I have a stationary scanning booth, made up of 19 poles with 6 to 7 raspberry PI cameras each.
I have previously used Agisoft Photoscan and use markers from this software, with which I could automatically make a Bounding Box and the alignment results were really good.
Now, I use Reality Capture (which I like more than Agisoft), I found that there was an experimental option of markers. Can I use them to automatically generate a bounding box? If not, can I generate control points that help me align automatically and efficiently?
The distance and position of cameras and markers are always same.
About this problem check the Re-use Alignment in the Help section of your software or you can find some information in this video: https://www.youtube.com/watch?v=5VyYLaNxHz0
If your cameras position is always the same, you can export the reconstruction region from one project and use this region in other new projects.
I would like to continue to use Reality Capture, but at this point in time, I am not aware of any way to use markers to automatically construct a reconstruction region with limitations of my rig…but I would like to verify what I believe to be true. If the camera position is static, I can export either control points, ground control points, and reconstruction box…or I can use the xmp method, exporting the xmp file…in either case, I reimport that data when processing a subsequent shoot, and camera locations and angles are exactly the same, the reconstruction zone will be “inserted” to the desired boundaries. However, my rig is mobile, built in a big custom trailer…the markers would be fixed permanently to the floor, but the columns ride in sleeves with springs above/below to minimize impact of a rough ride…so they move slightly. Each column of cameras can have slightly different angles/locations when moved to each location…a column might be .25 cm to the right or left or the column might be rotated a few degrees right or left due due to freedom in the mounting system. I also have several cameras that are adjusted per shoot…for example, a “rack” of cameras set to capture the face, based on the subject’s height. About 20 of my cameras have adjustable tilts that can be modified…all tilted lower if I’m doing a shoot for a youth sports team, or i might adjust several upward if the group it overall taller on average. I’m looking for help in figuring out a way to automate processing. Ideally, I would be able to print 3 targets, let the software identify them and name them (i.e., “cornerA”, “cornerB”, and “CornerC”) then enter a real world measurement from the center of corner A to Corner B, Corner B to Corner C, and a height.
Question… if I go to each individual camera registration and change relative position uncertainty relative to what I believe would be accurate depending on what actually happens with each camera, can I export the rig, bounding box info, point control info, xmp, etc. and expect the bounding box to work, even though the cameras move? Reality Capture is by far the best package available, and I’m surprised it doesn’t offer a better solution for automatic bounding box creation based on marker location…
The best way to use markers/control points to create a reconstruction region is to set the reconstruction region on control points. You can use that manually in RealityCapture in the Set Reconstruction Region dropdown, or through batch file (command line).
If you are doing it manually, use this tool and click on three control points and height of the reconstruction region can be set to anywhere. You can change it later in MESH MODEL tab / Info panel to a certain value.
If you want to use command line, there is a command: setReconstructionRegionOnCPs. Contact me on support@capturingreality.com and I will send you a CLI sample to see how to do it.