Since I was an intern at Image Sensing Systems working with the Autoscope in 1994, I thought our industry should be doing turning movement counts with computers instead of people. That’s why I was excited to be an early adopter of Miovision who promised automated turning movement counts. After working with Miovision for half a year, I figured out (1) they deliver accurate data, although we don’t know it’s truly automated and (2) they ate up my profit margin.
So after 16+ years of daydreaming about portable camera systems (good things take time and yes I should get a life), I built my own little camera system, COUNTcam, which has now evolved to the more refined COUNTcam 2. This is a reliable system that can be setup in about ten minutes at any intersection and record for up to five straight days!
(Thanks to Kimley-Horn who took the photo of our original system in action in Minneapolis way back in fall 2010).
So what about this post’s headline? Right around the time we were developing our first COUNTcam, TranSmart completed a corridor study of Highway 52 for MnDOT south of Rochester, MN. The initial scope called for about 50 tube counts and a.m./p.m. counts at 17 intersections. I brought up the idea of doing our 13 hour counts instead, which I thought would get them good data for their corridor study (giving data for peak hour analyses as well as traffic control warrant analyses) and we’d be able to do it significantly cheaper than the initially requested data.
MnDOT liked the idea of the turning movement counts, but 48 hours of data is their “standard.” We settled on a fair price for the 48 hour turning movement counts. Deploying people to do the same counts would be radically expensive, which is why they were originally thinking mechanical tube counts even though the data wouldn’t be as useful.
Our testing on the original COUNTcam showed we should have no problem recording 48 hours so of course I shot from the hip and said we could do it. Even though we hadn’t recorded anything longer than 24 hours for a project up until that point. Happily, it worked!
So then we went through a lot of hours of turning movement counts, which was made easier with the use of COUNTpro and COUNTpad 2, which allow us to toggle the video faster/slower while accurately counting (it still takes a person to do the counting, but this is a huge step forward from doing field counts with a jamar board).
These tools allowed us to accurately complete these counts at more than 5x real speed (the highway has 6000 cars a day and some of the cross streets have 300 cars a day – which would have been painfully slow at real speed).
2021 Update: A lot has changed since I wrote this post. 48-hour turning movement counts are commonplace and we even do seven-day turning movement counts. We also offer our countCLOUD processing service where we’ll turn your videos into turning movement count data. Processing for a 48 hour turning movement count is less than $350. Here’s how countCLOUD works and where to purchase.
I’m a traffic engineer in British Columbia for the Ministry of Transportation. I have been following your blog for over a year and I am very interested in your newly developed product. Can you put me on a list of some sort that would ensure I am kept in the loop?
I would love to try this out in my district and, if successful, sell the idea to my collegues around the province.
I would be interested in your video product for tmc’s. Please advise of your product cost and availablity. Curious when processing data in fast mode how do you record data for 15 minute intervals? MGM
This makes me think of where counts from signal controllers and existing detection fit into the picture. They can be used to spot trends in the short term, between iterations of full-blown turning movement counts using specialized equipment and/or people. Signal controller counts give us daily feedback at the expense of accuracy (due to shared lanes, long loops, etc.), and may not even be possible without adequate existing detection and central systems, versus the higher accuracy of specialized observations that, due to their cost, can only be done with a frequency that is on the order of years.
Just purchased a counter for BCMoT! Thanks Mike for the excellent customer service. I can’t wait to get the camera up and running….
Can I please have some details regarding your little camera and how it works? My team is in your situation of 1994 and wish we could get some things that can do thing with ‘ease’.
In particular I would like to know:
-whether it classify vehicles
-What Set up time would be
-What form of data it collects
-Battery life if it uses battery
-How it determine turning traffic
-Price of course
just to mention a few.
Thanks for reading my blog. Our system is still manual based (video the intersections and then do the counting in the office). However, our manual counting system allows the technician to toggle the video speed up and down. This results in significant time savings versus deploying technicians in the field.
The place to start would be our website. There are several videos explaining how components of the system work. Here is our Frequently Asked Questions page – http://www.countingcars.com/category_s/74.htm
Here are responses to your specific questions:
• Four classes can be counted.
• About a ten minute setup time the first time, down to about five minutes for a regular user – faster than putting out a pneumatic tube counter.
• Video is recorded in .avi files. Count files come out in excel files or jamar’s petra pro.
• Our COUNTcam 40 can record for 40 straight hours and our COUNTcam 120 can record for 120 straight hours.
• Pricing – it’s all on the website. We recommend starting with one of the COUNTkits. The COUNTkit 40 is our most popular – http://www.countingcars.com/product_p/7003.htm
After looking over the website, let me know what questions you still have. Cheers!
Surely freelance data entry people are already being hired over the Internet to reduce video to counts, for extremely low wages.