A LiDAR point cloud is a dense collection of three-dimensional points, and on its own it is hard to use. Classification assigns every point a class, ground, building, vegetation, water or noise, and that classification is what turns a cloud into a terrain model, a canopy height map or a building footprint. QGIS can do the whole workflow without paying for commercial software.

Start with data. Free LiDAR is available from many national mapping agencies, the USGS 3DEP programme in the United States, and open data portals in Europe, and the files usually arrive as LAS or LAZ. Load the file into QGIS and check the point count, the extent and whether the cloud already carries a classification. Many public datasets are already classified, in which case your job is checking and refining rather than starting from scratch.

If the cloud is unclassified, use the Processing Toolbox. QGIS ships tools for ground classification based on progressive morphological filters, and the LAStools-style approach is available through the processing providers. A practical sequence is to classify ground first, then buildings, then vegetation by height above ground. Ground points are the lowest returns in each neighbourhood; buildings are flat, elevated clusters; vegetation is scattered returns above the ground with gaps.

Once classified, the useful products follow quickly. Filter to ground points and build a digital terrain model, subtract the terrain from the surface model to get a canopy height model, and measure tree heights, building heights and volumes. Style the cloud by classification in the 3D map view, using a different colour per class, and the structure of the landscape becomes visible at a glance.

Watch out for common problems. Noise points below the ground or floating in the air will corrupt a terrain model, so filter them first. Dense vegetation can hide the ground entirely, leaving holes in the terrain, and steep terrain can confuse automatic ground classifiers. Always validate against a known surface, a surveyed spot height or a road centreline, before trusting the numbers. Classification is an iterative process: classify, visualise, fix the mistakes, and classify again.