High Performance Spatial Visualization of Traffic Data

Project summary:

The objectives of this project were to develop high performance spatial tools and techniques to generate critical visualizations of loop-detector traffic data collected at the Traffic Management Center in Minneapolis, Minnesota, and to explore new spatial data structures and algorithms to address the performance bottlenecks in the process of producing novel interactive traffic visualization.

The researchers identified the main performance bottlenecks to be the large size of the traffic datasets, which cannot fit inside main memories of common workstations. Researchers constructed a web-based, video-like visualization software package for observing rapid summarization of major trends. This package can be used to visualize the effects of a sudden increase in load on the traffic network after scheduled events, to assist in the planning of traffic management for similar future events.

The researchers extended the visualization package to support several data mining techniques (e.g., clustering, classification, and outlier detection). In addition, they also identified the performance bottlenecks in the generations of various visualizations and developed efficient algorithms to address the bottlenecks.