Pedestrian Safety or ‘Beyond the Induction Loop’

Image Source: Carlos Osorio/Toronto Star

One of the more complex challenges in traffic management is the safe road crossing of pedestrians. Pedestrians are the most vulnerable road users and especially children or the elderly might not even have an alternative to participating in traffic. There are ongoing efforts to increase the number of crosswalks and improve their design. The options range from various traffic lights over nudging elements for motorized traffic to different positions and the general designs of crosswalks.

However, all of these interventions are ineffective if they do not meet the natural needs and behaviors of pedestrians. Significant mismatches (missing or misplaced crosswalks) lead to a behavior known as ‘jaywalking’ or ‘rogue/wild crossing’. And hence to some of the most dangerous situations in urban traffic.

All that said: By observing pedestrian behavior,  it is easy to gain insight into the need for changes to the placement and design of crosswalks.


The key elements are surprisingly simple. For each counting element, do not just count expected traffic along lanes and intersection origin-destination tracks.

Simply add one or more counting lines in the middle of the road and have the optical sensor count the number of pedestrians crossing the road in each respective zone.

Image Source

This grants immediate insight into the acceptance of current installations and possible remedies such as: 

  • Relocation of protected zebra crossing
  • Obstacles such as fences to prevent unauthorized  crossing
  • Nudging signals to push pedestrians back to protected crossings


For far too long, we have focused on measuring the expected. Induction loops count traffic only in expected lanes. Infrared or photoelectric barriers count pedestrians at traffic signals.

Safety, however, is about knowing what happens unexpectedly: Cars and bicycles driving in the wrong lane or even in the wrong direction, pedestrians crossing where there is no crosswalk or traffic light, vehicles speeding, bicycles with high-velocity differences on narrow lanes, buses overtaking bicycles without a  safe distance. Improving safety in increasingly tense traffic situations requires better infrastructure and smart sensors capable of measuring the impact of any intervention. The first step can be taken tomorrow, if not today. Save lives.

Innsbruck, AustriaPedestrian Safety or ‘Beyond the Induction Loop’

Dumb Sensors (aka Inductive Loops) vs. Smart Swarm Sensors

We often think of students counting traffic as (until very recently) the old-school way of doing traffic counts (aka Students-as-a-Sensor (SaaS)). We think of Inductive Loops as the modern traffic counting sensor.

But when was the first real traffic counting sensor? The 1980s? 1970s? Try 1937! In an early project...traffic recorders….operated off a strip laid across the street, and used a six volt battery. Each hour it printed off a paper strip with the total for that hour…” (source)

In fact, current Inductive Loops are a slightly updated version of the same 1935 sensor. They may lie beneath the asphalt, but they are fundamentally the same sensor technology (minus the paper printer!).

Inductive Loops are:

  • Expensive and disruptive to deploy
  • Are prone to hardware failure
  • When they fail are expensive to replace and redeploy
  • Are not updatable
  • Cannot handle slow or stationary traffic
  • Are Inaccurate

“Dumb sensors, always remain dumb” (Wieland Alge)

People are often skeptical that a video-based sensor, using standard IP cameras and Dynamic Perception Technology, can be as accurate as ‘tried-and-tested’ Inductive Loops. This is in fact not the case, as two recent independent tests (by a globally active corporation with more than 50 years of experience in traffic management) have shown.

Parking Area Entry/Exit Test

Swarm Perception was deployed for one camera overlooking an entry-exit area to an outdoor UK parking area, where there were exiting Inductive Loops to count entries and exits. The Hikvision camera used costs just under €200.Over a period of a few weeks, the organization ran a live test comparing the accuracy of Inductive Loop, Swarm Perception, and Manual entry/exit counts of vehicles (day and night). Swarm Perception delivered significantly more accurate counts than the Inductive Loops. Induction Loops missed some vehicles (especially on entry) and also over-counted vehicles (especially on exit).


Why? Well, one example shows this. A car broke down at the exit, causing traffic chaos around it, this made a mess of the inductive loop counts (as cars were either stationary, drove too slowly, or drove around the Inductive Loops) while Swarm Perception continued to count as accurately as ever (without even breaking into sweat!).

Fast Highway Test

In a different test, by the same organization, Swarm Perception was deployed for one exiting standard IP camera overlooking a fast highway in Denmark, where there were exiting Inductive Loops.

Over a period of weeks, the organization compared traffic counts between the Inductive Loops, Swarm Perception, and Manual Counting over 15 minute periods. To make it more fun, the camera with the worst resolution and worst night vision configuration was chosen for the test. About 7500 vehicles pass the test location on a usual working day.

Yet again, Dynamic Perception Technology was shown to be more accurate than the Inductive Loops, even at night-time! This was despite having fast-moving vehicles further away from the camera.

Innsbruck, AustriaDumb Sensors (aka Inductive Loops) vs. Smart Swarm Sensors

How Can You Use Your Existing Infrastructure to Gain Smart Data?

You want to gain smart insights into your business but don’t want to exchange your entire camera system for smart cameras? Investing in expensive counting systems is not an option for you? Now you have the chance to gain smart insights cost-efficiently using your existing cameras. How does this work? With a device for your camera, called the Swarm Perception Box. It’s a small device that enables you to gather smart real-time data with the latest AI technology, right at the camera location. The installation only needs power, internet connectivity, and a connection to the camera.

Easily control what data you get

After the plug and play installation, you can configure your device in the Swarm central management platform, using our detailed documentation. The great advantage of this self-service setup in the Swarm Control Center is that you can define the data collection according to your needs and requirements, making it fast and smart. As the picture below shows, you can draw counting areas or entry-exit-zones on the camera’s view to gather location-specific information. It only takes you a few clicks to set up and requires no special computer skills or programming knowledge. You can adjust the setup for every intersection or road. When you think about it, every street or intersection is different, so you should be able to adapt your data gathering accordingly.

What smart real-time data can you get?

As well as traffic data, as briefly described before, you can also gather smart parking, retail, and public transport data. Here is a short list of just some of the possibilities:


  • Identify the number of people entering and leaving your retail space
  • Coming soon: Estimate customers’ age range and presented gender


  • Identify the flow of road users across intersections, sidewalks, and bike paths with entry and exit zones.

  • Identify cars, buses, trucks, cyclists, motorcyclists, and pedestrians with smart object classification


  • Recognize the occupancy rate of parking spots

  • Identify parking vehicle classes including cars, vans, motorcycles, and trucks

Public Transport

  • Identify the number of people in public transport vehicles and stops or stations
  • Coming soon: Estimate passengers’ age range and presented gender

What are the benefits?

  • It’s fast to deploy as it works with your exiting cameras
  • Make better decisions faster. You get real-time data fast, as we analyse video streams in real-time 24/7 at up to 30+ Frames per Second (FPS).
  • It’s easy to integrate into your infrastructure as we use standard Cloud, IoT and Edge Technologies
  • It’s massively scalable. 1000s of Swarm Perception Boxes could be easily managed via our central management
  • It’s much more affordable than counting systems or smart-cameras. We offer transparent and fair pricing, via a predictable all-inclusive annual fee.
  • You save time and money by purchasing only the smart component you need
  • Our smart box is a software defined sensor, meaning that new features and improvements are automatically delivered at no extra cost
Innsbruck, AustriaHow Can You Use Your Existing Infrastructure to Gain Smart Data?