lidar-processing

Using Python to process, filter, and classify LiDAR point clouds. Also supports mesh generation for 3D modeling.

Setup

1. Clone this repository: git clone https://github.com/stanley-wei/lidar-processing.git
2. Install the required packages (requires Python3): cd lidar-processing && pip install -r lidar-processing/requirements.txt

Usage

All LiDAR classification codes are as specified here.

Ground & Feature Extraction

python3 -m lidar-processing.classification.ground_extraction <LiDAR_FILE>
python3 -m lidar-processing.classification.feature_extraction <LiDAR_PATH>

The script lidar-processing/classification/ground_extraction takes as input a LiDAR .las/.laz file and outputs a .las/.laz with accompanying ground/non-ground class annotations.

The script lidar-processing/classification/feature_extraction takes as input a LiDAR file (or directory of LiDAR files) with classified ground points and outputs .csv file(s) containing a set of extracted features for every point. (One .csv for each input file.)

(Use python3 -m lidar-processing.classification.ground_extraction --help and python3 -m lidar-processing.classification.feature_extraction --help to view additional options.)

Classification

Training:

python3 -m lidar-processing.classification.train <DATASET_PATH>

This takes as input a directory of classified LiDAR .las/.laz files and trains a model to classify point types.

(Has classifier options; see python3 -m lidar-processing.classification.train -h for more details.)

Testing:

python3 -m lidar-processing.classification.test <DATASET_PATH> <CLASSIFIER_PATH>

This takes as input: (1) a directory of classified LiDAR .las/.laz files and (2) a joblib-pickled classifier with a function .predict(), then evaluates the performance of the classifier over the dataset.

3D Modeling

python3 -m lidar-processing.scripts.interpolate_and_mesh <LiDAR_FILE>

The script lidar-processing/scripts/interpolate_and_mesh.py takes as input a classified (i.e. split into buildings, ground, etc.) LiDAR .las/.laz file, and will output a 3D mesh. Supports modification of mesh output via image masking.

[Note: Many GIS applications support classification of LiDAR data; I personally used CloudCompare and LASTools for interfacing with LiDAR files.]

(See python -m lidar-processing.scripts.interpolate_and_mesh.py -h for more details)