Electronic people/tally counters

[GiTS]

Those of you who have worked retail may be familiar with the electronic counters stores install on doors to count the traffic entering and exiting the store.

I'm working on a project to improve the quality of the shuttle system at my university. I want to measure all the important aspects of the network ( how many people are going to where, when, and how long it is between nodes.) The shuttle system currently uses the transloc system to track the shuttle location realtime, although it does not take enough sample sizes atm, I am finding out if sample size can be upped. That system will give me the where, the when, and the how long travel is between nodes. But not the flow of people coming onthe shuttle and exiting. They currently measure the amount of people getting on the shuttle at various times and at various stops.

That is OK, but without knowing how many people get off it is hard to judge where people are going. I want to help solve this problem by installing electronic counters that can mesh with the transloc hardware or take accurate timing so that the number of people getting off at each stop and getting on at each stop can be measured. While this hardware doesn't nessecarily track where each person is going and coming from, rough estimates can be taken that should become more accurate with greater sample sizes.

If a directional people counter was built from scratch parts cost only (built for free by students), how much might it cost? It doesn't have to be from scratch, just something to meet our needs for 8 shuttles at less than $10,000.
If you spot errors in my methods or have other suggestions that would be helpful.

Thank you,
GiTS

 

[schip666!]

Your basic break-beam light and sensor on either side of the door thing should be pretty easy to build and install somewhere around torso height -- so you don't count legs instead of people. However, without forcing your subjects into single file you won't get good counts when multiple bodies overlap. Also, to distinguish bodies getting on from those getting off you would probably need two sensors spaced about a quarter body apart.

Other approaches might involve PIR (your basic automatic porch light) sensors to detect body presence, but I'm still not sure how you could distinguish individuals in a crowd and their directions. Further afield, video and image processing...but there you are burning up your $10K budget and will need some help from your Uni's EECS department...

The red-flag item is: "mesh with the transloc hardware"... I don't know what that is, but the "usual" issue with any commercial system is that it is not open to user fiddling. Do you know if your system has user configurable inputs, and if so, how you get data out of it?