The goal of my project is to create a camera system that can track a moving person. I've seen numerous related posts on the Arduino forum, but no completed projects, just ideas thrown around, and I would like to change that.
To begin with, I'll list the individual approaches I've come across:
Two GPS receivers, one on the person, the other on the camera, yielding the relative position of the person with respect to the camera. I've read that GPS relative positioning can yield decent accuracy, but I'm unsure whether GPS alone will yield high enough accuracy.
RF/Ultrasonic transmission: the person would have a RF/ultrasonic transmitter, and the camera would have multiple receivers. The camera would know how to align to the object due to variances in signal strength.
IMU 3d tracking. Essentially, through the use of gyroscopic/accelerometer measurements, the target's position could be tracked relative to its initial position.
Image based object tracking: I'm not really sure how this works. However, I am aware of two possible downsides. This method won't work if the camera loses view of the object. The arduino might not have the necessary processing power.
In addition to these individual approaches, it is possible to combine the data with a Kalman filter, or some similar method of sensor/data fusion.
While I have a basic understanding of these different methods, I am unsure of which method or combination of methods to pursue. I plan to play around with these different methods, but I would greatly appreciate the input and guidance of others.
RF/Ultrasonic transmission: the person would have a RF/ultrasonic transmitter, and the camera would have multiple receivers. The camera would know how to align to the object due to variances in signal strength.
Time-of-arrival or phase measurement might work, but signal strength probably will not.
SkiBum326:
The arduino might not have the necessary processing power.
It absolutely won't, by several orders of magnitude. To do real-time video processing (or any other form of image processing) you need a computer, not a microcontroller. The Arduino could be involved for steering the camera mount to track the target, but don't expect it to be involved in anything to do with image processing.
Maybe an IR beacon on your target like this.
Then a curved line of receivers that rotates around on a servo, turn the "nose" of your tracker towards the beacon when a signal is detected.
RF/Ultrasonic transmission: the person would have a RF/ultrasonic transmitter, and the camera would have multiple receivers. The camera would know how to align to the object due to variances in signal strength.
Time-of-arrival or phase measurement might work, but signal strength probably will not.
Before the GPS technology, people used to uss LORAN, a system to track ships with time-of-flight of RF signals to various towers and a master tower. You can do it with long distances, nautical miles, but meters of distance is too short. It only takes a bunch of nanoseconds to travel and the difference in time of flight is too short.
I've seen ultrasound tracking. Google "ben heck show episode with suitcase". He made a robot suitcase that follows him around. It's around 2 meters of distance but could work up to say 6 meters.
Nintendo wii remote control can also work. It gives you all IR objects (up to 6). so if the person is cooperating and always wears IR leds around his chest and back, the remote can tell where he is, up to a certain angle range. Still the range is a few meters and you need brighter IR leds for longer distance, say 10-20 meters.
The robot suitcase projects looks promising. In the end, I'd like this project to work up to 150 ft, but ultrasonic sensors are definitely a good start.
One thing I forgot to mention is that the transmitter (target that the camera is following) will be in someone's pocket.
Depending upon what material is between emitter and detector, wavelength of the IR, and power emitted,,,,, I have been able to detect IR emitter in a shirt pocket from 5 feet (do not have any idea what the max distance would be, did not test it) and UVb worked much better. However, is moot point. No way it'll work over 150 feet....... (sighs and wimpers)
I saw a neat trick on the Ben Heck Show... Ben made a suitcase robot that follows the owner. He used 3 ultrasonic ping sensors, but he disabled the ultrasonic transmitter on the 2 units on the suitcase. He used xbee modules on teh suitcase and on a pocket micro that the person it was supposed to follow wears. The person puts a little gadget in his back pocket that has an ultrasonic ping module facing to the rear. The suitcase triggers the pocket to "ping" wirelessly by the xbee modules, and then the logic on the micro in the suitcase times how long it takes the ping to reach each of the 2 sensors on the suitcase. It then calculates a PDI (differential) based on the timings so that the suitcase steers to follow the person. Was fairly impressive and accurate. They also modified the transmitter ping sensor by cutting the "can" shroud down to widen the beam, and they did the same on the receivers.
