Chad pictures

[10538]

I assume that laser measurements were also done from that exact same spot?

Yes, same exact spot.  Sorry about the crappy camera.  It turns out 11aug's camera is 4x the res as mine.  Luckily I set it to the highest res or it would have been worse. 

[spf33]

The rest is easy:
0.0032 x 1600 = 5.24mm (effective horizontal image area size)
0.0032 x 1200 = 3.93mm (effective vertical image area size)
Horizontal FoV = 47.179°
Diagonal FoV = 57,254°
Vertical FoV = 36,267°

screenshot of max fov calculations;

[nekitamo]

I got a confirmation from the author of EXIFTool - he acknowledged that his FOV formula is not correct for 4:3 aspect ratio cameras, but as most professional cameras use 3:2 aspect and the error is below 5%, he won't exactly "rush" to fix it.

Also, I've just finished my calculations for the first image and although I'm still preparing the description of the second part for posting here, I believe there's no need to keep the results hidden in the meantime. Here's the calibration image with a perspective-correct 3D ruler drawn between the two reference points:



And here's the same ruler translated into the perspective of Chad's image based on the assumption that both small fins in line with the ruler are of the same length (4.2m):



As you can see, the final result doesn't differ much from my preliminary estimation - the outer diameter of the central ring is again close to 4 meters.

[spf33]

very preliminary here, just thought you all might be interested in the progress;:




i used all the regular exif and manufacturer info to set this up with the plan to transfer most of the assets to a scene using chad's original 1169x767 scannedimage.jpg.
anyone have any idea what 1169x767 might be natively?  i know all chad's image photos, minus the cell phone, are about the same resolution so i am wondering if anyone knows what native res they all are cropped from.

the 3d scene using chad's image, i'm really going to have to wing it as far as fov, ccd sensor size and lens so even a good guess on the native res will help.

[nekitamo]

Not sure what you mean, but Chad's images fit the profile of i.e. 4.0" x 6.0" (3:2 aspect ratio) photographs scanned at (1169/6=) 200 dpi, then slightly cropped to hide margins. In this case 1169x767 could very well be the original, unresized image produced by the scanner.

While I was calculating my ruler translations I deduced the horizontal field of view angle for #4 to be around 35 degrees, but I believe it would be better if you don't use my findings so we can have two independent analysis eventually producing the same results

[nekitamo]

Looking at Spf33's work reminded me about another thing I wanted to ask, not to mention that spf probably needs this information, too.

So, if you don't mind: exactly how tall are you, elevenaugust? 

[onthefence]

So, if you don't mind: exactly how tall are you, elevenaugust? 

I think you would also need to know 10538's height to determine the angle of the camera with respect to the ground.

Which leads me to the question of angles and elevation. Would it help in your calculations to use surveyors equipment for all this photo analysis. I mean to determine the angle of objects in the sky relative to their plane against the ground, you should be given the angle of horizon as well as the actual on-site field of view angles.

I mean, is there a camera or attachment that will give you the following information, blue = angle of elevation, red = dots or rings showing actual angles in the field.



Something like this would come in very handy especially for Stephens site, where low resolution and possibly inaccurate USGS elevation maps are relied upon.

[spf33]

heh.  nekitamo, it's like walking down the same path a few steps behind you.
i too was going to ask how tall tonio is but forgot to in my excitement to post that last chad image showing the tilted drone.

onthefence, heck yes equipment that gives height and angle would incredibly helpful i'm discovering.

still not sure if the usgs dem for stephen i'm using contains errors or not.

here's exactly where i downloaded:
http://data.geocomm.com/catalog/US/61069/1688/group4-3.html
and here's the link to the bit about positional errors:
http://data.geocomm.com/readme/usgs/dem.html

not sure how i can determine if the dem i'm using were created before 2001 thus were corrected and the errors introduced.  think i might have to email and ask.

*edit to add -  here's my post on the geocommunity website asking for clarification on the dem errors.  sent an email to their help email account also:

http://spatialnews.geocomm.com/community/boards/viewpost.php?cat_id=2&mess_id=9214

[elevenaugust]

So, if you don't mind: exactly how tall are you, elevenaugust? 

I was waiting for this question!!
My height is 1.77m 

[nekitamo]

Thank you

But, here's another thing... I did some tests with my camera to check the accuracy of FoV angle calculated from EXIF data, and the result is pretty discouraging:



I believe this is mainly due to barrel/pincushion distortion which also varies with focal length. I'll try to devise some means of "calibration" in order to deal with this anomaly...

[spf33]

My height is 1.77m 

i had you at 5 inches taller.
so i'll have to rotate the camera down a little which shouldn't change the needed drone tilt too much.

nekitamo, i'll also redo this scene with whatever fov estimations we can come up with.

...photographs scanned at (1169/6=) 200 dpi, then slightly cropped to hide margins. In this case 1169x767 could very well be the original, unresized image produced by the scanner.

how did you determine the photos were scanned at 200dpi?
also, looking at chad's ScannedImage-5.jpg the size is 1171, not that i think 2 pixels are going to make all that much difference but i would like to be able to use the native res.

i'm not sure i understand your fov graph, can you explain further?
since max also calculates the same fov's you are getting, that would mean max is susceptible to these errors(?).

[10538]

very preliminary here, just thought you all might be interested in the progress;:

Very impressive Spf.  I had no idea the drone would have to be so close to the tree.  Do you need any camera data from me?  Let me know.

[nekitamo]

how did you determine the photos were scanned at 200dpi?

Well, by simply dividing the number of pixels with the supposed image size. So if it was 6" wide photograph, 1169/6=194.8 pixels/inch = 200 dpi. If the photo was smaller, more dpi was used and vice versa...

i'm not sure i understand your fov graph, can you explain further?
since max also calculates the same fov's you are getting, that would mean max is susceptible to these errors(?).

Actually, I believe the error comes from the camera, and my camera model is known for its substantial barrel distortion at low focal lenghts. Here's how I think barrel/pincushion distortion affects the FoV angle:



Maybe we should ask for some "calibration" shots with various focal lengths of some kind of known-size grid at known distance from the camera?

[spf33]

I had no idea the drone would have to be so close to the tree. 

and i had no idea there was even any tilt just looking at the photo.
but it could be a bit farther from that tree than i have in 3d. 
i didn't even check the dimensions of the drone torus in the 3d until right now...4.2 meters.

unless a hoaxer did exactly what we are doing here, distance measurements, satellite views, terrain data, camera exif data, 3d photogrammetry, i can't see them getting these tilts correctly and consistently or even at all.  does that make sense? 

at this point, the tilting which at first seemed fishy is beginning to point to authentic photos for me...but this feeling may change.

hopefully chad's scannedimage-2 and 5 are consistent with scannedimage.jpg.

[10538]

and i had no idea there was even any tilt just looking at the photo.
but it could be a bit farther from that tree than i have in 3d. 
i didn't even check the dimensions of the drone torus in the 3d until right now...4.2 meters.

unless a hoaxer did exactly what we are doing here, distance measurements, satellite views, terrain data, camera exif data, 3d photogrammetry, i can't see them getting these tilts correctly and consistently or even at all.  does that make sense? 

Do you mean, with respect to shadows and sun angles?  Or are you talking about visual perspectives?

One thing I am working on is analyzing the sun angles on the drones between the different Chad pics and against the background.