[aylons]

It's kind of the last option.

This kind of analysis is already done and important, but usually it takes installing a lot of sensors and giving an impulse to the structure. I have actually worked on a company that did this, using strain gauges.

The fact that this is one using a commercial camera is kind of impressive. I'll take a look on the papers someday, but the obvious way to do this would be to have a framerate at least the double of the frequencies being measured.

Being able to determine frequencies much higher than sample rate not only is non-trivial, but also alleviates immensely the cost of these measurements.