u mapped to x position, v mapped to y position, pressure mapped to hue, density mapped to luminance, magnitude mapped to size.

Figure 6 shows the line representation of the flow data.

Interpretation

1. The width of the lines gradually increases from left to right indicating the increase in the magnitude of the flow field.
2. As in Figure 5, pressure, represented by hue, decreases from left to right, and density, mapped to luminance, increases from left to right.
3. The orientation and variation in the width of the lines combined with the luminance helps us identify vortices.

Arrows

Figure 7: Arrow representation: u mapped to x position, v mapped to y position, pressure mapped to hue, density mapped to luminance, magnitude mapped to size.

Figure 7 shows the arrow representation of the flow data.

Interpretation

1. The direction of the arrows represent the flow of the vector field and helps us identify vortices.
2. As in Figure 5 and Figure 6, pressure, represented by hue, decreases from left to right, and density, mapped to luminance, increases from left to right.

Triangles

Figure 8: Triangle representation: u mapped to x position, v mapped to y position, pressure mapped to hue, density mapped to luminance.

Figure 8 shows the triangle representation of the flow data.

Interpretation

1. Two vertices of the triangle are anchored along the vertical axis, and the third vertex is placed in the direction of the flow vector.
2. As in Figure 5, Figure 6, and Figure 7, pressure, represented by hue, decreases from left to right, and density, mapped to luminance, increases from left to right.

Transparency

Figure 9: Transparency: u mapped to x position, v mapped to y position, magnitude mapped to size and transparency.

Figure 9 shows the use of transparency to represent the flow data. We noticed visual interference effects between transparency and luminance. Thus, we do not map an attribute to luminance when transparency is used.

Interpretation

1. The transparency gradually increasing from left to right indicates the increase in the magnitude of the flow field.
2. The transparency variations generate a sand-like or a cloud-like effect that help identify vortices.

Conclusions

We use rules of human perception to visualize flow orientations in an underlying flow field that harness the strengths and avoid the limitations of low-level human vision. Individual flow vectors are represented using graphical "glyphs" that vary their spatial position, color, and texture properties to encode the vector's attribute values. The result is a display that allows viewers to rapidly and accurately analyze, explore, compare, and discover within the flow data. Our method has the ability to represent large flow fields. Our technique is not restricted to astrophysics data. It can be applied in any situation where appropriate ranking attributes can be identified to control glyph placement. We would like to represent unsteady flow. We also plan to conduct simple validation studies to investigate our design choices and determine which glyph representations are preferred and under what circumstances. In the future, we also plan to make our flow visualization tool available to the scientific community for exploring and analyzing their datasets.

References

1

S. G. Eick. Engineering perceptually effective visualizations for abstract data. In Scientific Visualization, Overviews, Methodologies, and Techniques, pages 191-210, Washington, DC, USA, 1997. IEEE Computer Society.

2

R. M. Haralick, K. Shanmugam, and I. Dinstein. In IEEE Transactions on System, Man, and Cybernetics SMC-3, volume 3, pages 610-621, 1973.

3

C. G. Healey. Perceptual colors and textures for scientific visualization, 1998.

4

C. G. Healey. Formalizing artistic techniques and scientific visualization for painted renditions for complex information spaces. In Proceedings International Joint Conference on Artificial Intelligence 2001, pages 371-376, 2001.

5

R. A. Rao and G. L. Lohse. Towards a texture naming system: Identifying relevant dimensions in texture. In IEEE Visualization, pages 220-227, 1993.

6

A. P. Sawant. Dynamic Visualization of the Relationship Between Multiple Representations of an Abstract Information Space. M.S. thesis, North Carolina State University, USA, 2003.

7

L. G. Tateosian, B. M. Dennis, and C. G. Healey. Stevens dot patterns for 2D flow visualization. In APGV '06: Proceedings of the 3rd symposium on Applied perception in graphics and visualization, pages 93-100, 2006.

8

C. Weigle, W. Emigh, G. Liu, R. Taylor, J. T. Enns, and C. G. Healey. Oriented texture slivers: A technique for local value estimation of multiple scalar fields. In Proceedings Graphics Interface 2000, pages 163-170, Montr´eal, Canada, 2000.

Biography

Amit Prakash Sawant (amit.sawant@ncsu.edu) is a PhD student in the Computer Science Department at North Carolina State University. His research interests include computer graphics, scientific visualization, information visualization, visual perception, and databases. Website: http://www4.ncsu.edu/~apsawant.

Christopher G. Healey (healey@csc.ncsu.edu) is an associate professor in the Computer Science Department at North Carolina State University. His research interests include computer graphics, scientific visualization, perception and cognitive vision, color, texture, databases, and computational geometry. Website: http://www.csc.ncsu.edu/faculty/healey.