I suspect the only way to achieve what you want is by using a camera with image recognition software which is far beyond the capability of an Arduino.
An ESP-32 cam can do face recognition so it may be powerful enough but I suspect it would be easier to do this sort of project with a laptop and all the powerful software that is available for PCs
The fastest refresh rate on any GPS I've owned was 10 Hz. On a good day, you might be able to get lat & long to within six feet, but altitude is notoriously inaccurate.
So, I don't think your project is very practical, at least if you want close up video of your aircraft. As Robin2 says, perhaps image recognition is the way to go, but that's not really an Arduino thing. You might use a GPS to give the software a starting point for where to point the camera though.
There are antenna pointing systems to increase the comms range of autonomous aerial vehicles (UAVs) what some call drones that could probably be adapted to point cameras. The accuracy required for pointing an antenna is probably less than that needed for video, but the antenna pointing systems might be a good starting point.
The antenna pointing systems use the location of the UAV broadcast back to the base station via the telemetry system. The update rate of the GPS could be less than 30 Hz and still be used as Kalman filters allow for different sampling rates for the different sensors.
How much are looking to spend? I ask because you could use an real time kinematic (RTK) GPS system to obtain a much higher fidelity location. Check Sparkfun if you are interested in RTK as they have some affordable systems.
A hybrid system using the location-based pointing system to get the plane within the field of view of the imager and then use video processing to get the plane in the center of the FOV. One possible problem with an image-based system is the plane will look very different depending on the relative orientation that the imager sees. Using artificial intelligence might get you around the orientation issue at the expense of additional computational horsepower needed. As stated by others a standard Arduino does not have the horsepower to do video processing.
I would probably start with a UAV antenna pointing system and see if that gives you the required accuracy and implement RTK if it does not, this is assuming you want a robust solution and are willing to pay for it. One benefit of the UAV solution is an autopilot could be used on the plane just to provide the comms and location information with a base station on the ground. The autopilot would not have any connectivity to the planes controls. The software for the autopilot, base station and pointing system is all open source.
wwbrown:
How much are looking to spend? I ask because you could use an real time kinematic (RTK) GPS system to obtain a much higher fidelity location. Check Sparkfun if you are interested in RTK as they have some affordable systems.
I'm happy with a budget of $5,000 +/- a few thousand.
A hybrid system using the location-based pointing system to get the plane within the field of view of the imager and then use video processing to get the plane in the center of the FOV. One possible problem with an image-based system is the plane will look very different depending on the relative orientation that the imager sees. Using artificial intelligence might get you around the orientation issue at the expense of additional computational horsepower needed. As stated by others a standard Arduino does not have the horsepower to do video processing.
I have a lot of opencv experience and have written some software that already does pretty much what you describe, but it takes a lot of computing power to do it in real-time and is unfortunately not practical. I need something that I can pack up with me and take in my airplane which has very limited baggage space.
I would probably start with a UAV antenna pointing system and see if that gives you the required accuracy and implement RTK if it does not, this is assuming you want a robust solution and are willing to pay for it. One benefit of the UAV solution is an autopilot could be used on the plane just to provide the comms and location information with a base station on the ground. The autopilot would not have any connectivity to the planes controls. The software for the autopilot, base station and pointing system is all open source.
Maybe this helps,
wade
It all helps a lot because I had somehow missed that sparkfun had an RTK solution that seems reasonable.
But one thing I still need help with is the radio to communicate the position information to the base station.
But one thing I still need help with is the radio to communicate the position information to the base station.
The u-blox c94-m8p is a self contained RTK-GPS setup, with a built in radio link that can be used to report the position of the rover, relative to the base station. It is a bargain.
jtbourke:
I'm happy with a budget of $5,000 +/- a few thousand.
I have a lot of opencv experience and have written some software that already does pretty much what you describe, but it takes a lot of computing power to do it in real-time and is unfortunately not practical. I need something that I can pack up with me and take in my airplane which has very limited baggage space.
It all helps a lot because I had somehow missed that sparkfun had an RTK solution that seems reasonable.
But one thing I still need help with is the radio to communicate the position information to the base station.
Thank you!
Jim
Radio connectivity is a normal part of the autopilot as the it broadcasts at the minimum telemetry to the base station and the ground station transmits instructions to the AP in the air.
The heavy openCV computations will be done on the ground as that is where the pointer is, so you room and power to support a GPU or two if need be.
jremington:
The u-blox c94-m8p is a self contained RTK-GPS setup, with a built in radio link that can be used to report the position of the rover, relative to the base station. It is a bargain.
I had not seen the u-blox RTK solution you posted, I would be hard pressed to go with Sparkfun's product as you might as well go the King of GPS, ublox, at that price point.
In the performance review I linked, the author mentions that the radio link works perfectly at remarkably long range, at least 13 km line of site. It seems that the RTK solution is occasionally lost, though.
I've not tried any range tests like that, but in my short range tests, the performance of the setup is awe-inspiring.
