search

Applied-GeoSolutions / lidar2dems

Utilities using PDAL and PCL to create DTMs, DSMs, and CHMs from lidar data
89 stars 36 forks source link

Ground point classification in steep terrain #23

Open matthewhanson opened 9 years ago

matthewhanson commented 9 years ago

Need to verify if complex slope parameters only require more processing time, but don't produce worse results even in flat areas.

matthewhanson commented 9 years ago

I looked at a couple cases using different terrain types, and the larger the slope generally the higher the resulting DTM is (DSM is unchanged). Higher slope results in spots where the ground returns are higher (checked idw, min, and max - and they are all higher), even in flat areas. So there is an effect.

I think this remains an open question. Proper analysis would require more time, a variety of test terrain and runtime parameters (unfortunately classification takes a long time, so this will be a very computationally intensive question to answer properly).

F-Sullivan commented 9 years ago

I'm working on this issue for a subset of one of our problem sites. I tried to set the slope and cell size using the override, but I don't think it worked because the classified las files are named _s1c3.las regardless of the override values, and the results of the two different sets of parameters are identical.

F-Sullivan commented 9 years ago

The functions class_params and get_classification_filename (which calls class_params) in utils.py takes seems to be used to determine the filename and the slope and cell parameters. I'm unsure how the parameters dictionary works, but the name and slope/cell size is being determined incorrectly - not sure if this is using the except value in class_params, or if it is successfully trying to get the params in the try function (we classified as non-forest, flat, which has the same parameters as the except).