[Horse]

(03-29-2016, 04:11 PM)RodgerRoentgen Wrote: For the other side of the equation, we'll need to firm up the numbers you observe with the recordings you have made. 10 FPS is reasonable for your recordings. Also consider if you are compressing time when you record, we need to factor that in. For example, if playback is at 2x speed (1 min play = 2 min record), we need to account for that in the FPS value.

Glad you're up for number juggling.  Not compressing time when recording.  Only use 1.5x and 2x for playback to save time.

Quote:ok, we'll also want to consider the recorded vs. streamed FPS. While you may record @ 10, the stream could be less (or more). One way to check is to step through all 10 frames in 1 second. If there are any frames that are identical, you'll need to subtract the number of duplicates from the total FPS. For example, if in 10 frames, 2 frames are duplicated (4 frames total, 2 individual pairs), then a realized frame rate is 8 FPS.

The new tepco stream comes with duplicate frames.  In 10 frames half are duplicates.  That changed with the flowplayer.  In the old streams a spark was one frame; with the flowplayer, sparks are two frames, the extra frame a duplicate.  That’s why I guessed they were adjusting the frame rate on the new flowplayer.  In 10 frames, 5 frames are duplicated so the realized frame rate is 5 fps.  

Quote:The calculation of 1 dot per frame = .1mSv was not based on any hard numbers. To come up with that number I guessed many things. I guessed the following:

• Frame rate FPS
  

I poorly assumed 30 thinking back to NTSC, but clearly the live stream is more like 10

• Dose/Rate
  

The video just had 'over 20mSv/hr' in the description, so I guessed higher @ 25mSv/hr, but again we don't have any idea how accurate that number is.

• Dots per frame
  

even this I really cheated on, I roughly counted the dots in one quarter of one frame. This is a very unreliable number considering we can sample more data to get better averages.

The video of the interior of the reactor was probably taken with a better camera than the tepco feeds, so I’d stick with the 30fps.  The dose/rate guess @25mSv/hr is good enough for starters, considering that could still be a low-ball number.  The dots per frame count; that was smart counting a quarter and extrapolating.  The dots per frame would vary considerable, but a snapshot count is good enough for starters.  

Agree we should try to get a reliable count for the reference part of the ratio; I’ll look for more tepco video showing dots with a Geiger count, maybe Chas or Cali have links handy.  The ISS recordings are no help because they don’t include a radiation reading.  

I hope differing frame rates don’t affect the ratio we’re trying to build.  

Reference count at 30 fps  ::   tepco count at a realized 5 fps

I picked a tepco count of 20 dots per hour because that’s when the emissions get thick and I use it as an indicator that something is steaming up.  Once we get a realistic reference for the ratio, plugging in tepco counts should be easy.  

I know you’re worried about the accuracy of the numbers and that’s good.  Would like to plug in some numbers and see if we’re getting realistic results.

[RodgerRoentgen]

(03-29-2016, 08:03 PM)Horse Wrote: The video of the interior of the reactor was probably taken with a better camera than the tepco feeds, so I’d stick with the 30fps.  The dose/rate guess @25mSv/hr is good enough for starters, considering that could still be a low-ball number.  The dots per frame count; that was smart counting a quarter and extrapolating.  The dots per frame would vary considerable, but a snapshot count is good enough for starters.  

While I follow your train of thought, we must also remember that the cameras inserted into the PCV have to be very small in size. The camera, a light source(s), sensor payloads and some minimal mechanical device(s) for crude movement, all must fit through pipes 1-3 inches in size. In this case, they are probably sacrificing some video quality here to get everything to fit through. I'm thinking we estimate a little less for the small cameras in tight places. lets at least assume around 24pfs to be cautious. We still have another frame rate consideration though... more on that below.

(03-29-2016, 08:03 PM)Horse Wrote: Agree we should try to get a reliable count for the reference part of the ratio; I’ll look for more tepco video showing dots with a Geiger count, maybe Chas or Cali have links handy.  The ISS recordings are no help because they don’t include a radiation reading.  

ok, here is a possible first candidate:
https://www.youtube.com/watch?v=b8iVhnp9vFk

This video is TORTURE, but it does have a displayed count as well as SOME gamma distortion artifacts. It's a horrible video though. We might be able to pull out a sliver of data from this one. In only a few sections of the video we get to see the count the packbot is detecting. I picked out the following times with displayed counts (quite HIGH)
1:54 --- .85 Sv/hr
2:22 --- .87 Sv/hr
2:43 --- .84 Sv/hr
4:54 --- (can't tell...)
18:35 -- (displayed on larger screen in background but out of focus)
19:22 -- .19.Sv/hr
Again, this video sucks, but we can take some frames around those times that show dots. We might have to extrapolate counts for the entire frame based on the dark portion(s) that we can count dots in. It'll be fuzzy, but theoretically more accurate than our previous calculations. As we find more data like this and keep normalizing and combining it, the overall accuracy should increase.

The first 5 minutes or so we could probably get some data from. The rest of the video isn't that helpful. Also, why the hell cant they just put the camera they are FILMING A LAPTOP with on a TRIPOD?!? ARGH!!

(03-29-2016, 08:03 PM)Horse Wrote: I hope differing frame rates don’t affect the ratio we’re trying to build.  

This will always be a factor, but at least on your end it should be constant and predictable. Whenever more data is added from existing videos that have known rates, we'll need to very carefully consider the frame rate on several levels. While a camera might achieve 30FPS recording, we are not actually watching what the camera is seeing. Through the various layers of software, hardware and firmware that makes a live webcam stream possible (or youtube), more and more image and time compression is introduced, and this destroys some of the data we are looking for. We can use some good logic to take what data remains and get close to the data that has become obscured.

I can get into some detail about the image compression but it'll be a somewhat lengthy explaination. I think I'll start another thread to reference that train of thought. What I'll try to express in more detail is that even though camera capture 30FPS, when we watch it on the stream we lose lots of data and get 10FPS. The same applys to video we get from youtube. These videos are generally compressed again when uploaded, and even the flash player displaying the video on your computer can apply some compression to smooth out videos. All this adds up crappier image quality and less frames per second. Really we want to kind of do what you did and step through every frame in one second and see how many unique images are present for any video used for data.

(03-29-2016, 08:03 PM)Horse Wrote: I know you’re worried about the accuracy of the numbers and that’s good.  Would like to plug in some numbers and see if we’re getting realistic results.

Well, I believe we should strive for being as accurate as possible. I see plenty of speculation on radiation related to fukushima with nothing to back it up. We will also be speculating, but at least we have SOMETHING based on real data to back up anything we calculate. Ultimately we will never have an accurate number to say X amount of distortion on cam Y = Z dose of radiation. There are too many variables for us to nail down. But don't give up hope, this is still a good exercise and can help get a better understanding of all of the technologies involved. We can still get close in our calculation though, but we will never be able to check our work by standing in front of U4 webcam with a geiger counter...

[Horse]

Roger, I see getting a good reference will be the biggest challenge.  The youtubes may not be the best for counting or maybe a better way to say it;  if we find a youtube with rad readings and dots we should find the original tepco vid for the best count of a frame.  Tepco's vid library is hard to see titles and some vid's of interest went 404 on me.  Most of the vid's I checked had no time reference, extra work to guess a frame rate/sec.  FYI, I recorded some of the suggested youtubes and noticed odd, inconsistent frame duplicates when stepping thru so they have a low frame rate and also the youtube compression make it hard to work with.  The best candidate would be a tepco video with dots that have a time counter and a rad reading associated with the vid.    


Some samples of dots:

youtube vids

(Full) Fukushima JP packbot investigation reactor 3 (roger wants to start with this one)
https://www.youtube.com/watch?v=b8iVhnp9vFk

Reactor 3 PCV equipment hatch investigation 9/9/2015  (pia posted spark sample)
https://www.youtube.com/watch?v=jzAdZFU8N_c

Inside Fukushima I Nuke Plant Reactor 2 Containment Vessel - Longer Version (1/4)
https://www.youtube.com/watch?v=aR--2dASoJA


Tepco vids  http://www.tepco.co.jp/en/news/library/a...atid=61793

2011.4.20 The videos were taken by the Packbot inside the nuclear reactor building of Fukushima Daiichi Nuclear Power Station - 1st Floor, Nuclear Reactor Building, Unit 1
http://www.tepco.co.jp/en/news/library/a...atid=61793

[Horse]

https://www.youtube.com/watch?v=b8iVhnp9vFk
might be

2012.11.28 Fukushima Daiichi Nuclear Power Station Unit 3 PCV equipment hatch rail (Video recorded by Packbot)
http://www.tepco.co.jp/en/news/library/a...atid=62055

2012.11.28 Fukushima Daiichi Nuclear Power Station Unit 3 PCV gas control system duct inspection (Video recorded by Packbot)
http://www.tepco.co.jp/en/news/library/a...atid=62055

search by date nov 2012
http://photo.tepco.co.jp/en/photo/2012/201211-e.html

I downloaded the two videos, got windows media files. Would be interesting to get a couple reference sources to check for consistency of the rad rate and dose to the amount of sparks. Ideally, a screen full of just a few sparks would make it easier to count.

[Chasaha]

I think I found the video I had mentioned in a previous post by YT user BioNerd.

[UPDATED] radioactivity on video - just like inside chernobyl
https://www.youtube.com/watch?v=jFNvYA7731o

see the americium at 3:00 minutes in  

[Horse]

(04-05-2016, 01:19 PM)Chasaha Wrote: I think I found the video I had mentioned in a previous post by YT user BioNerd.

[UPDATED] radioactivity on video - just like inside chernobyl
https://www.youtube.com/watch?v=jFNvYA7731o

see the americium at 3:00 minutes in  

Thanks ChasAha, I liked this one.  The americium storm looked like the videos inside of a reactor.  The one I think might be useful for counting purposes and to get a rough idea of the rad range the tepco sparks might be in is the Cs-137 sample 0.250 u Ci. 30.07 years.  I recorded the video, stepped thru the first 20 frames and saved grabs.  I see 0 to 5 sparks each frame, mean of 2.5, assuming 10 frames per sec, for 25 sparks per second.  On the tepcams I'm worried about 25 sparks per hour at 5 frames per sec.  It's only a rough estimate as I haven't had time to actually do counts yet. The other samples in the video might give us comparative measures to tighten the reference.

[RodgerRoentgen]

I think I've found some really good starting source videos. Thanks for your video suggestions Horse, they led me to 4 videos on the Tepco site that have rad readings on screen and high rates that permit counting per frame.

The videos are these four:
http://www.tepco.co.jp/en/news/library/a...d=jilpgxe6
Download link
http://tepco.webcdn.stream.ne.jp/www11/t...20_01j.zip

http://www.tepco.co.jp/en/news/library/a...d=o852v7u0
Download link
http://tepco.webcdn.stream.ne.jp/www11/t...20_02j.zip

http://www.tepco.co.jp/en/news/library/a...d=erg32ld4
Download link
http://tepco.webcdn.stream.ne.jp/www11/t...17_01j.zip

http://www.tepco.co.jp/en/news/library/a...id=il57o19
Download link
http://tepco.webcdn.stream.ne.jp/www11/t...13_01j.zip

I started with an initial analysis from one of these videos. I'm going to start a new thread to cover the video/frame analysis process. It became a little more complicated than I initially anticipated, but all of that should render better results. These will be great baseline videos to start with. Once we have good data sets from these videos, we can analyze other videos like those above to confirm or adjust our calculations. Thanks all for finding the videos we have so far!

Horse, to calculate rates from the recordings you have, we'll need to iron out a few more details. Let me try to get the analysis process documented first so analysis can start if others want to jump in. Then we'll refine the video data you have as best as we can. Once thats done, we can at least make a calculation based on my initial analysis. I've been working on some spreadsheets to do the work for us. More on that soon...

For now, could someone do a quick analysis of any of the above videos to figure out the actual frame rate. This was the process of stepping through and counting all frames in 1 second and subtracting the duplicated frames. For some reason I am unable to go frame by frame with my media player at the moment.

[Horse]

Oh, those frame rates throw us a curve.  All this discussion of frame rates led me to look at the tepco videos a little closer.  The tepco camera feeds before 2016 were 10 frames per second and sparks lasted just one frame, each unique frame being one tenth of a second.  Videos before 2016 had more information and detail than the Flowplayer introduced by Tepco, only on their website, at the start of 2016 when they stopped all the other feeds.  Flowplayer displays 10 frames per second as well but only 5 frames are unique and 5 frames are duplicates thus the sparks last for 2 frames now, one unique frame and one duplicate frame.  Seeing half as much detail as before, it’s not surprising that on average I now find half as many sparks as last year and the decrease in the changing pixilation even seems to make the emissions look less vigorous when viewing.

[Horse]

A video recreating sparks or dots on a camera using radioactive sources as is seen on the tepcams.  See ChasAha video by bionerd23.  

https://www.youtube.com/watch?v=jFNvYA7731o

The video test showed various sources, a, b, and g creating sparks on the ccd output.  CCD sensor was removed from protective coverings to allow unrestricted emissions detections.  Alpha was said to burn the pixel to a constant white output.  Sparks were said to be beta.  

The video was a good demonstration that radiation does affect ccd cameras.  Propose further testing of common ccd cameras for more specifics to determine a radiation level and conditions that mimic the tepcam spark activity.  

Suggest using unadulterated cameras to determine what penetrates normal lens or casings and what distance and radioactivity rate are required.  

If a spark is found to be A, B, or G; what then is the distortion that occurs during increased spark activity?  

From the videos I’ve seen, here’s my guess to test.  Alpha radiation would normally be blocked by a lens or a casing and ccd sensor damage would not normally occur.  Most likely it is Beta radiation which can penetrate up to a few inches of solid material causing the random dots to appear without damaging the camera sensors.  If not, we have Gamma radiation that can easily penetrate the camera.  Would Neutron radiation cause any visual effect on the sensors besides making the camera itself radioactive?  

RogerRoentgen has Geiger counters and radioactive material to test with.  I supplied a usb camera with software to record the experiments.  

The first test with an unmodified usb camera and various radioactive sources up to 7700cpm did not produce any sparks.  

More testing is being planned.  We don’t have stronger radioactive sources to test with so we’ll try removing the camera’s protective lens and expose the ccd sensors closer to the sources.

[RodgerRoentgen]

(04-15-2016, 06:54 AM)Horse Wrote: Oh, those frame rates throw us a curve.  All this discussion of frame rates led me to look at the tepco videos a little closer.  The tepco camera feeds before 2016 were 10 frames per second and sparks lasted just one frame, each unique frame being one tenth of a second.  Videos before 2016 had more information and detail than the Flowplayer introduced by Tepco, only on their website, at the start of 2016 when they stopped all the other feeds.  Flowplayer displays 10 frames per second as well but only 5 frames are unique and 5 frames are duplicates thus the sparks last for 2 frames now, one unique frame and one duplicate frame.  Seeing half as much detail as before, it’s not surprising that on average I now find half as many sparks as last year and the decrease in the changing pixilation even seems to make the emissions look less vigorous when viewing.

Great Horse, this will be used in the calculation. I'd say based on your analysis, we can assume 10fps when the tepco cams were ASX streams (pre Oct 2015) and 5fps with flowplayer (after oct 2015). Yes, we are losing data in the dropped frames, but we should still be able to establish the same ratio if it is 10 or 5pfs.

---

Analysis and calculated results

---

I'm spinning my wheels here on how best to simplify the process. I have a working system to make a calculation of dose based on counts from the tepco cams. The attached spreadsheet includes my initial analysis as a base set of numbers to work from.

  CamAnalysis-inProgress-a1.xls (Size: 19 KB / Downloads: 298)

What really needs work in this spreadsheet is to permit entering multiple sample data results to generate some sort of useable mean or average that is then used to make the final calculation. As you can see in the spreadsheet, section 1 & 2 were my original analysis, section 3 is the area for taking all of the sample data inputs and making a more average number to use in sections 4 and 5 for the final calculation.

Hopefully others can at least see what is going on in the calculating process. My goal is to come up with a CPM from the samples that is related to the video (sample) size (in pixels). the size factor is important so we can compare to the relative pixel size of the tepco webcam. the CPM rate is used in the calculation of dose in creating a ratio of CPM to pixel area over time.

Horse, you can try entering different values in the B57 cell. this cell is where you enter 'how many seconds between counts' (average). I did my best to fill in all of the information you've given me so far for the tepco cams (frame rate and CPMs). I determined viewing size (pixels) for tepco cams  and also provide a 'masking %' to subtract out the portion of a frame where you can't catch a dot. Those are the areas with high lighting, sides of the buildings and portions of the vent stacks and cranes. I've estimated this portion to be around 25% of the frame where counting a dot isn't possible.

Finally at the end, we see the calculated value of dose in the various sv/h units. This calculation factors in all of the video sizes, frame rates, time rates and masked areas to give as acurate a calculation as I can estimate. Remember all yellow cells are calculated values and red cells are for data entry. Not all calculated values are used in the final calculation, but were there to help me better nderstand the data. Maybe others can look over the spreadsheet and find better ways but for now this is the first in progress working model.

[Horse]

Quote:HillbillyHoundDog
August 18, 2017 at 10:11 pm · Reply
HORSE, https://www.youtube.com/watch?v=uYrhWO_ZLYw&t=0s

Mini reactor- Sparks, just the same. Interesting comments…

A controlled, water shielded reactor was still releasing nuclear elements emitting Gamma radiation measured next to the pool and showing as the Cherenkov radiation sparks detected by the camera sensor.  At 7:07 in sparks are mentioned as being gamma.  The comments included measured radiation readings at the time when sparks are blinking.   A 10 mR/hr reading during a small spark burst on the video to roughly calibrate the Gamma of the spark bursts I still find on the tepcams.  

Breazeale Nuclear Reactor Start up, 500kW, 1MW, and Shut Down (ANNOTATED)
Alex Landress