CloudCompare forum

Hi,

I am very new to 3DMASC, but am trying to use it for ground point classification of two point clouds - one is lidar (L2) and one is photogrammetric. Currently, when I try to train the classifier, it seemingly takes forever and ultimately does not work. For the photogrammetry point cloud (pcx) I have segmented out ground points and non-ground points (like for CANUPO) and classified these in one cloud. For PC1 I use the point cloud I want to classify (although I have tried making it both smaller and less dense to see if it makes a difference).

Here is the parameter file:
# Defining labels for point clouds
cloud: PC1= # Label PC1 is associated with the input point
cloud: PCX=

# Core points
core_points: PCX

# Defining scales for neighborhood analysis
scales: 0.5;1;2 # Scales (neighborhood sphere diameters) for feature calculation

#FEATURES

# Z values with different statistical methods

feature: Z_SCx_MEAN_PC1
feature: Z_SCx_MODE_PC1
feature: Z_SCx_MEDIAN_PC1
feature: Z_SCx_STD_PC1


feature: ROUGH_SCx_PC1
feature: ANISO_SCx_PC1
feature: SPHER_SCx_PC1
feature: LINEA_SCx_PC1
feature: PLANA_SCx_PC1
feature: CURV_SCx_PC1
feature: Zmin_SCx_PC1

#feature: X_SC0_PC1
#feature: G_SC0_PC1
#feature: B_SC0_PC1

My goal is to optimize this method for ground point classification in steep alpine terrain with relatively dense vegetation, however the fact that I cannot make it run makes this process a bit difficult. I would love some help to solve this.

Also, if anyone has worked with 3DMASC for this type of terrain and have any tips, it would also be greatly appreciated.

Thanks!

What you are trying to do is not clear, please reformulate your objectives.
Please provide a subset of your data if you want some help (CloudCompare .bin file with point clouds explicitly named and their association with PCX and PC1).

I am trying to classify a point cloud into ground points and non-ground points (i.e. vegetation) with 3DMASC. I have both LiDAR and photogrammetry data and want to complete the operation on these separately. So far I have not gotten past the "Train classifier" step as it seemingly runs forever.

To make my PCX file I segmented out only ground points from my original PC and merged these together, and classified them with the Edit>Scalar fields>Add classification SF. I did the same for a selection of non-ground points, classifying my two clouds as 1 and 2, respectively and merging them into one. This is because the PCX requires a classification field.

For the PC1 I use an unclassified point cloud of the same area. I use the parameter file I have provided and the default settings for max depth, etc.

Have I fundamentally misunderstood how this plugin works?