If you work in the world of 3D, you've probably encountered the wall that is trying to include LiDAR scan data in a game engine. The problem is that LiDAR doesn't create a solid object, but rather throws millions of points into space, which for a graphics engine is basically a chaotic mess of coordinates with no surface to render.
Converting a LAS or binary file to an OBJ is not simply changing the file extension, but rather performing a process of surface reconstructionIt's not an easy task, and if you overdo it with the smoothing algorithms, you can end up with models that look like blurry smudges instead of a detailed city or mechanical part.
Understanding the raw material: The Point Cloud
To begin, it's important to understand that a point cloud is, literally, a bunch of X, Y, and Z coordinates. Unlike a CAD model, It has no faces or edgesTherefore, there is no concept of volume. It is a discrete representation; between one point and another there is an absolute void that the software must fill.
Depending on how they were captured, this data can be structured, such as those of terrestrial scanners that organize information into precise grids, or unstructured, typical of mobile or wearable scans that require much more intensive post-processing to remove noise and align features.
In the case of drones, the data usually comes in LAS or LAZ formats, optimized for geospatial applications. These files typically require a previous classification to separate, for example, what is vegetation from what are buildings, before trying to generate a grid that makes visual sense.
The technical path: From Points to Polygonal Mesh
The first major step is converting the cloud into a mesh, usually in STL or OBJ format. This process begins with a thorough noise cleaning, eliminating the so-called "outliers" or erroneous points that appear due to reflections or vibrations of the sensor during scanning.
Once the cloud is cleaned, reconstruction algorithms are used. The use of convex hull or Poisson reconstruction These are common methods, although if the parameters aren't properly adjusted, the result can be disappointing. The key lies in densification and smoothing, interpolating the missing areas to create a continuous surface.
When working with industrial parts, it is vital to closing gapsSince it is impossible to scan 100% of a part (there are always hidden areas), the technician must reconstruct those cavities based on the surrounding geometry so that the mesh is airtight and usable in graphics engines.
File formats and their usefulness in the pipeline
Not all formats serve the same purpose. OBJ format It is the workhorse for static geometry due to its universal compatibility, although it depends on an external .mtl file to manage materials, which sometimes causes the model to look gray when imported if we forget to attach said file.
- GLB/glTF: Considered the "JPEG of 3D", they are ideal for web and augmented reality as they package everything into a single binary file with PBR materials.
- fbx: The preferred option for complex animations and rigging in Unreal Engine or Unity.
- STL: The absolute standard for 3D printing, but limited since it does not save colors or textures.
- STEP/IGES: Mathematical precision formats (NURBS) used in engineering, completely different from triangular meshes.
If you're looking for maximum efficiency in a graphics engine, the ideal solution is to run the OBJ file through a process of topology optimization (remeshing) to reduce the polygon count. A model with millions of triangles will crash any graphics card; the goal is to maintain detail in the curves and simplify the flat areas.
Integration into Graphics Engines and Visualization
Once we have the OBJ file, the next challenge is the visual fidelityMany users get frustrated because the model looks bad when imported. This is usually due to a misalignment of the axes (Y-up versus Z-up) or because the face normals are inverted, making the object appear hollow or with black holes.
To achieve a professional finish, it is recommended to use PBR materials (Physically Based Rendering). While OBJ is limited in this respect, converting the final mesh to GLB allows the graphics engine to correctly interpret the surface roughness and metallicity, giving a much more realistic appearance.
In large-scale projects, such as city scanning, they are used batch processing pipelinesAn entire city cannot be loaded into a single OBJ file, so the scene is divided into sectors or "tiles", optimizing each one to ensure smooth loading in the graphics engine and avoid overloading the RAM.
To transform raw LiDAR data into a compelling visual experience, it's essential to master the workflow from cloud filtering, through polygon mesh generation, and finally to exporting to optimized formats like GLB or FBX, always ensuring that the geometric intention remains intact during each conversion.
Passionate writer about the world of bytes and technology in general. I love sharing my knowledge through writing, and that's what I'll do on this blog, show you all the most interesting things about gadgets, software, hardware, tech trends, and more. My goal is to help you navigate the digital world in a simple and entertaining way.