I was trying to find a good starting point, and was hoping anyone here could help. I was looking for a sensor setup that would allow me to track an object, without using a visual camera (like the color tracking or CMU camera systems out there).
I need a good amount of range, maybe 20’ - 50’. Ultimately, I would like to be able to achieve the same sort of tracking that the CMUCam demonstration video’s, just without the usage of a camera itself. Video: youtube.com/watch?v=GDOC9yZ7IHc
Could this be done with some sort of IR technology, RFID, Bluetooth, transceivers, sonor, etc? GPS cannot be used in this application.
I 'm sorry, to have object tracking you need vission which requires a camera.
IR detection can provide object avoidance but not tracking.
RFID will not do either because RFID stands for Radio Frequency IDentification, like the one use for toll road tags. In Texas, they are called EZ-Tags.
I think you’ll need to move the object pretty slowly for this to work. However, you could use two IR sensors and check which one saw the object the last time. Then you’ll just have to pan the servo in that direction until both sensors sees it again.
You could probably solve the vertical plane by applying a matrix of IR sensors, but I’m absolutely not sure.
The link to that page discusses edge detection and navigation. Tracking requires real time monitoring in bothe the X and y planes simultaneously. I’m not going to say that it is impossible to use IR in a clever manor to get some sort of tracking but its not practical. If it sees an object one second and then it moves somewhere else the sensor has to rescan for it.
For a project at school I had to make a “tracking” system that would allow a motorized cart to follow a person around.
I decided to use an infrared emitter which had to be buckled to the person’s belt, so the device had a source to track.
Although this isn’t exactly what you’re looking for, it would provide a simple tracking system, provided you’re willing to attach an emitter to the item you want to track.
It can’t be called object tracking because it isn’t object tracking. It’s edge detection and only works in the plane horizontal to the robot. If the object is above or below the robot, it will not be seen.
Object tracking works in 3D, not just 2D.
Not using the correct terms to describe a process causes confusion, which is not good.
I must admit that video was impressive. I have not seen tracking using edge detection and I was surprised at the speed at which it was able to do it. I would like to see the code to that.
I am able to put an emmiter source on the object, for tracking.
The tracking would need to be both on a horizontal and vertical planes.
Based on most of the suggestions, it seems like a visual camera source is not only the ideal solution, but quite possibly the most practical one as well?
I originally did not want to use a visual device due to some of the costs and room we have to work with, but sometimes there is no viable alternative. With that in mind, is the CMU Cam the suggested item for such application? Anyone know what the range capability of the CMU Cam would be off the top of their head?
How you track an object depends on the properties of the object being tracked, and the surrounding environment. If the object is capable of having some sort of beacon on it, then there are numerous approaches. A siimple one would be to use an IR object detection setup where the IR emitter is used as the beacon on the object being tracked, and the IR receiver is used as the object detector. The detector field of vision could be physically limited such that a servo can pan/tilt the detector until the beacon is detected and its location established.
if you can have an emitter on the tracked object then you could use an IR beacon and a iris + pin-diode array similar in concept to how heat seeker missles work. the pin-diodes are arranged in a geometric pattern with a lens and iris above them. the basic concept, the iris is wide open until it sees enough signal to track on one or more of the pin-diodes. it uses the strongest signal to “steer” the nose in an attempt to center the signal as it closes the iris to narrow the window and increase accuracy. if it looses the signal the iris opens and the logic attempts to re-acquire. wet hair, lather, rinse, repeat. In this case I would think the signal would be a modulated IR beam, say 38 to 40KHz and the pin-diodes would have a programmable gain amplifier and a real sharp bandpass filter, followed by a rectifier stage so you can read the signal in on an A/D converter. this way you could reject other “noise” and allow you to follow only the strongest source. You could use a TV remote as your “beacon” to track. you would need a lens, IR filter, and auto-iris (and its driver circuit) of course but those can be had from places like supercircuits in a standard C or CS mount. You could even use the cheapest B&W camera that will mount the lens/filter/iris as your platform to develop with and then replace the CCD with your diode array.
I think this would seriously freak the dog out… persistently follow it around, then stay a certain distance away, backing up if he charges… oh my head. If you could couple the IR range detector with a PIR sensor you could determine a warm target from the general environment.
I was planning to use a thermal sensor array for a similar application, but I was hoping for maybe a 5’ range. The range you are describing seems crazy far unless the trackee is going to be broadcasting its position.
This post is along the lines of what I am looking for.
For my use, I would like to have some sort of beacon to place on a stationary or moving object (house, car) to be able to track for aerial photography and videography. It would control a gimble (pan/tilt) on an RC airplane or helicopter as an example.
The beacon could be an LED with a CMUCam tracking it, but I’m not sure that an RC aircraft could carry a CMUCam and supporting electronics to communicate with it and the two servos for pan and tilt. The CMUCam offers the pan/tilt capability so you’d just need a pair of servo’s for the gimble.
I’m not sure if this would help, but Servo Magazine had an article on a webcam based laser range finder that worked by having the webcam look for the spot generated by the laser pointer and using that to calculate distance. I know it’s not distance detection that you’re doing, but the article might have something useful in it for tracking a beacon since they are basically tracking a projected beacon.
Here’s the deal:
You need x-y tracking and some sort of edge (or other way to find scale) if you want z. The reason cameras are so good is because they ARE a series of x-y detectors (and software can do scaling to give you z if you need it.)
For all-out crazy tracking, the best thing I could think would be to put a series of high-intensity IR LEDs on your target (like a triangle or arrow or something that your software could “look” for. If it’s asymmetric you could also use it for heading) and stack an IR pass filter on a camera for tracking (you’d also have to remove the IR blocking filter that comes in almost all cameras)
If you get too much noise you could look at subtracting two sequential images (one with the tracker on and one with it off) but that would require VERY precise timing or RF telemetry between the two objects and some motion correction for your images. Also, if you can find a color that isn’t in your surroundings you could do a color-pass filter on your set color to get rid of everything else.
Once you remove most of the image you can then search the remaining points for a shape that matches the LEDs you put on your target.
(I worked with some guys who were doing this kind of stuff for the FAA to land planes etc so I know it’s possible but not exactly… easy)
Given the processor power you’re going to want (remember that resolution goes up by a squared factor! The more accurate the more processor you’re going to need) I’d suggest getting a wireless camera and broadcasting back to a base station if you can.
The brighter your target’s emitters the easier this is going to be.
Good luck
~Jay
PS. Depending on the resolution you need you could mount a GPS module on your target and broadcast it to your tracker. Couple hundred bucks and you’d just get back coordinates… depends on what you want I guess…
Check out sparkfun(dot)com if you’re looking to go GPS.
