I have four depth cameras on a device and their relative positions are fixed. The cameras can perform simultaneous scans on an object from four different angles.
I'd like to register four images together as one complete 3D scan. Currently I'm manually registering the four scans together based on the overlapped area between the scans. Are there any methods to do the registration automatically? For example, an algorithm that can detect and extract the overlap area, then align the scans? And I might prefer a non-learning method as I don't have a GPU currently. I tried reading a few papers, but I'm not sure which one would be the best fit for me. Do you have any recommendations?
Supplementary information about the four scans:
- The overlap area between scans is not big, taking about 5-10% area of a single scan.
- Since the cameras are fixed, the overlapped area of the object can be determined.
- Special markers/stickers can be attached to the overlap surface of the object(like a colored sticker or a QR code?) if it will help with the detection.
Many thanks!
