Video-based traffic counts: efficient, flexible, accurate

Increasing traffic counts' accuracy, efficiency, and reliability is a never-ending process for traffic surveyors and municipalities. The same holds for one of our clients, 4-Traffic, a team of traffic survey specialists based in the Netherlands. 

Automation is the answer to all these struggles. By joining forces with GoodVision and making our video-based analytics tools a staple in their toolbox, 4-Traffic brought their traffic counts to a new level with a vehicle detection reliability rate of over 99%.

But you don’t have to believe the numbers alone: see how 4-Traffic applied GoodVision’s solutions to automated traffic counts in multiple locations across the Netherlands, how traffic counts are calculated, and what our partners gained through video-based traffic analytics.

Efficiency: traffic counts that do more, for less

Video-based traffic count systems have multiple advantages over manual methods, but one of the most important is efficiency. Unlike human surveyors, cameras can monitor traffic around the clock at multiple locations without losing focus or making simple errors. The automated method reduces traffic count costs over a long time and generates much more data, resulting in more accurate traffic analysis.

For these reasons, 4-Traffic sees traffic count efficiency as a key factor in all its projects. In Amersfoort, Royal HaskoningDHV, a leading global engineering consulting company, tasked the team with conducting a traffic analysis on an intersection near the Prins Bernhard barracks. Video-based monitoring enabled 4-Traffic to quickly determine the number of crossing movements of cyclists and pedestrians and the traffic flows of motorised vehicles. In just 14 days, our partners counted more than 182.000 vehicles.

The cameras can operate at any time, and thanks to their power supply, they can be independently deployed in the most convenient location. Images and recordings captured by the cameras are instantly turned into reports and numerical data available on the GoodVision Video Insights platform. This expedites the entire process compared to manual, pen-and-paper traffic count methods.

Flexibility: multimodal traffic counts in all situations

Automating traffic counts for straight highways with nothing but cars moving through is one thing; doing the same for complex intersections with mixed traffic is another. Still, multi-directional road sections affect traffic the most, so collecting data from roundabouts and intersections is extremely important. Then, there’s non-motorised traffic that also needs to be measured.

So, how does 4-Traffic use GoodVision’s traffic video analytics for various road surveys?

1. Intersections and roundabouts: GoodVision recognises the direction of vehicles travelling through the intersection or roundabout. Each manoeuvre can be tracked separately by defining the entry and exit points in Video Insights, e.g., only for vehicles turning right at the roundabout or entering the intersection through a specific branch.
   
   If the section includes crossings, the solution can detect and count the number of pedestrians and cyclists crossing. The roadway's direction and volume of cycle traffic can also be tracked.
   
   
2. Cycle paths: In GoodVision, users can define the zones where they want to measure traffic by drawing them on the camera feed in the interface. This allows tracking specific areas, such as cycle paths, without collecting data from the entire field of view.
   
   
3. Pedestrian crossings: A similar analysis can be carried out for pedestrian crossings to determine the volume and direction of walking traffic on multimodal road sections. Identifying zones where pedestrian and vehicle object trajectories cross often is particularly useful for road safety research or deciding whether crossings are positioned correctly.

The same method can be applied for pedestrian-exclusive zones like plazas or parks, e.g., to run footfall analytics for retail or public events.

For example, in Groningen, 4-Traffic was tasked not only with conducting a parking survey but also with recording traffic movements in four distinct locations around the city park. Using traffic cameras, they were able to easily track the number of pedestrians and cyclists passing, entering, and leaving the park in multiple areas. In total, 4-Traffic counted more than 113.000 cyclists spread across four locations over a 14-day study period.

It’s not just cars, pedestrians, and cycles that can be tracked. The algorithm can recognise eight different object classes and create custom ones. Thanks to that, users can measure and classify traffic even in the most complex, multimodal road sections.

Accuracy: near-flawless source of traffic data

Of course, neither efficiency nor adaptability would mean much if video-based traffic monitoring couldn’t deliver reliable data. Automated traffic counts are immune to human error but how accurate are they exactly?

GoodVision uses a proprietary AI algorithm to recognise and classify vehicles and manoeuvres with incredible accuracy. This rate can be further improved by optimising the collection and verification process. What’s 4-Traffic’s approach and how did it enable them to achieve near-100% traffic count accuracy?

During the analysis, 4-Traffic applies three different verification methods to warrant the highest accuracy of the collected figures:

1. Comparing incoming and outgoing traffic — The first step is to compare the volume of traffic entering the intersection to the amount of traffic leaving the intersection at all intersection branches. If both figures are equal, it’s a good sign that the results are correct and that the analysis is dependable.
2. Multiple cameras — Deploying additional cameras overlooking the same road section from various perspectives and comparing the data they collect is an easy way to boost monitoring accuracy. The extra cameras can be installed permanently or temporarily to avoid additional expenses. Either way, the lower error margin will pay off over time.
3. Manual verification—Though video-based traffic analytics systems are designed to eventually replace manual traffic counts, traditional surveying methods can still prove useful. 4-Traffic occasionally runs manual traffic counts and randomly compares its results with the data collected automatically to verify whether the latter is correct.

4-Traffic applied these verification methods during their traffic count project for Royal HaskoningDHV at the Aletta Jacobslaan – Geerdinksweg intersection in Hengelo. The team installed two cameras that independently monitored incoming and outgoing traffic at the intersection. The results were also cross-verified with manual counting based on the images. Combining the three methods enabled 4-Traffic to hit near-perfect traffic count accuracy rates: 99.7% for motorised and 97.1% for non-motorised traffic.

Moving forward with video-based traffic counts

In the age of rapidly developing cities and growingly intense mobility, municipalities and traffic engineers need access to ever more complex and accurate data, faster, and at a lower cost. This puts traffic surveyors under pressure to enhance their data collection methods without compromising quality or quantity.

4-Traffic has overcome this challenge by partnering with GoodVision to create a smart and automated source of traffic data and analytics. Want to join them and our other partners? Get in touch!

You can read more about our partner, 4-Traffic on their website.