Hacker News new | past | comments | ask | show | jobs | submit login
Non–line-of-sight imaging over 1.43km [pdf] (pnas.org)
81 points by walterbell 19 days ago | hide | past | favorite | 20 comments

So these guys successfully got images of objects sitting behind a solid barrier from nearly a mile away by analyzing photons scattered around the barrier.

I imagine the only way to hide objects from this technology is to enclose them well enough to prevent most photons that bounce off the objects from leaving the enclosure.

And this is just an early-stage proof-of-concept. There's a good chance the technology can and will improve significantly over time.

Just saying.

2s exposure time, 64x64 raster grid, looking behind a barrier 1.43km away.

Summary: by bouncing light off of a target barrier and having the light scatter around the barrier, you can compute an image of objects behind the barrier.

I was immediately reminded of the Dual Photography paper (well, I had to look it up first; I only remembered the "photographing a playing card that is facing away" demo at the end):

- http://graphics.stanford.edu/papers/dual_photography/

- https://youtu.be/p5_tpq5ejFQ

That's an absolutely science-fiction-esque paper. I can't believe that's real, and I'm surprised to see such little follow-up research. I assume it's not practical to generate the light transport data with any reasonable speed?

Isn't this just using the 'visible wall' as a really dull mirror? Or am I not getting something here.

What are some of the applications of this technology?

Edit: I'm also wondering if anyone has any idea as to the estimated cost of such a technology? Thousands, tens/hundred of thousands, millions? A quick search of lidar sensors seem to be in the range of $2500 - ~$10000 for ranges of sub 30 meters.

This technology seems similar(?) to lidar from what I can tell.

Teslas have been shown to prevent accidents by braking based on things drivers can’t see. https://www.mobilescout.com/tech/news/n77946/tesla-autopilot...

I imagine that idea could be improved more with this technology, but Tesla is very anti lidar so it’s have to be someone else.

Military, police, surveillance. The usual suspects when it comes to oppressive regimes like china.

On a more upbeat note though, imagine the possibility of producing a 3d image of your internals during keyhole surgery or inspecting remote structures like solar panels and dams via satellite.

My local police department would be really interested in this tech. They've been using automated license plate readers, facial recognition tech, drones, predictive policing, and have a Ring doorbell partnership.

In a very real way, American police departments are functioning in roles that the military serves in most dystopian societies.

Where do you live? Remind me not to visit.

Any major city in the United States has all of these features

It's really incredible that we just let ourselves slide into this surveillance dystopia, isn't it? I yearn for the halcyon days of August 2001.

Dirt cheap. Laser with galvo (like many laser projectors for light shows have), plus 3 photodiode detectors and optics.

Most of the cost will be in the telescopes.

If you don't count personnel time integrating/fiddling around, maybe $1000 at qty 1?

First one that comes to mind is military. Where to aim (or not aim) your high powered weapon that can shoot through a wall.

Electro-optical reconnaisance spacecrafts

Realistically there aren't any practical applications. It's a very cool demo though.

Seeing inside a building has no practical applications?

Or other non-line-of-sight things, like sensing through an aperture in a body?

This is neat. Assuming the usual things about costs going down and capabilities increasing, both exponentially of course... then one can imagine some novel and non-military applications.

For example, vehicles seeing around corners to avoid collisions or being able to see further into damaged industrial sites than robots can go today (i.e. Fukushima), or for cave explorations (especially if they improve significantly and start building images from multiple reflections - see around the corners that themselves are around corners).

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact