Las Campus

Airborne Lidar scanners can be used to obtain large scale point clouds of the earths surface that are a great foundation for creating 3D models of the real world. For my Master Thesis I used free available data from Bezirksregierung Köln to create a textured 3D Model of my university campus using a point-based rendering algorithm with this software.

Preprocessing

The Lidar point clouds contain objects that are either not relevant for most 3D Models and are hard to process and they typically have less data density on vertical surfaces, so they need to be preprocessed to get a better result.

Remove vegetation and Outliers from PointCloud

If available use shp data to add points for walls

Modeling

I decided to try a surface splatting approach for this project.

Use splatting to create surface

Add texture

Credit

Free las, shp and texture data from Bezirksregierung Köln NRW

Additional shp data from OpenStreetMap

Some of the Splatting Algorithms PhongSplatting and AdaSplats

Project Overview

small introduction to most important files

main

• specify name of las file, gml file, (osm) shp file and texture jpg here

Window

manage rendering, "camera", window and shaders
uses data stored in DataStructure

• move with W, A, S, D, shift, space
• look around with mouse
• F1 toggles rendering of normals and coordinate system
• F2 toggles rendering of splats vs points
• F3 toggles backface culling
• F4 toggles rendering of texture

DataIO

parse files and process LiDAR data (chapter 3)
gets called in DataStructure

reads and converts all the files

• las -> Point Cloud
• shp -> Polygons
• gml -> Buildings

function preprocessWalls

• convert shp Polygons to Buildings

function filterAndColorPoints

• filters vegetation from point cloud by removing multi return points that don't belong to no osm/shp building
• marks points as wall points for next step
• filters outliers

function insertWalls

• mixes gml and (osm) shp walls
• fills holes in point cloud by inserting additional points on walls from buildings
• removes old wall points afterwards

function readCache and writeCache

two types of cache:

• one after reading data and processing it in DataIO (point_cache)
• one after splatting (splat_cache)

DataStructure

holds data
uses DataIO to get cached/fresh data
performs AdaSplats if there is no splat_cache

holds data

• point cloud
• texture coordinates
• tangents for splats (points without splat have tangent1 = (0,0,0))

function adaSplats