The 9/11 Forum

David B. Benson · by **David B. Benson** on Thu Nov 05, 2009 9:46 pm

Trippy wrote:What I would appreciate is if femr2, OWE, and Szamboti could work out between them some common feature for synching their data streams up so that I can work with the 4 data sets (and any others produced) and actually compare apples with apples.

There isn't any. Instead, use minimum least squares estimation from residuals to align the various datasets.

The other thing I would like to see (and DBB might be the best bet here) is the displacement predicted by the current model (is it BV&L?)

The current model is B&V's.

I've seen DBB make several posts that suggest that he has access to at least partial data along this lines, and it would be great if this could be synched up with one of the sets of measured data.

I use OneWhiteEye's antenna mast feature measurements from the Sauret DVD video.

I would also consider using nobody's data, if he's willing to make his raw numeric data available.

So would I.

Anyway, for Major_Tom's question, there are no jolts above the noise level in OneWhiteEye's data. There are no deviations from the shape of the drop curve determined by the crush-down equation to be seen in OneWhiteEye's data (above the level of the noise). As smooth as silk.

Trippy · by **Trippy** on Thu Nov 05, 2009 10:01 pm

David B. Benson wrote:
Trippy wrote:What I would appreciate is if femr2, OWE, and Szamboti could work out between them some common feature for synching their data streams up so that I can work with the 4 data sets (and any others produced) and actually compare apples with apples.
There isn't any. Instead, use minimum least squares estimation from residuals to align the various datasets.

I thought about that, but if they're all using the Sauret footage (which AFAIK they are) then they should be able to say "This unambiguous event happened in the time labled frame blah through frame blah blah in my data set" (say for example the big expulsion of fire below the impact zone).

femr2 · by **femr2** on Sat Nov 07, 2009 2:31 pm

Trippy wrote:
David B. Benson wrote:
Trippy wrote:What I would appreciate is if femr2, OWE, and Szamboti could work out between them some common feature for synching their data streams up so that I can work with the 4 data sets (and any others produced) and actually compare apples with apples.
There isn't any. Instead, use minimum least squares estimation from residuals to align the various datasets.

I thought about that, but if they're all using the Sauret footage (which AFAIK they are) then they should be able to say "This unambiguous event happened in the time labled frame blah through frame blah blah in my data set" (say for example the big expulsion of fire below the impact zone).

No comfirmed details yet from OWE or Tony yet, though my recent data is great quality

http://femr2.ucoz.com/photo/6-0-232-3 (1192x721px/16.4Kb)

Feature Tracked:

Have placed the spreadsheet online for reference:
http://femr2.ucoz.com/load/1-1-0-22

OneWhiteEye · by **OneWhiteEye** on Sat Nov 07, 2009 5:27 pm

The forum has been hoppin lately. To be Churchill-esque, this is a posting rate up with which I cannot keep.

There are some requests for info, give me a chance to catch up and then compliance will follow soon.

Like the data, femr2; the looks of it anyway. Nice whack in there, can you visually check to see if the tracking is true over that hump?

SynthEyes seems very good. Probably need to drag myself into the 21st century, but the code-it-myself habit dies hard.

femr2 · by **femr2** on Sat Nov 07, 2009 6:25 pm

OneWhiteEye wrote:The forum has been hoppin lately. To be Churchill-esque, this is a posting rate up with which I cannot keep.

There are some requests for info, give me a chance to catch up and then compliance will follow soon.

Like the data, femr2; the looks of it anyway. Nice whack in there, can you visually check to see if the tracking is true over that hump?

SynthEyes seems very good. Probably need to drag myself into the 21st century, but the code-it-myself habit dies hard.

It's consistent across both upper and lower interlace frame traces, which would leave me to believe that it is indeed correct.

It's slightly later in descent than most traces, but a definite *whack*.

I'll post timeline correlation.

OneWhiteEye · by **OneWhiteEye** on Sat Nov 07, 2009 8:57 pm

Major_Tom wrote:Do they exist or not? Trippy, OWE, what are your current opinions?

80% yes, 20% no.

The upper block perimeter moves down, spandrels hammer the floor connections one by one and we see these little jolts along the roofline. Maybe?

Maybe.

OneWhiteEye · by **OneWhiteEye** on Sat Nov 07, 2009 9:08 pm

T_Szamboti wrote:Differentiating again to find acceleration would be the second derivative, and it isn't necessary to show whether or not deceleration occurred, so I don't understand why anyone is even talking about the third derivative.

Technically, a difference needs to be taken on the velocity to determine rate of change - but it is sufficient in this case to say whether a value is less than or greater to a prior value, subtly distinct. Visual inspection of a velocity graph plainly reveals slope changes, but that's doing differences with the eye.

Third derivatives on raw video data must be insane.

T_Szamboti · by **T_Szamboti** on Sun Nov 08, 2009 12:15 am

OneWhiteEye wrote:
T_Szamboti wrote:Differentiating again to find acceleration would be the second derivative, and it isn't necessary to show whether or not deceleration occurred, so I don't understand why anyone is even talking about the third derivative.

Technically, a difference needs to be taken on the velocity to determine rate of change - but it is sufficient in this case to say whether a value is less than or greater to a prior value, subtly distinct. Visual inspection of a velocity graph plainly reveals slope changes, but that's doing differences with the eye.

Third derivatives on raw video data must be insane.

You don't quantify what these purported small slope changes are, but it can generally be said that the velocity slope change would have had to be fairly high to show a transfer of momentum which would have been sufficient to cause propagation in a natural collapse.

Trippy · by **Trippy** on Sun Nov 08, 2009 12:26 am

T_Szamboti wrote:
OneWhiteEye wrote:
T_Szamboti wrote:Differentiating again to find acceleration would be the second derivative, and it isn't necessary to show whether or not deceleration occurred, so I don't understand why anyone is even talking about the third derivative.

Technically, a difference needs to be taken on the velocity to determine rate of change - but it is sufficient in this case to say whether a value is less than or greater to a prior value, subtly distinct. Visual inspection of a velocity graph plainly reveals slope changes, but that's doing differences with the eye.

Third derivatives on raw video data must be insane.

You don't quantify what these purported small slope changes are, but it can generally be said that the velocity slope change would have had to be fairly high to show a transfer of momentum which would have been sufficient to cause propagation in a natural collapse.

Context suggests that they're discussing the graph produced in this post which has been corrected for errors using the methods discussed over in the Sauret Data Synching thread.

OneWhiteEye · by **OneWhiteEye** on Sun Nov 08, 2009 12:27 am

T_Szamboti wrote:You don't quantify what these purported small slope changes are...

Sorry, I should have been clearer. I meant in general, decreases in velocity can be seen by the naked eye on a v vs t plot, if they exist. Excluding noise, any value that is less than a prior value indicates deceleration, if sampling interval is constant. Did NOT mean any specific data in this context.

femr2 · by **femr2** on Sun Nov 08, 2009 1:49 am

OneWhiteEye wrote:The forum has been hoppin lately. To be Churchill-esque, this is a posting rate up with which I cannot keep.

There are some requests for info, give me a chance to catch up and then compliance will follow soon.

Like the data, femr2; the looks of it anyway. Nice whack in there, can you visually check to see if the tracking is true over that hump?

Yes. It's rock solid. This graph compares the traces of the upper and lower field of the interlace frame. (My video is *unfolded*, so I essentially have two Sauret videos. The second one is 0.5*(1/29.970)s ahead of the other (half a frame).) (y axis values are in pixels, but you'll need to divide by 4 to get back to original video scale). It's velocity (pixels/frame)

SynthEyes seems very good. Probably need to drag myself into the 21st century, but the code-it-myself habit dies hard.

I'll try and put a video up somewhere of the actual trace, so it can be confirmed. Rock solid.

femr2 · by **femr2** on Mon Nov 09, 2009 1:27 pm

OneWhiteEye wrote:Nice whack in there, can you visually check to see if the tracking is true over that hump?

Further inspection reveals that the *whack* occurs at a point in time at which the tracked feature enters into shadow (which does distort the video image)

We may have to mark that portion of data as erronious.

OneWhiteEye · by **OneWhiteEye** on Mon Nov 09, 2009 3:41 pm

I'm familiar with the shadow problem. However, it may also be the big bump in Tony's data as well, and that's a different matter because it's not from the mast. Haven't checked closely but that's probably the location of a particular big bump in my data. More on that after a look.

femr2 wrote:We may have to mark that portion of data as erronious.

There are some potential mitigations involved, at least for a hand-coded solution. There is some metadata attached to my data, the purpose of which is to help assess the quality. With the mitigations for local illumination transients I have in mind, there will be more - even multiple channels of data for the same target. Some coding is in the works.

femr2 · by **femr2** on Tue Nov 10, 2009 12:13 am

OneWhiteEye wrote:I'm familiar with the shadow problem. However, it may also be the big bump in Tony's data as well, and that's a different matter because it's not from the mast. Haven't checked closely but that's probably the location of a particular big bump in my data. More on that after a look.

femr2 wrote:We may have to mark that portion of data as erronious.

There are some potential mitigations involved, at least for a hand-coded solution. There is some metadata attached to my data, the purpose of which is to help assess the quality. With the mitigations for local illumination transients I have in mind, there will be more - even multiple channels of data for the same target. Some coding is in the works.

With the upscaled footage I use, the distortion under shadow is not limited to illumiation, but a physical lower right extension of the pixel features (With a x4 pixel extrusion) which *bleeds* the tracking point down and right for a short period. The track remains true, but follows an artificially extruded position hook. Quite how to get around that I'm not sure (without re-positioning the track manually, which has all sorts of subjective problems attached)

That the latter end of the trace cannot be determined without using features higher in the frame (all of which suffer from similar shadow effects) is a tad unfortunate.

I'll keep working on the data to try and see if I can negate the shadow effects somehow. (Definitely need to nail the frame-sync so we can compare different data-sets properly.)

The 9/11 Forum

Missing Jolts found ???; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Re: Missing Jolts found; film at 11

Who is online