Scanning Workflow
Hello,
I’m using an Intel RealSense D455 RGB-D camera to scan indoor scenes, and I’d like some advice on the best workflow for capturing a room. Specifically, I’m trying to determine:
- Should I scan the room horizontally (panning across the room) or vertically (scanning from floor to ceiling)?
- What percentage of overlap between scans would be optimal to ensure proper stitching during post-processing?
Additionally, are there any official or recommended guidelines for indoor scanning workflows with RGB-D cameras? Any software suggestions or tips for achieving the best results would be greatly appreciated!
-
Hi Dimouzoi Instead of stitching together multiple individual views, it will be easiest if you move the camera around the room in real-time to build up a scan. If you have a Windows computer then a D455-compatible commercial software product called DotProduct Dot3D is well suited to this application.
https://www.dotproduct3d.com/realsense.html
https://www.intelrealsense.com/3d-scanning-an-environment/
A free alternative that can also build up a scan in real-time as the camera is moved around is the RealSense SDK's C++ program example rs-kinfu
https://github.com/IntelRealSense/librealsense/tree/master/wrappers/opencv/kinfu
If you would prefer to capture individual views and stitch them later then scanning horizontally left-right would be best, as the camera's view is wider in the horizontal than the vertical.
The overlap between views is something that would likely have to be worked out through trial and error experimentation. The image below of two RealSense cameras performing stitching into a single image from different angles may be a helpful visual guide though.
The image below provides guidance about approximately how much of a room a D455 could see at 1280x720 resolution.
Please sign in to leave a comment.
Comments
1 comment