IndividualTreeDetection ignores the min_height parameter

[williamlidberg]

The output shapefile from IndividualTreeDetection contains points with very low values and sometimes even negative values even when the min_height parameter is set.

Here is my code:

# container directories
filtered_las_dir = '/tmp/filtered_las_dir/'
height_dir = '/output_tree_dir/'
min_height = 1

def main(laz_dir, tree_dir, min_height):

    print('create dem for normalization')
    start = time.time() 
    """
    converts all laz tiles in a directory to a digital elevation raster.
    """  
    wbt.set_verbose_mode(True)
    wbt.set_working_dir(laz_dir)
    wbt.lidar_tin_gridding(parameter="elevation", 
    returns="last", # A DEM or DTM is usually obtained from the "last" returns, a DSM uses "first" returns (or better, use the lidar_digital_surface_model tool)
    resolution=1, # This is the spatial resolution of the output raster in meters and should depend on application needs and point density.
    exclude_cls= "0,1,3,4,5,6,7,8,10,11,12,13,14,15,16,17,18", # Example of classified points to be excluded from analysis i.e. class 9 is water.
    minz=None,
    maxz=None,
    max_triangle_edge_length=None
    )
 
    print('Normalize point cloud.')
    for tile in tqdm(os.listdir(laz_dir)):
        if tile.endswith('.laz'):
            intile = laz_dir + tile
            filter_out = filtered_las_dir + tile
            height = height_dir + tile
            indtm = laz_dir + tile.replace('.laz','.tif')

            wbt.normalize_lidar(
                i = intile, 
                output = height, 
                dtm = indtm
            )
            
    print('Detecting individual trees from point cloud.')        
    wbt.set_working_dir(height_dir)
    wbt.individual_tree_detection(
        i=None, 
        output=tree_dir, 
        min_search_radius=1.0, 
        min_height=min_height, 
        max_search_radius=10, 
        max_height=50, 
        only_use_veg=False
    )

[jblindsay]

This is odd for sure. Are you able to send me a sample data set for which you encountered this issue so that I can do some testing and hopefully figure out the cause for the problem?

Just as a side note, I'm a little surprised that you're using WBT for Python scripting still given you have WbW. I find myself writing all my Wb-based scripts in WbW these days.

[williamlidberg]

Here is a link to my data.

The reason why I am not using WbW for this is that I am doing this for someone else. The code will live in a docker image that others can use and I am not comfortable using my WbW license like that.

[jblindsay]

Thanks for providing the data. Have you normalized your input point cloud using the NormalizeLidar tool beforehand? Otherwise the point z values are elevations and not height above ground, as you may desire.

That makes perfect sense why you'd be using WbT for this rather than WbW. I was just worried if there was some deficiency that you might have found in WbW as to why this application wouldn't be better suited to it. But I guess that makes a lot of sense. Thanks for sharing.

[williamlidberg]

This data is the raw version of my data. sorry for not mentioning that. I did use the normalise lidar tool before applying the individual tree detection tool. The code is in the original message above.

First I created a DEM from the laz file and then used that to normalize the data to elevation above ground. Finally I used the normalised data as input to the individual tree detection algorithm.

[jblindsay]

Silly me, I should have noticed that.

[jblindsay]

Taking your data set and normalizing it, then applying the IndividualTreeDetection tool, I must admit that I'm not seeing any output points that are less than 1 m, which was the value used for min_z. I can't seem to replicate.

[williamlidberg]

I did some more tests and seems to be related to setting a working directory instead of specifying an input file.
wbt.set_working_dir(normalized_laz_dir)

This works:

min_height = 1
intile = 'C:/william/filtered/18A003_61600_3750_25.laz'
trees = 'C:/william/18A003_61600_3750_25.shp'

wbt.individual_tree_detection(
    i=intile, 
    output=trees, 
    min_search_radius=1.0, 
    min_height=min_height, 
    max_search_radius=10, 
    max_height=50, 
    only_use_veg=False
)

But not this:

filtered_las_dir = 'C:/william/filtered/'
wbt.set_working_dir(filtered_las_dir)
trees = 'C:/william/test_output/'
min_height = 1

wbt.individual_tree_detection(
    i=None, 
    output=trees, 
    min_search_radius=1.0, 
    min_height=min_height, 
    max_search_radius=10, 
    max_height=50, 
    only_use_veg=False
)

I want to use the excellent parallel processing and edge "fixing" that your lidar tools use when working with multiple files.

[jblindsay]

That was the missing piece for figuring out the issue. Turns out that I was missing an if statement to filter out points with z < min_height for the batch processing branch. I've fixed it now and pushed to the repo.