Tweet
Help and advice is needed, please. I am starting a project to create 3D models of Viking-age weapons using 3df Zephyr Lite, but I am struggling to get a good reconstruction from my photos.
I am a beginner both with 3D photogrammetry and Zephyr. The object is photographed on a turntable in a light box, so lighting should be good. The camera is a point-and-shoot on full manual mode, mounted on a tripod, and on a shutter delay so there should be no camera shake. I tried both a Android cell phone camera (not as good images) and a fancy mirrorless camera (images about the same).
I started with some simple objects, such as paperweights with textured, detailed surfaces, and got excellent reconstructions. I tried some replica Viking weapons which I dulled (to avoid reflections from the shiny surfaces) with powder spray, and got very good reconstructions, although the sharp edges of the weapons were never as crisp or detailed as I wished - they looked like moths had been eating the edges, rather than smooth, straight edges.
I can't do anything to mess with the historical artifacts when I scan them. I can only scan them as is without any sprays or targets attached. I am unable to get good reconstructions with the one artifact I have to work with.
First, the results are inconsistent. I can run the same set of photos multiple times, with what I believe are the same settings, and the results range from no reconstruction to something that is recognizable. Typically, only a small number of cameras are accepted, and the sparse point cloud is just a jumbled mess with ghost points all throughout the volume, and further processing doesn't help. I use the deep preset from the default menu for reconstruction, and attempts to use advanced reconstruction settings has not helped much. And occasionally, the reconstruction from the same photo set isn't too bad, except for the rough edges.
Any suggestions on how I can consistently get a good reconstruction? An example set of photos of a spearhead is here in a zip file:
and the best reconstruction from that set is here (most of the time, the reconstruction failed from this set):
Many thanks.
I am a beginner both with 3D photogrammetry and Zephyr. The object is photographed on a turntable in a light box, so lighting should be good. The camera is a point-and-shoot on full manual mode, mounted on a tripod, and on a shutter delay so there should be no camera shake. I tried both a Android cell phone camera (not as good images) and a fancy mirrorless camera (images about the same).
I started with some simple objects, such as paperweights with textured, detailed surfaces, and got excellent reconstructions. I tried some replica Viking weapons which I dulled (to avoid reflections from the shiny surfaces) with powder spray, and got very good reconstructions, although the sharp edges of the weapons were never as crisp or detailed as I wished - they looked like moths had been eating the edges, rather than smooth, straight edges.
I can't do anything to mess with the historical artifacts when I scan them. I can only scan them as is without any sprays or targets attached. I am unable to get good reconstructions with the one artifact I have to work with.
First, the results are inconsistent. I can run the same set of photos multiple times, with what I believe are the same settings, and the results range from no reconstruction to something that is recognizable. Typically, only a small number of cameras are accepted, and the sparse point cloud is just a jumbled mess with ghost points all throughout the volume, and further processing doesn't help. I use the deep preset from the default menu for reconstruction, and attempts to use advanced reconstruction settings has not helped much. And occasionally, the reconstruction from the same photo set isn't too bad, except for the rough edges.
Any suggestions on how I can consistently get a good reconstruction? An example set of photos of a spearhead is here in a zip file:
and the best reconstruction from that set is here (most of the time, the reconstruction failed from this set):
Many thanks.
Comment