CloudCompare forum

Hi Daniel and all users,

First, thanks for this incredible work and generosity it is quite inspiring!

My question: Is there a way to specify more precisely the cell height we want on the rasterize tool (or a way to work around it to get the same results)?

Explanations:

If, for instance, we want to generate a Digital Terrain Model (which I can see in the forum many people use the ccRasterize tool for) we select to keep the minimum height in each cell. We then have a lot of chances that some extremes values will be given to some cells, which are not representative of the cell general's ground height . I think we could get closer to the "real ground values" if we could select the 5th percentile of the height value in each cell or something similar.

We would also get better results for Digital Surface Models by using the 95th percentile rather than the maximum height.

Simlar request had been done here https://www.danielgm.net/cc/forum/viewt ... ell+height I dont' know if there's has been some updat since?

I totally understand that this would require more calculation effort and I whish I had the programmin skills to implement it myself, but it's unfortunatly not the case :(

Maybe there's already a tool in place that can allow to work around and get the same results? Or we can add it to the ToDo list?

Thanks again!

Indeed, I'll have to add this to the TODO list.

Meanwhile, have you tried the qCSF plugin to extract the bare earth?

Best

Daniel

Yes I've tried CSF plugin but I get better results selecting the minimum cell hight with ccRaterize tool.

When I compare the DEM's generated by the algorithm and ccRaterize tool with 300 control points for which I have an exact ground elevation, I get the following results:

RMSE (CSF cloth res=4m) : 0.427502484
RMSE (CSF cloth res =0.5m) : 0.522642096
RMSE (minimum cell height with ccRasterize in a 4m grid): 0.322222492

Med error (CSF cloth res=4m): 0.338998
Med error (CSF cloth res =0.5m): 0.449
Med error (minimum cell height with ccRasterize in a 4m grid): 0.164

So in this specific case ccRasterrize seems to do a better job extracting an accurate DEM than CSF. We would get even better results with 5th percentile heigth :) !

Thanks for the quick reply!