I am currently working on point cloud data analysis, trying to label objects which are not ground or vegetation e.t.c. So far I tried many clustering algorithms, with moderate success.
In my best model I did the following: clustered data in colorspace (r,g,b), clustered data in x,y,z coordinates and then combined these two models where points which belonged into the same two cathegories were merged into the one super-cathegory. K-means and mean-shift did both equally good, with K-means being much faster.
My question is, am I on the right track or there are much better solutions to this? Are there some algorithms(ideally as part of some major python machine learning library, but I am also open to use c++ or implement it myself) which work well to do something like grouping points together by color and position, ideally not in separate way (like I did)?
P.S.: Keep in mind I work with very large datasets (~100M points), so good performance is essential to me.