On Thu, 28 Mar 2013 12:05:10 -0700 (PDT), Frank Krygowski
<
frkr...@gmail.com> wrote:
>> <
http://www.youtube.com/watch?v=ERJklRkh_pc>
>> <
http://www.youtube.com/watch?v=-DMrZz-LsM0>
>> <
http://www.traficon.com/information/news_detail.jsp?id=162>
>Those examples did look like they worked well.
Amazingly well. It would detect a bicyclist almost instantly when it
entered the detection zone (as marked). The trapezoid outline is
probably the valid detection window or area of interest, where the
system will look for input. I have the same feature on my motion
detecting security camera, where I can eliminate motion from areas
inhabited by critters (near the floor), or are visible through
windows. I played the video enough times to see how it works. It
looks for contrast, not brightness, and then checks if the contrast
area is moving.
>But locally, I've had
>cyclists tell me about a camera-controlled intersection where they
>were not detected.
The ones in the above videos are IR cameras that cost a small fortune.
They also have to be cooled (Peltier junction) and include a heater to
prevent condensation on the optics.
>And a well-connected cycling activist in a distant
>city was the one who told me about using a cutout to calibrate the
>camera.
If they were using a dummy, the camera system is using visible light,
and not IR. Unless the dummy was pre-heated to skin temperature, it
would not be detected by an IR system.
>Maybe different camera companies use different technology?
Yep. There's IR, IR plus Doppler radar, pattern matching, outline
matching, background subtraction, flying spot scanner (to cover the
entire intersection with one device), etc. I dug through Google
Patents and found more than I want to cite.
>I note
>that the "window" on your examples spanned the entire intersection. I
>like that; but ISTR seeing calibration images online from (apparently)
>a different camera vendor, which showed multiple and separate windows,
>one "looking at" the front of each lane.
My guess(tm) is that an IR system can more easily recognize objects
than a visible light system. It can therefore look at the entire
intersection without worrying about clutter from pedestrians, birds,
critters, wind blown debris, rain reflections, etc. However, the
visible light system need all the help they can get. So, they outline
the probable location of where vehicles and detected objects are
expected to be seen. If the bicycle isn't in the expected location,
it's going to be missed.
>I also note that the one you've pointed to uses infrared. I wonder if
>there are any using visible light instead?
Certainly there are visible light systems. Here's you basic speed
trap camera system:
<
http://www.google.com/patents?id=P1wLAAAAEBAJ>
See the references under:
<
http://www.google.com/patents/US5041828>
for far more examples. IR isn't very common because it's VERY
expensive.
<
http://www.bicyclinginfo.org/bikesafe/case_studies/casestudy.cfm?CS_NUM=707>
Loops are approximately $500 each; for bike detection, there
are typically two loops per direction of travel. Video detection
is approximately $35,000 for a complete intersection installation.
Pedestrian/bicyclist push buttons with the conduit and conductor
to the controller cabinet is approximately $1500; each pole with
push button is about $300.
That's $35,000 for 4 visible light cameras. My guess(tm) is the IR
cameras are about 2-3 times as expensive.
>In one discussion, a
>cyclist talked about opening his jacket wide to increase the size of
>his visible image, to aid in detection. Of course, that would also
>make him appear "hotter" to an infrared camera.
Nope. I've done some IR imaging. Again, it's the contrast between
the IR source and the background that makes it work. There's plenty
of contrast between a skin temperature of 37C and a background temp of
about 20 to 25C. However, some body heat leaking through a jacket
will be about the same temperature as the background. I just verified
this with my IR thermometer. My shirt temperature was only about 5C
warmer than the room background. Perhaps if the rider took off his
shirt, it might work, but not while wearing anything.
>I guess I should track down the various vendors and see how their
>technologies vary.
<
http://www.trafficcontrolcorp.com/manufacturers.html>
Not great, but a good start.