Outdoor Photogrammetry: Issues and Questions

[kagidusko]

Hello all,

 

I am currently trying to test the use of CRP for outdoor forensic scenarios in common SE U.S. environments like oak hammocks and pine flat-woods. I am using plastic teaching skeletons to represent human remains scatters of varying types. I have a couple issues I was hoping folks could weigh in about and also wanted to see if anyone had any good articles/books I should maybe dive more into that might help on this particular subject:

 

• It is incredibly hard to walk in between the scatter without moving anything. Would it be worthwhile to set up a tall tripod/monopod (8ft or so) and shoot around the scene? I feel as though I'd be losing too much detail, and accuracy of the scene is key. Currently I am shooting about 5.5-6ft off the ground, eye-level height for me for the most part.
• I am shooting with the aperture priority setting, iso 100, in RAW format. Using a Canon Rebel T5i and shooting at about 23-24mm, depending on what I set it too before taking auto focus off. I have also completed some test runs using the CHI scale bars. I'm still getting a little blur on details, so much so that I can't really detect marker points on about half the images at times. Should I be shooting differently?
• I'm often having to shoot in weird patterns/positions as we're trying to make the scenes as real as possible. Am I tanking accuracy by moving all over the place?
• When processing, for the widest scatters, I have had to "detect markers" within the program prior to the initial photo alignment. Does this drastically alter error to do so before a sparse point cloud is generated?
• I'm finding that the models are getting super complex with all of the leaf and pine needle litter. Is there any way to control for that without loosing accuracy of info on "bones"? 
• I am getting some wavy-ness along some of the bones, or it appears that they are in a haze. Any idea what might account for this and how I can correct it?

Thanks in advance for any insight!

 

 

 

[leszekp]

You might try shooting at a slightly higher ISO (up to 400); you'll get faster shutter speeds to reduce the blur, and the added noise is likely to be small. You might also try processing the dense cloud at UltraHigh to get better resolution for your models. Using a camera pole can get you higher camera elevations, for more angles. With a faster shutter, you can try holding the camera pole obliquely to get good angles above the scene without disturbing the area.

[kagidusko]

Here's a medium run, decimated version of what I'm trying to do, if it helps to see. Am I just not going to have a lot of definition because of the complexity of all the pine needles? Anything I could have done better besides the advice above?

 

https://sketchfab.com/models/dcd3b12c1ec844ae84ea5dc531f9987b

 

Pass: thesis

[cdschroer]

Leszekp has some good advice.  In addition, I'd add that Marlin developed a harness system that keeps the camera at about 5-6 feet off the ground, and away from your body.  We have used it to walk around areas we want to capture.  You can see a picture of it here: https://www.facebook.com/culturalheritageimaging/photos/a.356002304439118.78214.127514043954613/1469987609707243/?type=3&theater

 

The sketchfab model looks to be really decimated, so it's hard to tell what's going on.  How many shots did you take?  What's the resolution of the camera?  And are you getting enough look angles?  In my experience the "wavy" pattern on a surface is an indication of not enough look angles on that part of the model. You can also run a report in Photoscan and look at the color coded chart for look angles.  It IS tricky to walk around and get this right, but it can be done, with some planning and practice. The pine needles do make it harder, so you want to keep outside of your area of interest as best you can.  You could use a longer pole than what is shown in the image above.  We have successfully captured an areas about 10 feet by 10 feet using the length pole in the photo, and keeping outside of the area we are shooting.

[kagidusko]

Carla,

 

Thanks for the response!

 

I've since re-run that set on high, though still having similar issues. Link to model HERE and password is: thesis. Unfortunately, re-shooting isn't an option at this point, but I'd like to know what I'm doing wrong to discuss in my further research section and to try again down the road. At some point I think I'll have to put together Marlin's contraption-it does indeed look like it would do the trick and keep me out of the area. Marlin, if you see this, let me know what you'd charge to put one together! 

 

Shot with a Canon RebelT5i

Focal length 23mm 

146 images (shot in raw, processed to tiff)

18 megapixel

 

I thought I got enough look angles based on what I saw in the report, but I may be misreading it. I've uploaded HERE. The report is for the original, non-decimated model. 

 

Thanks for the insight!

[cdschroer]

Kevin -

 

I've looked at this a bit.  I think from the spread of photos shown in the report, that there may be some issues in the middle section of the area shot, due to how spread out the images are. Overall, I think this is a tricky scenario because of the moving pine needles, and also the shape of the bones.  You could do a pass over the top (as you did) and get the placement of the bones and probably decent measurable data on the length of bones. You would likely ned additional shots at more inclined angles to get more of the bone structures themselves.  I'm not sure if this is clear, and part of the question is what your goals are in doing this type of capture?  Would you go back and image the bones separately "in the round" or at least some of them - and the initial scatter is to document how they are laying in the ground in a measurable way?  Overall, my biggest piece of advice would be to get a pole and harness and try it again (as I already mentioned)

 

Carla