PLANAR PROJECTION OF MOBILE LASER SCANNING DATA IN TUNNELS 
J. A. Goncalves t Mendes, R. t Araüjo r E., Oliveira, A. 2 Boavida, J. > 
! Universidade do Porto — Faculdade de Ciéncias, Rua Campo Alegre, 687, 4169-007, Porto, Portugal - 
jagoncal@fc.up.pt, rlpmendes@gmail.com, emilia.s.araujo@gmail.com 
? Artescan — 3D Scanning, IPN Incubadora de Empresas, 3030-199 Coimbra, Portugal — 
(adrianooliveira,jboavida)@artescan.net 
KEY WORDS: Laser scanning, point cloud, automation, adjustment, rectification 
ABSTRACT: 
Laser scanning is now a common technology in the surveying and monitoring of large engineering infrastructures, such as tunnels, 
both in motorways and railways. Extended possibilities exist now with the mobile terrestrial laser scanning systems, which produce 
very large data sets that need efficient processing techniques in order to facilitate their exploitation and usability. 
This paper deals with the implementation of a methodology for processing and presenting 3D point clouds acquired by laser 
scanning in tunnels, making use of the approximately cylindrical shape of tunnels. There is a need for a 2D presentation of the 3D 
point clouds, in order to facilitate the inspection of important features as well as to easily obtain their spatial location. 
An algorithm was developed to treat automatically point clouds obtained in tunnels in order to produce rectified images that can be 
analysed. 
Tests were carried with data acquired with static and mobile Riegl laser scanning systems, by Artescan company, in highway tunnels 
in Portugal and Spain, with very satisfactory results. The final planar image is an alternative way of data presentation where image 
analysis tools can be used to analyze the laser intensity in order to detect problems in the tunnel structure. 
1. INTRODUCTION 
Laser scanning systems are an important technology in the 
monitoring of engineering infrastructures. The very large data 
sets of point clouds provide a detailed geometric description of 
the observed objects, which is useful to monitor the conditions 
of the structure (Yoon, et al, 2009). For several years laser 
scanning was static, with equipment mounted on a tripod. In 
recent years mobile systems, integrating very accurate 
positioning and navigation (INS-GNSS) became operational and 
a commercial solution in last decade (Hunter, 2009). Fast laser 
surveys along roads and railways became possible, allowing for 
accurate monitoring of infrastructure conditions. Automation in 
data analysis has also developed, allowing for faster results 
(Biosca and Lerma, 2008) 
Most of the systems now used incorporate digital cameras, 
synchronized with the laser, in a way that the point cloud can be 
coloured in order to create a realistic model of the objects. The 
image component provides important information, for example 
to detect possible damages in structures. 
The intensity of the laser pulse returns can also be treated in the 
same way as image data. It may be of particular interest in 
structures not illuminated, such as tunnels, where the detection 
of cracks, moisture, water intrusions may be detectable in the 
laser intensity data (Nuttens et al., 2010). 
In the case of laser intensity or any imagery from other sensors, 
optical or thermal, the data volumes are very large and visual 
inspection may not be sufficient to detect problems on the 
surface of tunnels being monitored. Standard image processing 
techniques can be used for an automatic detection but it requires 
that surfaces are transformed onto planes in order that the data 
can be represented as a planar image. This is a problem of 
transforming a surface into a plane, similar to a map projection. 
Our problem is then the identification of the main surface in the 
point cloud, decide which points are not on the surface, and then 
108 
define it mathematically. In a rough tunnel it is likely that points 
may be slightly separated from the surface, being described by 
the distance to this reference surface, in the same way as a 
height. Once the surface is defined it is then transformed in 
some manner into a plan. Points are described by their (x,y) 
coordinates on the plan, and possibly a distance to the reference 
surface, and can be mapped uniquely onto the original 3D 
position. 
A simple tunnel with the shape of a cylinder is the simplest 
problem. It can be unfolded, transforming it in a plane, without 
any deformation. The shape of the cylinder cross-section is not 
a problem, provided that it is always the same. The resulting 
plane has two coordinates that are the distance along the 
cylinder axis and the distance along the cross-section. 
Real tunnels differ from this situation in that they are curved 
and can be modelled as juxtaposed sections of a torus, in a 
snake shape. Figure 1 shows a torus and two sections of a torus 
juxtaposed. 
  
Figure 1. Representation of a torus and the juxtaposition of two 
torus sections. 
This paper describes an algorithm that takes a point cloud and 
identifies the tunnel alignment, fits curves to the cross-sections 
profiles and divides the tunnel in small sections, mapping it then 
into a plane. The curvature of tunnels, especially in railways, is 
small, normally with curvature radius of several kilometres. The 
deformation that will occur is very small, not degrading the 
perc 
surfz 
This 
Scan 
CS6 
been 
railv 
mair 
seve 
aval 
recti 
tool 
Figi 
The 
sma 
with 
rail 
sho 
mor 
Height (m) 
The 
unf 
lase 
The 
mal 
anal 
The 
ST/ 
The 
can 
veh