Low Resolution Wood Members scanning

[elongo88]

Hi,
we are every interested in implementing the NVIDIA Jetson in our Architecture + Design School at Virginia Tech. We want to start with a simple project. We need to do a fairly low resolution scan of wood members. For this we need: 1) detect the object, 2) detect its edges and the nodes that the pole has 3) We need to build a 3D model out of the edges and the nodes detected.

I have seen is possible to do 1) and 2) with the NVIDIA Jetson Nano. But what I am totally sure about is if we can extract/export the edges and the nodes detected, to build up the 3D model afterwards. I mean, we can take 16 photos of the pole (each taken after rotating the object 22.5 degrees) and with the edges of each photo we can rebuild a 3D model. That is enough data for us, for creating CNC toolpaths.

Is it possible with the Jetson Nano, detected the edges and convert them to bitmap or CAD file format?
Or would this be something that would have to code ourselves?

Thank you,

Esteban Londono
School of Architecture + Design
Biodesign Lab
Virginia Tech, Blacksburg, VA

[dusty_nv]

Hi @elongo88, from what you describe, I think this would be something that you would need to code up (i.e. using OpenCV, or PyTorch, ect)

To simply detect objects, you could train your own object detector like shown here: https://github.com/dusty-nv/jetson-inference/blob/master/docs/pytorch-collect-detection.md

To detect the edges of the wood members, perhaps a simple Hough line transform with OpenCV would work? https://learnopencv.com/hough-transform-with-opencv-c-python/

Wish you the best of luck with your project!