Traffic Analysis

A nature preserve is suffering from dumping of chemical waste. At their disposal was one year’s worth of license plate recognition data, recording what vehicles goes where in the park at what time. The goal was simple: catch the car dumping the waste.

Spot the one car dumping chemical waste from over 170.000 datapoints from a license-plate recognition system? In the 2017 IEEE VAST challenge co-founder Bram Cappers singlehandedly outscored more than 50 6-person industry teams working on a case for over three months.

Traffic Analysis

Anomaly Detection

From amongst all the detected license plate, our software was automatically able to categorise cars into several categories:

  • Consumer vehicles, coming for a visit to the preserve
  • Park rangers, working at the park
  • Organised trips of busses into the preserve

But one car stood out! Already on its first pass over the data, our detection software spotted a single vehicle that:

  • Visited the park at unusual hours
  • Stayed in the park an unusual amount of time
  • Drove an unusual type of car
  • Visited the park unusually often

Indeed, the anomalous car was responsible for dumping the waste

Result

Where over 50 industry teams comprised of up to 6 people took three months to tackle this challenge. AnalyzeData co-founder used our state-of-the-art anomaly detection software to found the outliers in a matter of hours!

Although we’re proud to say we took first place in this challenge, unfortunately it was an offline challenge. But imagine if we could make this business case a reality: a park ranger could have been deployed to the vehicle even before they could dump their waste, preventing large-scale damage to the local natural ecosystem.

IEEE Vast