Infrared cameras for detecting people inside cars

  1. For an experiment we are doing, we need to be able to accurately detect people inside a car moving at high speed. Any suggestions for an IR/other camera that can help us effectively do it? The car has a hot engine, so I am concerned an IR camera may or may not work. Any thoughts.

    Cost not a big concern though would like it to be <$2K if possible.
  2. jcsd
  3. Mech_Engineer

    Mech_Engineer 2,347
    Science Advisor
    Gold Member

    A long wave infrared camera wouldn't work because glass is opaque that far into the infrared spectrum... My guess would be a camera system in the near-infrared spectrum (night vision rather than thermal) that can recognize a head shape.
    Last edited: Mar 5, 2010
  4. berkeman

    Staff: Mentor

    Too funny -- I started laughing until I realized that you want to detect how many people there are in the moving car. I was thinking, "of course there are people in moving cars!" :rofl:

    What angles do you have available? Just from the sides, or can you put a camera facing the oncoming cars on an overcrossing support?

    Welcome to the PF, BTW.
  5. berkeman

    Staff: Mentor

    BTW, what is the application? Something like catching carpool cheaters or something?
  6. Attached Files:

  7. Hahaha, you are right. I should have been more careful with my words :-)

    Yes, it is for a carpool violation application. Yes, we can put camera facing the cars on overcrossing support. Thanks for any ideas.
  8. I saw a video a while back about removing the IR filter from a DSLR and replacing it with a different filter will make the camera take IR photographs, but I am not to sure on this claim.
  9. berkeman

    Staff: Mentor

    It's common for digital cameras and camcorders to be able to image the near infrared. Just look at your TV IR remote control in your cell phone camera or other digital camera..

    The IR filter that you mention was likely to block near IR from the imaging CCD array of the DSLR. Removing it (or replacint it with an IR transparent filter) would let it image in the near IR.
  10. Mech_Engineer

    Mech_Engineer 2,347
    Science Advisor
    Gold Member

    There is a niche of digital photographers that like to take infrared photos. From the factory, the camera comes with a filter that filters NIR and transmits visible. The photographers remove the filter and replace it with one that transmits NIR, and blocks visible so they can take a pure NIR picture.

    Digital SLR's have actually created a boom in this hobby, especially ones that have a live view option because you can see in real-time what the infrared picture will look like helping with focusing and exposure.
  11. berkeman

    Staff: Mentor

    Cool info, ME. I didn't know that.
  12. Removing the IR cutoff filter from a camera will allow you to detect light out to abut 1 um wavelength. This is due to the bandgap of silicon so you can't do much to change it.

    For 1 um to 1.7 um or up to about 2.5 um InGaAs detector based cameras are commonly used. In this range water absorption makes people appear dark skinned so people look the same regardless of their skin color in the visible. You won't see a thermal response unless something is very hot, I don't know off hand but I would guess over 300 C, certainly not heat from a person.

    For 3-5um or 8-12 um (mid or long wave infrared) the lowest cost would be a microbolometer based camera. You would need to see if the thermal response time was adequate for your speeds. If not, there are compound semiconductor materials (InSb, HgCdTe, etc.) in that range which can be used in high speed detectors.
  13. Thanks davidrit. It certainly looks like I need someone who understands physics of light to get involved. I am an electrical engineer, so I can see some angles of what you talk about, but I sure need to get some help to build this camera.
  14. mgb_phys

    mgb_phys 8,809
    Science Advisor
    Homework Helper

    It might be practical if you can point the camera through the windscreen.
    A lot depends on the environment - for a good signal you need a biog temperature difference between the person and background. So would work great in Canada in winter, or in the US where everyone has the AC on near-freezing, wouldn't work if the ambient temperature is near 37C

    Uncooled systems (microbolometers/ ferro-electric) can generally do about 100mK NETD at video frame rates - affordable ones are cheap-webcam resolution so you would need one camera/lane and would only get a couple of blobs as images of the passengers

    Of course this wouldn't work for a person in the backseat unless you had multiple cameras - even then with something like an SUV it would be hard to spot other occupants (this is why Police are very nervous, and trigger happy, when stopping large SUVs with blacked out rear windows)
  15. Mech_Engineer

    Mech_Engineer 2,347
    Science Advisor
    Gold Member

    You can make most digital sensors sensitive to near IR (800-900nm) but this wouldn't give you thermal data, so you would probably need an infrared illuminator and some software to detect a head shape.

    All of this is moot, because mid and long wave IR (3-14um) doesn't pass through glass!!! Indeed, most automotive glass is made to reflect these thermal infrared regions to help insulate the vehicle. You would only get a thermal signature of the windshield (or perhaps a reflection since it will likely act like a mirror), not the people behind it. A thermal camera is useless for trying to see inside a car.
  16. Mech Engineer, no need to shout. I didn't disagree with you I merely offered a 30 second breakdown of the infrared into a few commonly used spectra to help increase the OP's understanding of detecting infrared. Judging from the OP's response he understood that even if you didn't.
  17. Having read with interest the posts, is the project impractical?
  18. Mech_Engineer

    Mech_Engineer 2,347
    Science Advisor
    Gold Member

    The project is difficult but not impossible. It's doubtful you'll be able to detect 100% of people in vehicles, but you can probably see if there are ones in the front seats using a night vision camera in the near-infrared spectrum.
  19. Unfortunately that is not good news I guess. We would be OK with being pessimistic by accounting for people when they are not actually in the car and hence missing out on some potential carpool violators. But missing out people in back seats would be pretty bad, since that would open up problems that commuters would complain that they actually had a person sleeping in the back seat etc, but were still tagged.

    I wonder if there are more expensive non IR options...
  20. I have watched this post and have a single observation.
    This gets dangerously close to violating personal rights, and subverting criminal law
    If I get a ticket, based upon false or true data, I still do not get to go to court and
    "face my accuser" as is required by law
    there is no accuser
    Its my opinion that if you want to monitor a persons activity, you are assuming all persons are guilty just by monitoring them
    stop light cameras, auto speed detectors are in this catagory.

    if its such a problem, hire A PERSON (which would provide a job) and have them do the work, not a mindless automated machine
    sorry, end of rant

Know someone interested in this topic? Share this thead via email, Google+, Twitter, or Facebook

Have something to add?
Similar discussions for: Infrared cameras for detecting people inside cars