There are a few ways you might get feedback from the kite, without using vision.
1) GPS. Don't rule this out simply because the kite's base might be in a different location each time you use it. To counter this, you simply need a GPS receiver on both the kite and the base. Take both readings, and convert them into Cartesian coordinates. The difference between these coordinates is the offset of the kite from the base. Note that while GPS offers fairly low accuracy, its resolution is better than its accuracy. The kite and the base will both have the same position error, and so the kite's location relative to the base can be (reasonably) accurately calculated.
2) IMU. A 9-axis Inertial Measurement Unit, will help. There are some pretty small, lightweight parts available, like the MPU-9150 from Invensense. If you use a Kalman Filter to combine the GPS reading with the IMU reading, you can hugely increase your position resolution.
(Sorry, that's a similar part, but the same size)
3) Load cells. The reason I asked how easily you could fly a kite with your eyes shut, is because this will give you some idea of how useful the information from Load Cells will be. Perhaps you wouldn't be able to fly the kite brilliantly, but I bet you could keep it in the air. Each kite string will need a 3-axis force measurement so that you know the exact force and angle of each string. Arrange the three load cells like this:
Each one has a string coming perpendicularly out of the top, in the direction of the load cell's maximum sensitivity. Tie the three strings together, and to the main kite string. The sum of the three measured force vectors will be the tension and direction of that kite string. Make sure that the angle between the strings is greater than the maximum angle of the kite string, otherwise one of the three strings may go slack, making for a false measurement.
4) Pressure. I'm sure you could learn a lot by measuring the air pressure at several points on the kite's inside surface. There are some tiny lightweight barometric sensors available, like the BMP085.
It's even available from Sparkfun on a breakout board. These sensors will also help you to measure the kite's altitude (if they're out of the wind), and you can even use them to measure the windspeed if you place two inside a pitot tube.
Four sensors and three Pitot tubes, plus one sensor on the ground, will give you wind speed and direction, and altitude.
Doesn't the sensor have a separated receiver?
As you said it could be that the received signal is highly attenuated and the SNR is so low you cant see it. One way to find out if there is a copy of the pulse inside the data is using a math tool called auto-correlation, but I doubt that a simple ultrasonic parking sensor uses it.
To have a better idea you can try to track the PCB and find out if you can see a filtered signal.
Best Answer
At 33kHz, the emitted wavelength is:
"Nearest distance is 0.4in (1cm)"
The wavelength of one pulse is the same order magnitude as the detection distance.
The ultrasonic sensors I have seen, where I found adequate data, seem to take 2-3 cycles to get to full oscillation, and a similar number of cycles to stop (IIRC a little longer to stop than start). IIRC the active pulse duration was several cycles, so 8-11 cycles in total.
I can't find the link, and I can't remember if the 'stop emitting' was being actively driven, or it was just switching off the power. So you might be able to do a little better by driving the device more effectively.
Purely based on that, I'd expect the minimum detection distance would need to be slightly longer than the length of the pulses round trip journey to make it straightforward to do.
You might be able to characterise the system in such a way that it can detect reflected sound during the 'stop' phase, maybe halving the minimum distance. It might even be better to emit for longer periods, though you might then be unlucky in the interaction of emitter and room reflections.
I believe some bats emit at a wide range of frequencies to help detect target type and distance.
If the 1cm minimum detection distance is a key requirement, maybe look at higher frequency devices or multiple devices. The higher frequency emitters I looked at had much narrower 'beam angles' than lower frequency devices though. So that might introduce a different constraint.