Components - Image Processing

[MUSIC PLAYING]

You might have seen the other film, and you know that I'm interested in tether dynamics. And part of that is flying kites and doing experiments. But in those experiments, I need to know where the tether is, and what it's doing over time.

And it's easy to see the tether from here. It's nice and black and fairly visible. But from way far away, so that we can see the entire tether, it gets a bit difficult to see what this line is doing. So we're going to need to do something about that in terms of image processing.

[MUSIC PLAYING]

I think we could probably get the whole kite. So we're going to set up the cameras in two directions, because we want to record the dynamics of what the kite string is actually doing. And from the pictures, we're going to try to find the tether in the sky.

So it's quite cloudy, so it will be quite difficult. But hopefully, with some clever algorithms and image processing, we can find out where the tether is and what it's doing over time.

So on the video camera, you might not actually be able to see the line, because it's lower resolution. But hopefully on the skills, we'll be able to work out sort of where the line is just based on averaging different directions over time.

[MUSIC PLAYING]

I'm going to connect this camera up with the other camera with this wire. And they're both connected to a trigger. And with the trigger, they'll both shoot at the same time. So we'll get one picture every second. And from that, hopefully, we can work out some dynamics. There's a limit to the frequency you can work out with one picture per second, but at this point, we're going to give it a try.

OK, so I can't take credit for a lot of the hard work that went into this project. There's a master's student, Richard Shaw, and he did a lot of work for his master's figuring out how you could find a tether on a background of clouds, or trees, or all of those kind of things. And he came up with a really clever method, in terms of image processing, to find this tether. And he runs around in the rain a lot.

It is really starting to rain. Like everything, you can't really control the weather. So we're going to try to get some really quick shots in here, and then pack everything up. So the kite's flying. Hopefully, we don't have any lightning or thunder. Looks like the kite's in the image. Starting to rain a lot. Is it ready, is it set up, Richard?

Are we good to go? Richard, are we ready?

Yeah, we're ready.

I can hear them in the distance. So you can hear the camera going about every second. Click, click, click. And it's nice and wet. OK, so I think that's enough standing around in the rain. Hopefully we have some good shots. We can go analyse them on the laptop and see if we got anything useful, and can see where this tether is on this very cloudy, rainy sky.

Different problem. Why that and not that? So maybe we should increase this, because then it'll, even if it diverges incorrectly, the angle that we'll look within might find it again. No, that doesn't work. Try cropping the kite out.

I think it's where it goes wrong here, straightaway, and then it just keeps going on.

Yeah, that's probably true.

We've come inside and taken refuge from the rain. We're a little bit wet, but in true Cambridge fashion, it's now sunny outside. We've uploaded some of the photos, and we're doing a little bit of analysis. We're playing around with different parameters to see if we can, indeed, find the tether in this image that really just looks like a bunch of clouds.

So this is the original image. And here we have what we've been able to find. So we've been able to find the bottom of the tether. You can see the tether is actually right there, but it starts to diverge a little bit right at the top, right where the kite is. So that's not ideal.

So we're going to try playing around with a few more parameters and see if we can, indeed, find the whole tether.

Could you reduce the number of points? I mean, increase the number of points.

That it's averaging?

Yeah. Uh, no. Increase the number of points.

Ah, that it's averaging. Control-Z that and increase the angle, two, four, two. Hopefully it works, Oh.

Oh, no.

No. Hmm, it just goes wrong right there. This is the boring bit of sitting in front of a computer and looking at code. That might help.

Disaster.

It's not a disaster. 75% of it is good. Four or two?

Four.

Four. Let's try four. Hey, there we go. Success. OK, so the bar's getting a little busy now. We've been here for a while. But we have managed to get this programme to work, and we've found the tether.

So this is our original image. You can see the kite in the top right-hand corner. And it's really hard. If you don't know where the tether is, you can't really see it. But with the programme, if I just run it quickly, it's going to try to find that image.

So what the programme is doing-- see, it's already found it. It was actually pretty quick. And you can see now it's spotted that tether. So we found the tether in this image, even though it's all cloudy and there's this background.

So what the programme is doing is it's sort of breaking it up into horizontal chunks, and then running a filter. And this philtre runs at a bunch of different angles, because we know the tether is going to lie in one of those angles. So even if we can't see the tether at one point, because there's clouds in the background, and different tones, along one sort of line there's going to be a peak, a spike, in the intensity of the pixels.

So we do that for the entire image, running it up and down the entire thing in these chunks. And then, where we have the highest one, that's where we actually start. So we pick that point, and then from that point, work up and down. And then we need only to look within a certain distance and a certain angle from the spot that we found.

So this programme will be really useful in terms of experimental validation. With two cameras looking at this, we can find the tether in 3D and work out the dynamics. I don't need to attach any sensors to the actual cable itself. And I can get dynamic response.

So it's really useful. I'm pretty proud of this entire programme. It's really exciting and it will be really useful for my Ph.D.

[MUSIC PLAYING]