Scan Objects Using Zed Open Capture

Software
1

Hi, I’m using zed-open-capture framework. What I’m trying to achieve is an application that scans objects at a 360 degrees( amputated legs, souvenirs or other static items. ) then convert it to a point cloud of the object.

So far I’ve been able to capture the image data and convert it to a satisfying depth map but I’m unable to align two separate depth maps of the image which are taken at different points. But I’m unable combine them. Process has to be done without human intervention so I can’t use point selected align method.

[!NOTE]
I’ve tried to use ICP algorithm but it didn’t produce any meaningful results so I guess that there has to be a initial rough aligning to make it work. But I couldn’t find a way to do it.

2

Hi @ates
Why are you using the ZED Open Capture library? Are you working on a host device that is not CUDA-capable?

Concerning point cloud fusion, normally ICP is the method to use, but it often requires an initial estimation of the camera poses to not fall into local minimum conditions.

3

Hi, excuse my late reply.
Yes I’m running on a machine which is not CUDA compatible. What would you recommend for me to initial guess.

4

You can use any PC with NVIDIA GPU, RTX is now recommended.
You can also use an NVIDIA Jetson device.