Point Cloud vs Mesh: What's the Difference and Which Do You Need?
A point cloud and a mesh can come from the same drone flight and look alike on screen, but they store different things. A point cloud is millions of separate measured points. A mesh is a continuous surface of triangles stretched between points. Knowing the difference tells you which one to compute volumes from, which one to show a client and which one to send to a web viewer.
What a point cloud stores
A point cloud is a list of 3D points, each with X, Y and Z coordinates and usually more: a colour from the photos, a class such as ground, vegetation or building, and for laser scans an intensity. There is nothing between the points. Zoom in far enough and you see through the cloud, because it is a set of samples, not a surface.
That is its strength for measurement. Every point is an observation, classes let you separate the ground from what stands on it, and a bare-earth surface can be built from the ground class alone. In photogrammetry the dense cloud is the first full product; the mesh, the elevation models and the orthophoto are all derived from it.
What a mesh stores
A mesh is a set of vertices joined into triangles that together form a closed or open surface. It can carry a texture: images wrapped onto the triangles so the model looks like the real site from any angle. Because the surface is continuous, a mesh renders as a solid object, light and shade work on it, and it stays readable when you zoom in.
The triangles are an interpretation. Between two vertices the mesh assumes a flat face, so it fills small gaps and can also smooth over real detail. Per-point classes are usually not part of a mesh, which is why ground filtering is done on the cloud, not on the model.
How a mesh is built from a point cloud
The step is called surface reconstruction, and there are two families. A 2.5D surface keeps one height for each position on the ground, like a terrain model or a TIN; it is the right model for ground, stockpiles and pit floors and is simple to triangulate. A full 3D surface can describe vertical walls, quarry faces and overhangs; it is usually built by fitting an implicit surface through the points and extracting triangles from it.
Stream Capture produces both from the same dense cloud: a surface mesh for terrain work and, by default, a full 3D surface mesh that can follow walls and overhangs. Each chart of the model then takes its colour from the photo that sees it best, judged by projected area, how central it is in the frame and sharpness, and charts are colour-balanced across seams.
Point cloud vs mesh at a glance
The two are not rivals. Most projects keep the cloud for measurement and the mesh for viewing, and both come out of the same processing run.
| Point cloud | Mesh | |
|---|---|---|
| What it stores | Separate 3D points with colour and class | Vertices joined into triangles, often with a texture |
| Between samples | Nothing: gaps stay visible | A flat face: gaps are bridged |
| Ground classification | Per point (ground, vegetation, building, noise) | Usually none |
| Best for | Elevation models, volumes, ground filtering, change | Presentation, inspection, web and 3D viewers |
| Usual formats | LAS, LAZ, COPC, PLY, CSV | OBJ, GLB (glTF), PLY |
Which one should you use?
Compute volumes, contours and cut and fill from the cloud or from the DSM and DTM built from it: those are the measured products, and the ground class is what makes a bare-earth model possible. Use the mesh where shape and appearance matter: showing a site to a client, inspecting a structure, placing a model in a web or 3D viewer. If walls matter, fly oblique photos as well as a top-down grid, because a straight-down camera sees very little of a vertical face and neither the cloud nor the mesh can show what no photo saw.
Frequently asked questions
What is the difference between a point cloud and a mesh?
A point cloud is a set of separate 3D points with colour and often a class; there is nothing between them. A mesh joins points into triangles to form a continuous surface, which can carry a photo texture.
Is a mesh more accurate than a point cloud?
No. A mesh is derived from the cloud. It can smooth out noise, but it also bridges gaps with flat triangles and can round off real detail, so the cloud remains the measured product.
Can you convert a point cloud to a mesh?
Yes. Surface reconstruction turns a cloud into a mesh, either as a 2.5D surface with one height per position or as a full 3D surface that can follow walls and overhangs.
Should I calculate volumes from the point cloud or the mesh?
From the cloud, or from the DSM and DTM built from it. The ground class in the cloud is what makes a bare-earth surface possible, and the elevation models are the measured grid.
Is a TIN a mesh?
Yes, a TIN is a 2.5D triangle mesh: each position on the ground has one height. It describes terrain well but cannot describe a vertical wall or an overhang.
Which file formats are used for point clouds and meshes?
Point clouds are usually LAS or LAZ, also COPC, PLY or CSV. Meshes are usually OBJ or GLB (glTF), and PLY can hold either.
Build a classified point cloud and a textured mesh from one flight
Native Windows · fully offline · free beta