3D Reconstruction
I want to use a D435i to do 3D reconstruction . My target is small object like doll or things on table. My plan is: first I get pointclouds from D435i, then I calculate camera's pose and put pointclouds together. I don't know whether the plan is feaseble or not, and if there is any paper which discusses this or project which has done this. If you can recommend some papers or projects, it would be helpful to me.
-
It sounds as though you would like to use a single D435i to take multiple point cloud scans of an object and stitch them together into a single 360 degree model.
If you have a Windows PC, the easiest approach would be to use the free LIPScan 3D program, which can scan objects on a rotating turntable to generate a 3D model. It can export .obj 3D model files, .ply point cloud files or .stl CAD model files.
https://www.microsoft.com/en-gb/p/lipscan/9nc85cq8b1p9
https://www.lips-hci.com/product?product_id=16
If you have a non-Windows PC, using an external library may be the way to go, like in this example discussion:
https://github.com/IntelRealSense/librealsense/issues/1941
The subject of 3D scanning has also been covered on the RealSense blog. If you scroll down the blog page to a line with a list of topics, you can find all past Intel RealSense blog articles about scanning by clicking on the '3D Scanning' tag.
-
Pyntcloud for the Python language has an MIT Licence, meaning that you can freely modify and distribute it so long as you include the original author's copyright message in the code.
https://github.com/daavoo/pyntcloud
There is also CloudCompare, which has a GPL Licence and a point cloud Merge function.
https://www.cloudcompare.org/doc/wiki/index.php?title=Merge
Both Pyntcloud and CloudCompare have been shown to work with the RealSense 400 Series cameras.
-
You can also use the Point Cloud Library to combine multiple point clouds together.
https://github.com/PointCloudLibrary/pcl
If you know the approximate location of each camera (which you do with the D435i) you can use the Iterative Closet Point algorithm that is implemented in the library similar to this tutorial.
http://pointclouds.org/documentation/tutorials/pairwise_incremental_registration.php
Please sign in to leave a comment.
Comments
6 comments