Telraam Sensor V2

New Sensor • Geschreven door Kris Vanherle op 10 February 2021

Telraam has started a new project to develop a completely new sensor and we need your help building it! In this blog-post, we explain our approach, timeline and explain how you can contribute to this new project.

As reported in our blogpost end of last year, our goal is to start developing a completely new purpose-built traffic counting sensor to replace the current Raspberry Pi-based system. We’re delighted we can now share with you the new sensor development trajectory has been approved for VLAIO funding, giving us the means to be ambitious. We’re partnering with Bagaar and Kapernikov to make this happen. Bagaar is an Antwerp based software company focusing on smart IoT-solutions. Kapernikov will be responsible for the image recognition based on artificial intelligence.

Our ambition is to have a fully functional prototype of this new sensor ready by the first half of 2022. While we are sure we’ll face many challenges, we are confident the new sensor will be a major improvement compared to the current system. You can at least expect:

  1. Robustness of the software & electronics: far fewer technical outages, more stable data transmission to the server. No more manual resets and inexplicable outages!
  2. more robust and elegant hardware design: a better solution for fixing to the window and a better solution for the camera view. We know the current Raspberry-Pi based sensor doesn’t have the most elegant casing and the camera mount isn’t ideal. Bagaar’s expertise in product design will ensure a smooth and elegant design for the sensor that will not bother you to mount to your bedroom window!
  3. an easier and fail-safe installation procedure: Was it a hassle for you to install your Telraam sensor and link it to your home wifi? Join the club… with this new sensor, we want to drastically simplify the setup procedure and avoid linking to your wifi altogether. Just take it out of the (beautiful) box, mount to the window and off you go!
  4. increased accuracy of the traffic counting data: While we’re fairly satisfied with accuracy of most devices (+/ 85% for cars, for most of the devices), we know the current Raspberry Pi-based Telraam-sensor struggles in some locations and has a hard time distinguishing between cars and large vehicles or accurately detecting bikes and pedestrians. We will completely overhaul the detection algorithm and adopt artificial intelligence (AI) for the detection. Kapernikov’s feasibility study has shown that with the latest technology, we can achieve privacy-preserving AI computing directly on the sensor (“AI on the edge”), so there’s no need to transmit images to the database, just the counting data.

If feasible, we’ll develop a sensor that is capable of traffic counting in dark conditions and we’ll develop a battery/solar-powered system, without a need for external electricity supply. We expect it will be challenging to achieve these goals, but aim high!

Now, how can you help us in this new project?
  • Have your say in what should be the priority features of the new sensor. We want to know what’s important for you and will align the development, insofar technically feasible. Check Telraam’s Kampsite on the new sensor and provide your feedback.
  • We’ll need lots of video-material to train the new AI-based sensor. For that we also need help! Find out here how you help us training the new sensor by supplying video’s.