The Center for Transportation Studies at the University of Minnesota has released a report that explores approaches to distinguish bicycles from pedestrians for traffic counting applications.

The primary goal of the study is to develop a practical vision-based bicycle counting system that is capable of automatically processing video stream data of traffic scenes involving bicycle and pedestrian activity and estimating their traffic counts.

The authors argue that, once deployed, vision-based counting systems can potentially be the least expensive and easiest method to do counting. This is due to the fact that cameras themselves are relatively inexpensive and easy to use.  In addition, the authors believe that in many places communities will be able to leverage the already existent security cameras, which can further reduce expenses.  Algorithms that perform image analysis and object classification are presented and discussed