CloudCompare forum

I'm currently manually segmenting some concrete pads at a camp ground, and it got me thinking: why can't we use the RGB or Hex notation color codes for batch selection?

Similar to functionality found in image editing software like paint.net or even photoshop: there is a "magic wand" pixel selection tool of which the user can modify the sensitivity for. You click on the area where the color is and it selects all the matching pixels. Now, you can enable it to select only those that are neighboring OR you can set it to apply the tool to any pixels within the image with the same/similar code values.

With a point cloud dataset this would be very helpful, especially if you're wanting to segment these to a subset for classification/training. This could be applied to things like asphalt, pavement markings/painted objects, buildings...really anything that is artificial to the environment. It would make it a little easier than just clicking around polygons which - sure - is a useful tool but can be time consuming if things are far apart or there are a lot of instances.

Anyway, thought I'd request it.

Ah, that's definitely an interesting feature but it's also tricky to implement ... I'll add this to the TODO list, but I wouldn't be too optimistic ;)

Consider checking

Plugins--->Colorimetric Segmenter--> Filter RGB

PablerasBCN wrote: ↑Thu Jun 13, 2024 10:42 am Consider checking

Plugins--->Colorimetric Segmenter--> Filter RGB

Ah-ha, now isn't that something....this is why I love this system, it has so much functionality built into it. It is a little hidden at times, but I suppose that is why this community exists. While this will take some practice and familiarization with what works and what doesn't, this is certainly another great way to segment some subsamples that could then be used for training.

Thanks!

glad it helped.

I had to segment asphalt from a mobile mapping system in the past.

Another technique I used was to using the histogram clustering try to pack the similarly looking colors for easier selection. You could also convert RGB into scalar field and with then use then use the histogram to visually and in realtime try to isolate the area of interest. The issue with the RGB segmenter tool is that is a bit of a blind shot.

I also tried to build a workflow where I would rasterize hte ground and export as tiff, in photoshop use the magic wand etc to create a mask. Bring back the tiff as geotiff (need to rebuild the geo data with a tool form the original tiff export) etc.. but was a bit too clunky workflow...

One day someone with skills may bring 2D segmentation into this program and project the ressults into the 3D cloud using models like segment anything or so... meanwhile...
viewtopic.php?t=6300

edit:seeing this oldy post I made time ago I see people added some links to projects using this for 3D!