My use of Autodesk Forge Reality Capture
I am using Autodesk Forge Reality Capture to create precise 3D models of small target objects. These models are then further adjusted in CAD, for eventual 3D printing.
MY GOAL: I am trying to improve the precision of my 3D models by providing a survey file with known markers (GCPs) positions and I have consulted the API documents here.
My Questions:
- Which coordinate system string to specify given my (non-geographical) arbitrary origin imaging application?
- Is there is a way to provide cartesian coordinates locations in the < gpc > tag, and if so, what is the syntax?
- When I should define a location as a GPC point and when I should define a location as a "checkpoint"?
My Motivation for providing survey file to Forge
Without providing survey information, I am finding that the resulting Forge models are significantly distorted, even when I use images that provided good models previously with RECAP.
When I previously modeled with the RECAP software, I initially had similarly distorted models until I provided RECAP with my marker positions. After I provided precise marker positions to RECAP, I got excellent models.
So I expect that if I can provide the marker locations to Forge I will again get very accurate output models.
However, because my application is not geographical, I want to provide my marker positions to Forge in terms of either spherical coordinates or cartesian coordinates from my own arbitrarily defined origin point.
I can provide distances in millimeters, and angles in degrees.
I can't figure out from the documentation which coordinate system name to provide.
I believe that I could probably provide the spherical coordinates from my arbitrary origin in the < gcp > tag, if I knew which coordinate system is appropriate for altitudes measured in millimeters. But I can't determine which coordinate system string I need to provide.
I see the very lengthy list of known coordinate system strings referred to in the documentation, but they are just labels with no explanation of what they are.
I can provide cartesian coordinates, but I don't know which coordinate system string would be appropriate for that. And even if I did, the < gcp > tags in the example don't illustrate any x= y= z= format.
Can someone provide me more information on use of survey files for my application?
About my set up
I have more than 20 images from known vantage points all around a controlled target imaging zone.
All around the target zone I have fiduciary marker points, with at least 3 markers visible in each image, and frequently more.
For every marker I know which images that marker appears in, and at which xpixel and ypixel location it is centered.
BTW, in RECAP these locations were defined in terms of % of the frame, rather than pixel positions. That enables more precision than a single pixel. Is that no longer possible?
I also know the physical location of each marker in 3-space, relative to an arbitrary origin (0,0,0) at the center of my imaging zone. I have these locations in terms of cartesian coordinates in millimeters, as well as in spherical coordinates forms.