WindReader/balloon following device


First of all, what I'm asking here is not trivial, I know. If it was, someone would have invented it before. Im looking for ideas to solve the problem.

The story: What I actually want is to measure wind on different levels, up to a few thousand feet. At the moment, we do this with a so called WindReader from the WindsWay (or so) company. This is a device mounted on a tripod with a monocular. We fill a balloon with a normed amount of helium and let it go. We follow this ballon by moving the monoculars and the device loggs this motion and calculates (actually a software on the PC) the wind directions and speeds. If it is early in the morning, we add a LED-Throwie (small LED and battery, we do not even try to retrieve them) to the balloon. Unfortunately, those devices cost a hell lot of money (I heard something about 10k) and only a few of them exist.
I was thinking about ways to measure winds, but sure is, we need to get something up there. I thought about drones or something similar, but something expensive always needs to be retrieved, and we're such in a hurry that there is no way waiting for the drone to come back. This also excludes a GPS attached to the balloon. So I'm back to this WindReader concept.
The known WindReader is quite a "simple" thing to rebuild. Mount monoculars on a two axis device, attach wheels to turn these axes and some gears, measure the time and numbers of turns and make the calculations. Let's forget about the calculations part for now, I do know this will be complicated, but not now ;).
I am now thinking about something which does this fully automatically. And here I need your help. My ideas for now are: Add an IR-LED to the balloon and follow this with a sensor. Problems here will be that I guess there is no chance following this over some thousand feet and the sun also emits IR-rays which will overshine the LED during day. Another idea would be with a camera and object recognition (and a zooming lens of course), but I know this is beyond Arduino possibilities.

Does anyone have another idea? Or could help me find the right corner for the camera recognition part?

Every input is appreciated!

Some kind of optical corner reflector on the balloon, and a powerful modulated IR source on the ground?

PI might be able to handle the video tracking if the source is strong enough or simply tracking the entire balloon. Arduino doesn't have the mental capacity to process videos.

OpenCV will work on the Pi; I saw one example where someone did face tracking at 5 frames/second which should be OK. It's pretty involved stuff but if you're just trying to identify a single-color balloon it shouldn't be that hard. There's also reacTIVision which is kinda plug and play.

Does the tracking work by predicting the vertical speed of the balloon based on buoyancy/drag? I'd have guessed that was pretty inaccurate but presumably it works well enough for what you need.

I would have thought the tracking camera / object recognition approach would be your best bet, but you've already pointed out the obstacles that would need to be overcome. I suggest that optical tracking is probably not viable if the sun is against you.

Since you haven't mentioned any constraints for the ground station, I'd have thought a laptop would be the natural choice.

The camera and mechanism used to zoom/focus/steer it seems fundamental to the working range of this project, but I'd have thought you could prove the concept fairly cheaply using a DIY mount controlled by an Arduino. Making something accurate enough to track a balloon up to thousands of feet will be pretty challenging - perhaps that explains the price tag of the commercial system. Can you afford to illuminate the balloon - do you have the resources, is it safe and legal to project that much focused light/heat? I don't know what laws are applicable, but it seems to me that a high powered emitter could be construed as a potential weapon - people have already been sent to prison for shining lasers at aircraft.

Does the tracking work by predicting the vertical speed of the balloon based on buoyancy/drag? I'd have guessed that was pretty inaccurate but presumably it works well enough for what you need.

Our commercial solution does it with calculating the level, depending on the time since launch. As I said, the width of the balloon is normed, meanind that we put a measuring cord around it while filling. To be honest, there is no way to proof this calculation, but it seems that it is accurate enough.

For the moment, I'll try it the following way: The device will get an Arduino which controls the motors/steppers, which align a webcam. This webcam is connected to the PC (laptop), where a processing script analyses the picture and sends the corrections needed to the Arduino.

I'll tell you when there's progress to this!

It would be possible to make this much more accurate by providing two sensors far enough distance apart that you can use triangulation to calculate the position.