Removing outliers (noise) from Lidar (point cloud) data in large datasets

[Fabio_Secanho]

Hello,

I'm processing a LAS dataset with 1,200+ tiles and about 500gb. ArcGIS Pro is doing fine creating rasters based on models (e.g., DSM, DEM, etc.). However, I need to remove outliers from this dataset. When testing different tools, the "Classify LAS Noise" produced the best results. My project area is mostly covered with forest, so my understanding is that "Isolation" is the most adequate method here. The problem is that this operation is taking a massive amount of time. Perhaps more than weeks if I let it run for all tiles (my computer is top-notch). Do you have any suggestions on how to proceed? Is this the only approach to properly remove outliers in a systematic way for the full area?

Please, see some images of the outliers. Most of the area is fine, with a few outliers here and there. These areas of the images are the worst by far.

Thank you very much!

[PrajwolSubedi]

Hi, I am also facing similar issues. Did your problem get solved?

[CodyBenkelman]

sorry I did not see the original post from Feb 2021 - I assume Fabio found a solution?

This is the correct GP tool to use - but it should not take days or weeks to process 1200 tiles.  I'd suggest making a LAS dataset with a small number of tiles to test the processing time.

Cody B

[PrajwolSubedi]

I am encountering a similar challenge with the presence of outliers in the data, which I am attempting to address using ArcGIS Pro. To mitigate this, I have employed the PDAL translate tool to remove class 7 data, which is typically considered noise. Additionally, I have utilized the 'Classify LAS Noise' tool. However, both approaches have led to a significant reduction in the quality of the resulting digital surface model compared to the original data.

Can you give me the solution?