Skip to main content
All CollectionsFLOW - real-time monitoringUse-case tutorials - How-to Series
Shoulder parked vehicle detection - How-To Series in FLOW Insights
Shoulder parked vehicle detection - How-To Series in FLOW Insights

Learn about dangers of shoulder parking and how you can detect shoulder parked vehicles to help you prevent accidents.

DataFromSky avatar
Written by DataFromSky
Updated over a week ago

Stopping the car on the shoulder or in the middle of the road does not only influence the traffic flow negatively, but it can also be dangerous. If such a situation happens, it's important to detect it immediately to be able to warn other drivers and possibly assist the driver with clearing the way if their car broke down. Our AI video analysis framework FLOW is able to detect stationary vehicles in specific locations. You can even measure the duration of blocking. Read this step-by-step guide and learn how easy it is to set up shoulder parked vehicle detection in FLOW and in turn improve road safety.

If you are new to FLOW - read this article or watch our webinar to get familiar with the basics. Links to some helpful FLOW guides can be found below.

What will you learn from this article?

  1. How to set up live detector of shoulder parked vehicles

  2. How to track the blocking time of the vehicles

  3. How to get real-time notifications of shoulder-parked vehicles

If you do not have the FLOW Insights demo kit yet - download it here for free!


First, we drag and drop a category filter to the canvas. We double-click it to open its menu and we tick boxes to select the following categories: car, light, heavy, and bus as shown in the picture below to not detect any other objects such as animals.

Second, we create a zone in which we want to detect shoulder parked vehicles by drawing until we connect the lines to form at least a three-sided shape. We always select slightly larger areas to account for vehicles being shown outside their lanes. After creating it we name the zone "shoulder parked vehicle detection zone". You can draw any number of zones in the analytics. After the zone/s are created we drag them over to the canvas on the right side and connect them to the category filter.

We also need to tick “Now” under the time mode of the zone. This is needed to count only currently present shoulder parked vehicles instead of vehicles detected over time.

In order to detect the shoulder parked vehicle, we need to add a Stationary duration motion filter to the canvas and connect it to the shoulder parked detection zone. When we have the filter selected, we select “Value” then “Greater” and then type in the number based on your needs.

Next, we add “value” widget to the stationary duration filter. When we add the value widget we name it “current number of shoulder parked vehicles” In the widget settings we select now mode and set colors scheme - green if no car is present in the zone and orange and red for if there is 1 or more cars present in the zone. We do this so we can better notice if there is a problem when monitoring the dashboard.

To have a better overview of the situation we can add a distribution, trajectory view, and table widgets to the category filter in the canvas. We can then monitor the situation using the dashboard tab where we can see the number of shoulder parked vehicles detected as well as the current situation on camera.

To count all the blocking cars from multiple zones we can add the Union Set operator in the canvas and connect it to all the zones for detecting blocking vehicles. To collect long-term data on this you can add a value widget with whole history settings. This will give you the total number of the blocking cars that were detected.

To create alerts with open API. Simply add sinks from the sink menu, in a similar way you would add a widget but on the left side of the element instead. Learn more about API.


You have learned how you can detect shoulder parked vehicles and how you can get real-time notifications about them. You can get data from FLOW to your highway message boards to alert the drivers of this possibly dangerous situation or you can have someone who can respond to the situation such as police or highway maintenance service to remove the shoulder parked vehicle or add warning signalization in case the road is blocked significantly.

Setting up detection with FLOW is easy thanks to its intuitive visual programming interface and seamless integration using either open API or I/O interfaces. A variety of connectivity options is available including LoRaWAN, LTE, Wifi or Ethernet. FLOW can be used for a variety of use-cases and has a perfect solution for every situation. The FLOW edge devices that we offer are the TrafficCamera used for simple two-way roads and the TrafficEmbedded capable of perfect detection on complex junction using a single camera but also detection from of up to 6 cameras with simpler scenes. Lastly, we offer the centralized solution the TrafficEnterprise - ideal for city-based server processing. It is also possible to use existing camera networks with the Traffic Enterprise making it very cost-efficient.

We hope you enjoyed the guide!

To browse more articles about FLOW click here.

Do you have any questions? Do not hesitate to contact us here.


• Do you have more questions? Contact us!

• Follow us on Linkedin.

• Learn more about FLOW.

• Visit our homepage to view products and news.

Thank you for your interest! Have a great day!

Let the traffic flow - with FLOW!

Did this answer your question?