Shop OBEX P1 Docs P2 Docs Learn Events
PIR for warm body tracking — Parallax Forums

PIR for warm body tracking

jaegjaeg Posts: 156
edited 2012-12-02 08:47 in Accessories
Would a PIR sensor (one without a fresnel lens and support circuitry I.E. the one offered by Parallax rather one of the ones at the top here: http://www.futurlec.com/PIR_Sensors.shtml) be suitable for an attempt at warm body tracking? I would like my robot to know that it is pointed at someone and have a rough idea as to what side the person is on. From what I understand it gives an analog signal.

Any caveats to this idea?

Comments

  • Mike GreenMike Green Posts: 23,101
    edited 2012-11-30 20:27
    The PIR sensor offered by Parallax has the Fresnel lens and support circuitry that outputs a digital signal and has a considerable time lag involved. The sensors at Futurlec don't come with either Fresnel lens or the support circuitry. You could have an analog output, but, with the time lag, it would be difficult to track a warm body. Much better would be to set up a fixed array that could indicate what quadrants in the sensor's view see a subject. If the subject is moving slowly enough, you could track it based on the active quadrants and the relative signal values. I suspect you would do much better with something like a laser tracker or an IR sensitive version of something like a CMUcam.
  • LeonLeon Posts: 7,620
    edited 2012-12-01 00:45
    When I worked for Irisys they used their technology (http://www.irisys.co.uk/) to develop a prototype thermal tracker, fitted to a motor-controlled camera, enabling it to track people and display a video image of them. If one stood close to it, it slewed to one's groin area, being the warmest part of the body!
  • Martin_HMartin_H Posts: 4,051
    edited 2012-12-01 08:34
    I did a POC using two PIR sensors to see if a robot could use them to find a moving person. It sort of worked, but as Mike Green said they are slow. Here's a video:

    The bottom line is that your target needs to cooperate. I bought an xband motion detector to see if I can build something a little bit faster.
  • Duane DegnDuane Degn Posts: 10,588
    edited 2012-12-01 10:16
    Have you seen Phil's project?

    It uses a more expensive sensor so it might not be an appealing option. I'm also not sure if the sensor is currently available in small quantities.
  • LeonLeon Posts: 7,620
    edited 2012-12-01 12:22
    The sensor wasn't available when I checked recently.
  • Tracy AllenTracy Allen Posts: 6,658
    edited 2012-12-01 17:09
    Jaeg,

    Those sensors from Futurlec feature "Dual Compensating Elements". That means that there are two thermal elements side by side in the package, but connected electrically in series opposition. What that means is that if both of them heat up (or cool down) at exactly the same time, there will be no electrical output. To get an electrical output, one or the other has to heat up first. That is provided by the IR lens, which focuses the IR energy. As a warm body moves across the path, the focal point sweeps across the elements and generates an output signal. Many different types of lenses are available, and often they have multiple facets, multiple focal points, so that the movement of the warm body causes multiple output pulses. Other lenses may have a single line focal point.

    Under ideal circumstances, it is possible to ascertain the direction of motion of the warm body by analyzing the signal. It is biphasic. That is, one direction of motion causes an upward deflection and then a downward deflection, vice versa for the opposite direction. With multiple sensors lined up it could could conceivably collect position information.

    It all depend on motion detection of static warm bodies. Either your bot would have to scan with a line lens, or the external body would have to move.
  • Tracy AllenTracy Allen Posts: 6,658
    edited 2012-12-01 17:18
    With regard to speed, the sensor elements respond to thermal changes on the order of 0.05 second. To make use of that speed, you have to process the raw analog blips. A number of years ago I built such a circuit to count bats flying through a portal. At first I was hoping it would work to discern the direction of motion using the biphasic signal, but the bats did not usually cooperate. They often made circling paths through the portal, or many individuals overlapping, and the signals did not look at all biphasic. So the detector became more an indication of general activity rather than a precise counter.
  • jaegjaeg Posts: 156
    edited 2012-12-02 08:47
    Thanks for the responses everyone. I may just use one PIR sensor so the robot while sitting still can know that something is in the room and eliminate the tracking feature for now.
Sign In or Register to comment.