Video or pictures for making a model of a small object

[Ejmidd2913]

Hi All, Brand new here. Love the software. Have been wanting to do photogrammetry for a while and with quarantine finally found the time.

??????I have a quick question and sorry if its posted here or in a tutorial I searched but couldnt find anything. Is it better to take individual images or extract images from video? Also does it affect quality using 2 different cameras?

I am photographing old toys (small cars) on a turntable using a 16mp DSLR with macro lens and a Samsung Galaxy S10 (12mp i think )so I can take 2 pictures at once (phone and DSLR). Putting everything to the software to create the model and then exporting it a a .stl file to 3d print. Sucessfully printed 2 models with varying degrees of detail. I am taking pics at about 4-5 different angles every 10 degrees. I was going to try extracting imaged from video to go thru the process quicker but I read video is not as good for getting images. My camera can do 30fps my phone 60fps in ultra HD if that matters.

Before I spend hours testing it wanted to see your thoughts or suggestions #1 if my process is overkill for what I am doing #2 if 2 different cameras make a difference in quality and #3 if images extra from video with my equipment will make a huge difference. Thanks in advance!

[cam3d]

Hi Ejmidd2913 welcome to the forum! Great to hear you love the software

To answer your question:

As a general rule, if you can take still images rather than video, you're going to have better 3D reconstruction results. This is because video adds an extra layer of complexity and frames still need to be extracted from the video file to process the data. Video files are typically lower resolution than stills, and can have detrimental elements like interlacing and motion blur etc...

Personally I like to stick to one camera for the sake of image consistency, however 3DF Zephyr can definitely handle multiple inputs.

If you do an A-B comparision of single camera vs your dual camera setup, you'll be able to identify if two camera types works better for you (weighing up time/quality/etc....) If the time savings make sense and there's minimal detriment, by all means use both!

4-5 elevations (angles) and 10 degree increments sounds about right for most small objects, if you have a object with lots of occlusions/thin elements you might want to increase the increments to 7.5 degrees, or, if you have very simple objects with minimal occlusions/thin elements, you can get away with 15 degree increments. If you have a slight offset between each elevation and their given increments, you can get a bit more coverage on objects.

What is very important is having everything you want to reconstruct covered by 3 images, giving lots of reference information for Zephyr to work with.

Try to get the whole object in frame so you make the most of the camera sensor/pixels, and if you want more detail, you can then supplement this 'global' pass with tighter framing close up on the surface.

It's worth noting that having enough overlap between images side by side is just as important as not moving in too far on your subject for this secondary close pass. If you zoom right in, the details and feature points Zephyr work with might change too dramatically to be able to successfully orient all your cameras.

There's a lot of trial and error to get things exactly how you want them with many variables to account for. I hope this has been informative and if you have any specific questions feel free to ask!

[Ejmidd2913]

Thank you very much for the response. I will try the single camera method and no video and see what happens!