Camera that Sees through Smoke and Fog Underway

tomschuring writes "The Age has a story about IATIA, who have been given $2.7 million by the Defence Department to fund development of a military spy camera capable of seeing through fog, smoke and dust storms. The technology uses a highly sophisticated camera that captures three images simultaneously through a single lens. Images thus resolved from between the particles making up fog, smoke, and dust storms are formed into a single picture of the hidden target."

Comments

Dense Camera Arrays for seeing through bushes

[vectra14]

These guys at stanford have done some really amazing stuff that's directly related. Except that they has literally dozens of cameras (as seen in their ppt), and their research seems to concentrate on multifocal image reconstruction (see ppt slides, presentation is quite good)

Link (has cool results links)

Re:Dense Camera Arrays for seeing through bushes

[ajna]

The Stanford work is actually entirely different. They utilize parallax -- in other words, their cameras are in physically distinct locations and see the scene with different perspectives. The IATIA work utilizes a single point of view, with images captured with the focal plane at the desired location and then slightly fore and aft. Read more here, at a Columbia site.

Quantitative phase microscopy is a relatively new technique that can generate phase images and phase-amplitude images. In practice, to obtain a quantitative phase image one collects an in-focus image and very slightly positively and negatively defocused images, and uses these data to estimate the differential with respect to the defocus of the image. These images (a through-focal series) can be easily obtained in our system with our z-motion nano-positioner. The resulting data can be solved to yield the phase distribution by Fourier-transform methods. Results are obtained by essentially solving an optical transport equation. Significantly, the phase that is obtained does not have to be unwrapped, as is required for interferometry.

I'd be lying if I told you I completely understand the quoted paragraph, specifically what "essentially solving an optical transport equation" refers to, but I'm sure some cursory googling will lead the curious to specifics, certainly more than googling on terms in the article summary would yield.

Keith Nugent

[metlin]

Hmm, Keith Nugent is fairly well known in some niche areas of optics. If I remember right, his initial work on the use of x-rays and the like to compensate for normal visible hindrances were met with some opposition, but he is quite famous otherwise.

That was because, ironically, this was developed as a method to visualize biological stuff, and some felt that his methods would not quite be suitable for such a task. His ideas were to use various parameters such as phase, intensity and angle of vision to extract information which could be correlated and converge to recreate images with minimal amount of information, which later gained acceptance.

I guess he developed on that technique, and later on evolved to have the military to take notice. Interesting neverthless.

Already exists

[leabre]

I was in the US Navy from 1994-1996 and the damage control teams already have a special camera (forget what it is called) that can see through dense smoke (the type you would expect from a jet fuel fire or amunition fire on a ship) and help you to see clearly through the smoke.

Wonder what makes the camera in this article so different from the technology the Navy already uses... I'm sure the current navy breed is much more advanced than it was 10 years ago.

Thanks,
Leabre

my thoughts

[Large Bogon Collider]

I'm not a 100% sure, but the technique involves phase shift. As light of a single frequency passes though an medium, its phase is altered and light propagation is delayed. If you can computationally filter out all out of phase shift information caused by fog, for example, you can "see" what the hidden object looked like. This process is quite CPU intensive. It seemed that about a grayscale SVGA sized image (0.41 mp) took 1.5 secs on a PIV 2.4GHz to calculate. This should improve with algorith tweaking and using FPGAs.

This may also have medical applications in terms of optical imaging - see through the patient (arms and legs only, probably). Shine a bright light at the patient. Capture the ealiest photos that emerge (the ones that had a direct path to the camera). Ignore slow photons (ones that were absorbed and release or bounced around). Voila, instant imaging without x-rays. IIRC, this was in development years ago.