Supercomputers to help Solve Urban Traffic Problems

These days, whenever we look up at a traffic light at any busy intersection, we will often see a camera there watching us back. It is a possibility that the cameras have been installed for a higher purpose. They could actually be monitoring traffic conditions and collecting visuals for any collisions. But is it possible that they are doing even more? Will they assist planners as they optimize the flow of traffic or even identify places that are most accident-prone? And perhaps most importantly, will they be able to do this without some poor people having to view hours upon hours of traffic footage?

Scientists who work at the Texas Advanced Computing Center (TACC), located at the University of Texas in Austin sincerely think so. They have been working to create new tools that will allow for sophisticated traffic analyses that the employment of data mining and deep learning.

Supercomputers Learn to Recognize Traffic

At a recent IEEE International Conference about Big Data, they presented info about some new deep learning assets that utilized raw camera footage of traffic from the City of Austin cameras and were able to recognize various objects – and then subsequently characterize the ways those objects interact and move about. This data is then able to be queried and analyzed by traffic officials to determine various traffic situations such as the number of cars that may be driving down a one-way street the wrong way, for example.

“We are hoping to develop a flexible and efficient system to aid traffic researchers and decision-makers for dynamic, real-life analysis needs,” stated Weijia Xu, who is a scientist that is leading the Data Mining & Statistics Group at TACC. “We don’t want to build a turn-key solution for a single, specific problem. We want to explore means that may be helpful for a number of analytical needs, even those that may pop up in the future.”

This algorithm was created for traffic analysis to automatically label any possible object from the raw traffic data, tracking objects from a comparison with other objects which had been previously recognized, and discover the various relationships among these different objects.

Supercomputers Tracking Traffic Patterns

After these scientists developed a system that could label, track and analyze traffic, they then applied the system to two actual examples: counting the number of vehicles that travel down a road and recognizing close encounters between pedestrians and vehicles.

A supercomputer traffic system would automatically count the number of vehicles within a 10-minute video clip, and then present preliminary results indicating that the tool had a  95% accuracy overall.

Being able to comprehend and understand traffic volumes as well as the traffic distribution across time is vital to verifying various transportation models and then evaluating the performance of this transportation network, says Natalia Ruiz Juri, who is a research associate and also the director of the Network Modeling Center at UT’s Center for Transportation Research.

“Current practice often relies on the use of expensive sensors for continuous data collection or on traffic studies that sample traffic volumes for a few days during selected time periods,” she added. “The use of artificial intelligence to automatically generate traffic volumes from existing cameras would provide a much broader spatial and temporal coverage of the transportation network, facilitating the generation of valuable datasets to support innovative research and to understand the impact of traffic management and operation decisions.”

In regards to possible close encounters, using the system allowed scientists to automatically find several cases where pedestrians and vehicles were in very close proximity to each other. None of the encounters were real-life dangers, but they did demonstrate just how the system can find dangerous places and locations without any human intervention at all.

“The City of Austin is committed to ending traffic fatalities, and video analytics will be a powerful tool to help us pinpoint potentially dangerous locations,” noted Jen Duthie, who is a consulting engineer from the City of Austin and a collaborator for the project. “We can direct our resources toward fixing problem locations before an injury or fatality occurs.”