[GIS] Classifying LiDAR data to mid vegetation using LAStools

classificationlastoolslidar

I want to classify LiDAR vegetation points in low and mid vegetation, using LAStools.

High vegetation is simple to get, using top of canopy (first return), but how about low vegetation, and mid? Following ASPRS standards they are class 3 and 4, respectively. How could I separately isolate these classes using LAStools, in order to create DEMs from them?

Best Answer

You can use lasheight to classify points at different intervals above ground, assuming you have already classified ground points. For example 0-1 meter above ground -> low veg, and 1-6 meter above ground -> med veg. The height values for these 2 classes vary from project to project, as there is no set rule. You can use the -classify_between option to set your tolerances. See the readme for more details.

For example, to keep only points between 1 and 6 units above the ground, and save out to 'mid_veg.las', enter the following command:

lasheight -i lidar.las -o mid_veg.las -drop_below 1.0 -drop_above 6.0

Related Solutions

[GIS] Classifying LiDAR data with LAStools

LASTools can perform a ground classification using "lasground" and then can perform some limited feature classification using "lasclassify". The performance and quality of feature classification in point clouds is strongly influenced by the type of landscape collected. Some landscapes just do not lend themselves to acceptable automated results. The best feature classification is provided when the vendor performs a "hands on" approach where considerable editing is performed after a given ground/feature classification algorithm is applied.

I am not sure how many tools are implemented in the ArcGIS LAS Toolbox. You may want to look at the available LAS commandline executable tools for more detailed analysis.

If your data was collected in an area with forest canopy or topographic relief I would recommend using the MCC algorithm for ground classification. This method uses a Thin Plate Spline and produces very good results. I am not overly fond of TIN based methods because of the surface bias, distortion and noise that they tend to produce.

[GIS] Classifying LiDAR ground points using laspy

I was able to come up with a solution. The following code works, but it feels like there is a much simpler way to do this operation. If anyone has a cleaner or more efficiect way to filter and reclassify lidar with laspy I'd love to accept that answer.

import laspy
import numpy as np
from scipy.spatial import cKDTree

in_file = laspy.file.File(input_file, mode = "r")
point_records = in_file.points.copy()

#filter all single return points of classes 3, 4, or 5 (vegetation)
single_veg = np.where(np.logical_and(np.logical_or(np.logical_or(
                                                   in_file.raw_classification == 3, 
                                                   in_file.raw_classification == 4), 
                                                   in_file.raw_classification == 5), in_file.num_returns == 1))

# pull out full point records of filtered points, and create an XYZ array for KDTree
single_veg_points = in_file.points[single_veg]     
single_veg_points_xyz = np.array((single_veg_points['point']['X'], 
                                  single_veg_points['point']['Y'],
                                  single_veg_points['point']['Z'])).transpose()

#filter class 2 points (ground) create XYZ array for KDTree           
ground_only = np.where(in_file.raw_classification == 2)
ground_points = in_file.points[ground_only]

ground_points_xyz = np.array((ground_points['point']['X'],
                              ground_points['point']['Y'],
                              ground_points['point']['Z'])).transpose()

#create a KDTree to query against
ctree = cKDTree(ground_points_xyz)

#For every single return veg point query against all points within 20 meters.
#If a higher elevation ground point is nearby, then change the classification 
#from vegetation to ground in the original point array.
for idx, record in enumerate(single_veg_points_xyz):
    neighbors = ctree.query_ball_point(record, 2000)
    for neighbor in neighbors:
        neighbor_z = ground_points[neighbor]['point']['Z']
        record_z = single_veg_points[idx]['point']['Z']
        if neighbor_z >= record_z:
            single_veg_points[idx]['point']['raw_classification'] = 2

#update points just once
point_records[single_veg] = single_veg_points

out_file = laspy.file.File(output_file, mode = "w", header=in_file.header)
out_file.points = point_records
out_file.close()

Best Answer

Related Solutions

[GIS] Classifying LiDAR data with LAStools

[GIS] Classifying LiDAR ground points using laspy

Related Question