Camera takes pic through *edge* of glass pane
August 22, 2018 1:47 PM   Subscribe

The newatlass link has research from University of Utah System has poor resolution now but is supposed to be good for "industrial applications". The light passes through the pane of glass and "about 1 percent of it is scattered" out to the edges of the pane, effectively sampling the light going through. There it's bounced off reflective tape to a sensor and algorith crunches data to produce image of what passed through glass. Hopefully algorithm/crunching gets better. Ref(s): A paper on the research was recently published in the journal Optics Express. University of Utah link with Video
posted by aleph (25 comments total) 16 users marked this as a favorite
 
That's pretty awesome. It's right up there with using windows as microphones.
posted by GuyZero at 2:22 PM on August 22, 2018 [5 favorites]


Or lasers used as cameras that can see around corners
posted by Popular Ethics at 2:28 PM on August 22, 2018 [1 favorite]


So what are the practical applications of this? Cameras built into windowpanes that give very distorted video or picture of whats through the window... instead of looking through it?
posted by GoblinHoney at 2:46 PM on August 22, 2018


So what are the practical applications of this? Cameras built into windowpanes that give very distorted video or picture of whats through the window... instead of looking through it?

A camera that you can see through- you could put something like this into a pair of goggles and the actual camera part would be on the side.
posted by dilaudid at 2:53 PM on August 22, 2018 [5 favorites]


Neato!
posted by Greg_Ace at 2:56 PM on August 22, 2018


So what are the practical applications of this? Cameras built into windowpanes that give very distorted video or picture of whats through the window... instead of looking through it?

Distortions are easy to correct.

What we get is the long overdue videoconferencing equipment that lets you look directly at each other.
posted by Tell Me No Lies at 2:56 PM on August 22, 2018 [16 favorites]


So what are the practical applications of this?

If they can manage to improve the image quality, and I suspect they will, then I'd think cameras in potentially every window, and you won't even know they are there!
posted by Mister Fabulous at 3:05 PM on August 22, 2018 [1 favorite]


So what are the practical applications of this? Cameras built into windowpanes that give very distorted video or picture of whats through the window... instead of looking through it?

Unobtrusive security cameras on windows. Unobtrusive object sensors on car windows.

The obvious extension of this (hinted at in the article) is to put multiple sensors on each edge of the window and combine them to get a higher quality image. I'll bet with enough training data it can get remarkably good. If it could be made good enough to replace the forward camera on phones it could solve the notch problem.
posted by jedicus at 3:05 PM on August 22, 2018 [3 favorites]


> So what are the practical applications of this?

Laptop cameras you can't put tape over.
posted by Reclusive Novelist Thomas Pynchon at 3:06 PM on August 22, 2018 [32 favorites]


Edge-to-edge cellphone screens that can also take selfies.

Also, the telescreens from Big Brother that watch you back.
posted by BungaDunga at 3:14 PM on August 22, 2018 [3 favorites]


it'll be super fun when every network-connected computer is basically a palantir.
posted by Reclusive Novelist Thomas Pynchon at 3:18 PM on August 22, 2018 [8 favorites]


The obvious extension of this (hinted at in the article) is to put multiple sensors on each edge of the window and combine them to get a higher quality image. I'll bet with enough training data it can get remarkably good. If it could be made good enough to replace the forward camera on phones it could solve the notch problem.

...and replace it with the petal problem.
posted by The Tensor at 3:53 PM on August 22, 2018 [2 favorites]


Neat. A decade ago I saw a Perspex artwork that emitted daylight from its edges and thought 'hmm, that's basically information, how could it be useful?'

A lot of the time you don't need a picture, spectral bands are enough to identify clothing, skin color. Or simple objects fluorescing in the room ... The human body literally glimmers PLOS link

I can't find it right now bu a while back I saw a system that could create an image of a room interior (and the room dimensions) from the light coming out of an open door - it seems like its the same effect.

I imagine a lot of (even more) meetings in windowless rooms.
posted by unearthed at 3:58 PM on August 22, 2018 [4 favorites]


What we get is the long overdue videoconferencing equipment that lets you look directly at each other.

Ha! I am working on exactly that problem. And without giving too much away, there are much easier ways to solve this problem than the technique linked in the article.
posted by ryanrs at 4:33 PM on August 22, 2018 [7 favorites]


I've wondered about if sensors get cheap enough to have a loop of them all the way around the edge. Then, instead of reflecting the light at the edges, absorb it between the sensors to reduce the noise. Might be a simpler/faster way to get the info.
posted by aleph at 4:42 PM on August 22, 2018


And if it gets accurate enough it can always work in glasses/goggles along their edges and provide invisible sensors for what the person is viewing. Lots of possibilities in that one.
posted by aleph at 6:10 PM on August 22, 2018


… and so easily messed up by a simple scratch across the pane.

As someone who makes edge-lit signage and displays sometimes, perspex scratches if you look at it funny.
posted by scruss at 6:18 PM on August 22, 2018


>> What we get is the long overdue videoconferencing equipment that lets you look directly at each other.
>Ha! I am working on exactly that problem. And without giving too much away, there are much easier ways to solve this problem than the technique linked in the article.


Well quit reading MetaFilter and go make the thing work! :-)

Seriously though, I can think of a number of ways to approach the product but the fact that no one has shipped it yet leads me to believe there are non-obvious problems to be solved.
posted by Tell Me No Lies at 6:21 PM on August 22, 2018


It does work! You can totally tell if the other person is looking at you, or over your shoulder. It's pretty cool.

non-obvious problems to be solved

We want a lot of money for it.
posted by ryanrs at 1:30 AM on August 23, 2018 [3 favorites]


Here, let me add a second-order consequence.

Researchers at MIT a few years ago demonstrated that you don't need a laser microphone to pick up sound remotely: passive high-definition video imaging can be used to recover audio from vibrations in objects like a bag of crisps or a potted plant (source: The Visual Microphone: Passive Recovery of Sound from Video). If an object vibrates when sound passes through it, with a good enough camera you can extrac the audio signal.

Now we're talking about effectively invisible video cameras. If you build a video sensor into the frame of any glass object, you can obviously use the glass surface as a mike ... what's less obvious is that if your video sensor is good enough, you can use any semi-rigid surface in view as an acoustic pick-up. So, presumably, you have a mike that can focus in on any conversation at a cocktail party and listen in on just the one you want to eavesdrop on, if there's a suitable resonant surface in view. Such as, oh, a ceiling tile, or a glass of wine, or a paper tissue.

The NSA will be overjoyed.
posted by cstross at 2:16 AM on August 23, 2018 [4 favorites]


>> non-obvious problems to be solved
>
> We want a lot of money for it.


I’m not sure that qualifies as a non-obvious problem.
posted by Tell Me No Lies at 8:36 AM on August 23, 2018


I'd like to see more about the theoretical maximum of information recoverable in this method. The examples are crude, but I suspect that may be in part because they have no other choice.
posted by Glomar response at 9:38 AM on August 23, 2018


I'd like to see more about the theoretical maximum of information recoverable in this method. The examples are crude, but I suspect that may be in part because they have no other choice.

I have a feeling a lot of that is going to depend on the medium and how many sensors they use.
posted by Tell Me No Lies at 10:15 AM on August 23, 2018


You got you panopticon in my pyrex!
You got your pyrex in my panopticon!
posted by ...possums at 3:18 PM on August 23, 2018


I’m not sure that qualifies as a non-obvious problem.

Yeah, the problem is fairly straightforward. It's the solution that's been non-obvious.
posted by ryanrs at 3:52 PM on August 23, 2018


« Older TEMPLE, GRENADE LAUNCHERS, NO ODDJOB   |   “NERF THIS!” Newer »


This thread has been archived and is closed to new comments