Hello everyone,
thanksalot for Telraam and this community! My sensor/system is now running for around four weeks now and it finally noticed some heavy vehicles, which is great!
Bikes and two-wheelers still have the largest share in my Telraam’s data which doesn’t match my subjective perception at this time of year (e.g. when there really large crowds of people wandering around on Christmas, it merely matched the daily average of counts).
Is there a chance, the device is counting pedestrians or cars as two-wheelers? If so, is there a chance this will change? Thanksalot for any help!
First of all: Welcome to Telraam Talks! Secondly: Good question, I’m sure @Telraam can answer your question in more detail, but the post below might already give you an idea of potential inaccuracies in Telraam counts.
Good question! I’ll try and explain it how Telraam counts and classifies the counted objects as pedestrian / cyclist / car or heavy vehicle.
The initial classification happens between pedestrians vs. two-wheelers vs; vehicles (=cars + heavy vehicles) based on two parameters: fullness value and axis ratio Fullness means how well this red shaded area fills up the space inside the black rectangle.
Axisratio is the ratio of the two arrows.
After this, there is a classification between cars and heavy vehicles within the class “vehicles”. This is done based on a cut-off value. However, a calibration (period) is needed so that this cut-off value is optimally calculated for your location.
So your conclusion (larger groups pedestrians or two-wheelers) could be right. If there is a large group of pedestrians walking by and the fullness and axisration looks more like the parameters of a vehicle, it will be counted as a vehicle. Same thing for larger two-wheelers or groups of two-wheelers.
The following articles on our FAQ page explain the classification more in detail and if your interested in knowing more in detail how the classification works (and when / why there could be misclassification) I suggest taking a look at this one: How are the different types of vehicles classified and differentiated?
Does this help understanding why your counting data are what they are?
Unfortunately the counting algorithm will stay the way it is and errors like this won’t be solved. We are thinking however of making something so Telraam owners can put remarks on their dashboard (for certain days) to explain some numbers, so other people who are looking at the data know to interpret (certain) data with the right background information.
We have been comparing video feed to Telraam counts since the installation 2 months ago and Bike traffic is significantly off. We have countless videos with counts and comparisons to Telraam reports, and have posted the videos to a Google drive for access to Telraam. We’ve requested help from our contact, and have even suggested we think the speed of bikes might be a factor. The response was if we want more accurate counts, we should spend thousands of dollars on a better system. Not quite what we were looking for in assistance. After following Telraam’s placement instructions, we were then told it might be too close to the road (we’re already at 15 degrees). We were getting reasonably accurate counts going the direction closest to the camera (uphill), but the other side of the road (downhill) was off by 50%+. The camera was placed where the most amount of car/bike/pedestrian traffic was. We have since moved the camera and it is now 5 meters from the road, 3 meters up, and on the opposite side of the road where the least amount of car and pedestrian traffic is, but still has view of 100% of bike traffic. Now the far side (uphill) is sometimes very accurate, but when it’s off, it’s significantly off (30%+). Downhill closest to the camera is off by 50%+ still. We suspect the speed of the bikes might be the issue (they can be up to 55 kph). We understand some undercounts due to groups of bikes but we are seeing the videos and accounting for them. We also understand if a bike and a car pass at the same time, it might miss the bike and we account for that also. Looking for any additional suggestions to help us with our bike counts. Here were our latest counts:
Secondly: Good question, I’m sure 