Cycle counting with a web cam

In February 2005, a Cambridge burglar was jailed based on evidence from a web cam. (http://masl.to/?N2771108A, www.grisby.org/burglar.html)

After an earlier burglary, the home owner had set up the camera to detect motion and record images directly onto his Internet site.

Idea

I discovered that the web cam I had recently purchased came with some bundled software that can detect motion. It gave me an idea. Back in 2003 I’d attended the international cycling conference, Velo-City, in Paris, and sat through a session on machines for automated cycle counts. (www.camcycle.org.uk/newsletters/51/article11.html)

There are a few of these machines in Cambridge, but the Campaign has found out that they are often very inaccurate. Normally traffic counting in Cambridge is done by humans, and so it’s expensive data. In particular the data for cycling is very sparse. So I decided to see if my web cam could count bikes.

Action

Using a web-cam to monitor movements in Thoday Street
Image as described adjacent

I set up the camera to look out and down on Thoday Street from my bedroom window. It took quite a while to get the camera positioned correctly and to handle the varying light conditions between inside and outside the house. (My neighbours must have thought I was spying on them or up to something weird.) Then I had to calibrate the motion sensor so that it would detect pedestrians, bikes and motor vehicles. When I got it right, I configured the computer to record five seconds of video whenever motion in the street was detected.

I chose a wet Friday in February, and over the day it recorded 1000 videos in 264MB of data. To process this data I had to sit through all these videos and note down what I saw. This took about two hours. There were some problems with the software and camera: it didn’t auto-adjust its exposure very well, images were very noisy at times and it seemed to be better at detecting rain than pedestrians.

I was mostly interested in how many cyclists cycle against the one-way system along the street. The results were (between 9:15 & 16:45):

  • Motor Traffic: 363
  • With-Flow Cycles: 46
  • Against-Flow Cycles: 14
  • Pedestrians: 94
  • There was also one car against the flow.

My camera could not see the pavement on my side of the street. It is probable that a few speeding cars and slow bikes were missed by the camera. But it was probably equally likely to count a cyclist whether they cycled with or against the flow.

Result

An important result emerged from this experiment: that approximately 25% of cyclists in the street go against the flow.

The city council is considering permitting two-way cycling in many of the one way streets in the Petersfield and Romsey Town areas. This sort of data could help in making the case for two-way cycling in your street. Over the summer I should like to do some more counts like this around the city. I’d be interested in hearing from any members who could help, either by providing a recording location or with equipment. In particular, a fully mobile unit could be made using a laptop with a long-life battery. If you have some spare equipment like that, or any other ideas, I’d be delighted to hear from you via the Campaign contact addresses.

Simon Nuttall