Hacker News new | past | comments | ask | show | jobs | submit login
Fukushima: Radiation Detection Using Artifacts in CCCD Image? (hackerfactor.com)
98 points by brudgers on April 17, 2011 | hide | past | favorite | 43 comments



I'd just punt this one. Cover the lens so that the image is otherwise black, get a good calibrated Geiger counter and wander around taking pictures of your lens cover and noting the counter's readings.

Go back and count the light dots in your black photos and see if they correlate with the counter readings. Use your fancy exposure-time/fstop/ccd-size math after you're sure there's something going on.

For that matter, bring along a crummy cmos sensor camera as well. More sensors, more fun.


That's not punting, that's calibrating.


As a Brit living in the US it took me over a year (and several misunderstanding) to realize that there is a difference in the use of the word 'punt' between US and UK. The main confusion was between:

  1. (Informal) To cease doing something; give up [1]
  2. (Chiefly British Slang) To gamble [1]
Often the context allows for either of those to make sense. Off topic, I know, but maybe it will help others avoid the confusion.

1. http://www.thefreedictionary.com/punt


Yes, thanks for that. I should know better by now. Once while living in Germany I asked someone in my German-peppered-with-english-words-for-german-words-I-didn't-know if they were concerned about the amount of preservatives in some sort of shelf stable cheese food.

They heard Präservativ when I probably meant Konservierungsmittel. Hilarity ensued.


If anyone's wondering, Präservativ means "condom" in German. My favorite false cognate in German is "das Gift", which in English is in fact "poison".


It also bugs me that there's no clear and simple equivalent of "schenken". (There are words like `bestow', but they don't quite cut it.)


For 1, at least as I hear it used in the tech industry, I think there is some connotation of doing it later. More like defer than give up entirely.


The (American) meaning I associate with "punting" is to defer the responsibility to someone else.


However punting (as in American Football manoeuvre) is not that well known a term in the UK. Perhaps punting as gambling refers to standing upright in a flat-bottomed boat an hoping to remain dry?


It is a standard term in rugby, anyone in the UK who knows about that sport is familiar with that definition. It's more commonly just called a 'drop kick' but people would know what punt meant in the context of sport.


Yes, you could go on to create a chart that you could extrapolate future readings from, but I was just trying to answer the question of correlation between photo artifacts and actual radiation.

You'd want to go back with a plan and find a wide range of different radiation levels to get a calibrated scale worth using. Your iphone almost certainly has a wacky non-linear response to gamma rays.


Also make sure to get a variety of angles to the suspected source. The sensor is pretty planar, so will probably have a much smaller response when it's edge-on versus facing directly towards or away.


Good thinking. If this effect turned out to be true, directionality would be useful data as well. It could be like foursquare, but for getting nuked.


I think the problem with trying to determine radiation fluence is that a lot of the dots appearing are likely due to damage to the detector elements, rather than instantaneous detections.

I've seen CCTV used to monitor high radiation areas and the images are always riddled with artifacts from damaged sensor elements, even when no radiation is present. They have to replace the cameras on a regular basis.


I tried using a cheap webcam CCD to make a true RNG a while back, but I had the same problem. After a while I got permanent damaged elements and it ruined the randomness.

http://pics.brizzly.com/1P3J.jpg


You can see this in photos and movies from Chernobyl. In some pictures, like those taken from the roof, there's a noticeable tint to the bottom of the film as the radiation comes up from below. There's also a lot of noise like in the picture shown here but much worse, in some movies from inside the reactor. Pretty creepy. You can see some of the noise in this video, but the bad youtube quality makes it difficult: http://www.youtube.com/watch?v=jKnFurg5-Ag


Do you have a link for those photographs? I would think the film would be saturated generally rather than from below. Also in an SLR the top of the resulting image would be towards the ground while the film was in the camera, no?


Your image reversal point is a good one. It might be that it was reversed in the images, I don't really remember.

I couldn't find the photographs, but they are mentioned at some pointsin the "Battle for Chernobyl" movie: http://topdocumentaryfilms.com/the-battle-of-chernobyl/ (Highly recommended movie, btw. Very freaky.)

(Also noticed that it's at Google video. Watch it while you can!)


Depending on the exposure, you can also see noise in photos taken in space. http://upload.wikimedia.org/wikipedia/commons/archive/9/95/2... is a decent example. The cameras used by the ISS crew are pretty high-quality, so I doubt all those hot spots are from a bad sensor. They're most likely cosmic rays.


Looks like taken with rather dim lighting, i.e. high-iso. In which case all these dots in the dark area could be explained as hot pixels (pixels that have a tiny bit of leakage that then shows up when massively amplified). Of course, that leakage might be due to comsic rays ;-) but the amount of them doesn't look worse than what you might get on earth (depening upon circumstances).

Also, with really long exposures, this leakage tends to add up, too.


EXIF data days the exposure time was 1/1600th of a second at ISO200. The earth is really bright and that is a ton of hot spots for such a low ISO, fast exposure.


Thanks for the link. At about 2:30 you can see lots of 'stars' in the video. http://www.youtube.com/watch?v=jKnFurg5-Ag&t=2m28s


You _may_ be able to determine the number of radioactivity events (becquerels). But you will not be able to determine the type, or the strength of the radiation. So you will not be able to measure rem.

And rem is what matters for most purposes.


On the contrary, like one of the comments on that page says, you are likely to only detect gamma. Alpha and beta will be stopped long before getting to the sensor.


Not all gamma is the same, the energy matters a lot in terms of how it affects the body. And of course when calculating the dose you need to know the energy - you add the total energy, not the total number of photons.

Alpha and beta from distant sources will be blocked of course, but if there are radioactive isotopes in the environment they may decay right near the sensor, and you won't be able to tell.

Distant gamma is a much smaller concern than radioactive isotopes in the immediate environment.


True about the gamma energy dependence, of course, but alpha and beta won't really get to the sensor unless the decay is inside the camera. And even gamma doesn't have that long range, even in air. You pretty much have to have contamination nearby.


For reference, you can comfortably detect, measure and map ground produced gamma radiation from natural elements from a distance of 50 - 80 metres in the air above.

There's also a respectable number of cosmic ray events in the atmosphere all the way down to ground level leading to high energy gamma counts as a matter of course.

Radioactive Iodine and Ceasium particles (in small numbers admittedly) made their way across the Pacific and became gamma emitters caught up in the air scrubbers of the University of Washington in Seattle.


Well, cosmic rays can have energies far, far higher than you get from typical radioactive decay, so they (or gammas produced by them when they hit the atmosphere) have a lot more penetration.


Why so much hate? It's an interesting question, not trying to use the numbers to create safety policy or anything.


If that came across as hate, then sorry.

I'm just pointing out where it doesn't work, that's all.


Yes it is possible, and I know people are doing it.

But it is a too short time to get an accurate reading on the level, a single photograph is exposed under a very short time.

And a camera is probably not shielded against radiation, so it might be impossible to predict how long the actual time of exposure is.


I'm a Navy physician, with an undergrad degree in physics, working at one of the Joint Task Forces assigned to the whole Fukushima mess.

If this had been taken in Tokyo, which is at background, I would have said it's something else. But right there, at the reactors, they are still very much in the several mrem/hr regime. Someone could probably calculate this. I don't think the "speed value" from the EV system is going to be of much use for this calculation, though the "aperture value" will. Nor does the color of the pixel tell you anything other than the color of the pixel that was hit by a gamma ray.

Never the less, a simple count could give you a pretty good first order estimate. Assume the radiation here is around 10 mrem/hr on 15 March (there may be a more accurate published estimate out there, I'm just going on trends . . . I'll reboot into Windows and see if I can get you more accurate info off our share drive).

[Edit]: Okay, so the 15th was the day of the biggest releases, ranging up to 12000 μsV/hr. Reactor 2 had an explosion at 0610 local, the reactor 4 fire was from 0600 to 1116 and there was a second explosion at reactor 2 at 1006. The 12000 μSv/hr peak is around 0930 with a second peak of 8800 at about 1000. I'm just looking at a graph, but it appear to decay down to around 300 μSv/hr over the rest of the day, then there's another, unlabeled event at about midnight. The metadata says the image was taken at 0858. This is in the middle of the upsurge. The area could be anywhere between 1000 μSv/hr and 12000 μSv/hr.

Are there other pictures available from different times?


There are a ton of other photos here:

http://www.tepco.co.jp/en/news/110311/

I'm not quite sure which are camera original or which have single pixel artifacts like that, though. Dr. Neal Krawetz, who runs that blog, would know though. He's an expert in digital image forensics.

I suppose you could just zoom in, count pixels, then try to get timestamps & locations out of the images via JPEGsnoop or similar tools for reading the metadata, but you'd have to contend with however many different cameras they use, etc.


The first of the packbot pictures from today also has a small number of blown pictures.

http://www.tepco.co.jp/en/news/110311/images/110417_1f_1_1.j...

taken at 1138 on 4/17.

In thinking a bit more, I guess without knowing how that Sony camera clears its registers, we don't know how long it took for the wells to collect those gamma rays (eg: are the registered "cleansed" just before each shot? And then cleared again after the information is transferred off the sensor?). However, the "ISO" setting probably doesn't matter. The aperture will matter essentially as much as the lens matters, and considering the lens is gathering in light from that entire area, that may be significant, I just don't know.


Aperture doesn't matter. The gammas that make those bright pixels will comfortably penetrate the shutter, lens, and the entire camera body. It's the time between erase and readout that matters, as you say, and it will act as a 4pi dipole detector.


Keep in mind that CCDs can be sensitive enough to background that they're usable as chaos for random number generation[1]. I am not a physicist or photographer, but it may be that some of those spots are the result of settings, damage, camera quality (an old Sony?), or just responding to lower level radiation than you'd expect.

[1] http://www.lavarnd.org/what/process.html


Some HD Video flyovers done by the Japanese Self Defense Force show various abnormalities in quality once they fly over the damaged reactors. See this video[0] from the 2m40s mark.

[0] http://www.youtube.com/watch?v=M22Gt4sswEA#t=2m40s


I only see weird distortions which I would interpret as aliasing between the readout rate of the camera and the vibrations from the helicopter. Any radiation should show up as noise.


I'm pretty sure you'd only be detecting alpha and/or beta particles, depending on the materials in the camera. Gamma detection would really require either a Geiger counter or a pair production type of detector, or a APD/PMT detector.


Cameras are the universal sensor.


They mostly work for electromagnetic radiation.


It's funny, but there are very few things in modern physics which aren't measured via electromagnetic radiation.


Indeed. And cameras probably also pick up some ionizing radiation of different kinds.




Consider applying for YC's Spring batch! Applications are open till Feb 11.

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

Search: