Merge color image

Q&A
,
1
  • Mech-Vision : 2.1.2

I know that there is a step in Mech-Vision to merge point clouds, but is there a way to similarly merge color images as well?

The reason I’m asking is because I would like to use deep learning for depalletizing.

I am planning to either use two cameras or use a single camera to take images from two different positions with precise movement measurement.

Thank you very much.

2

Hi @K.Y,

Apologies for the late response.

Currently, Mech-Vision does not have a built-in Step specifically for merging 2D color images the way Merge Point Cloud works for point clouds.

However, here are a few approaches for your depalletizing deep learning use case:

  1. Single camera, single position: If the field of view can cover the entire bin, using a single capture is the simplest approach.

  2. Use the merged point cloud for DL: Mech-DLK models can work with point cloud data. If you merge point clouds from two captures, you can run the DL model on the merged 3D data.

  3. Custom Script Step: You could use OpenCV in a Custom Script Step to stitch two images together, though this requires calibration of the camera positions.

Could you share more details about your setup (working distance, bin size, why a single capture is insufficient)? This will help us suggest the best approach.

3

Hi! I noticed your question has received a reply. Has your issue been resolved?

  • If resolved, please click the “Solution” button below the reply to mark it — this helps others with similar issues :pray:
  • If not yet resolved, please share an update and we’ll continue to follow up.