DETECTION OF BRICKS IN A MASONRY WALL 
George Sithole 
University of Cape Town, School of Architecture, Planning and Geomatics, 
Private Bag, 7701, Cape Town, South Africa - george.sithole@uct.ac.za 
Commission V, WG 3 
KEY WORDS: Laser scanning, Terrestrial, Segmentation, Detection, Algorithms 
ABSTRACT: 
Laser scanning is being increasingly applied in the documentation of buildings and archaeological sites. The documentation process 
sometimes involves making an inventory of bricks in walls. This paper proposes an automatic/semi-automatic segmentation method 
for detecting bricks in walls, in particular masonry walls. In developing the detection method the problems with detecting bricks in a 
3D point cloud obtained by terrestrial laser scanning are first examined. The method starts with a 3D triangulation of a 3D point 
cloud obtained by terrestrial laser scanning. Associated with each point in the cloud are reflectance values and an RGB triplet. The 
method then uses the triangulation, reflectance and RGB triplets to effect a proximity-based segmentation. A technique for selecting 
ideal segmentation parameters is also proposed. The method is tested on the point cloud of three different walls and the results of the 
brick detection are presented. Early results are promising. 
1. INTRODUCTION 
Laser scanning is being increasingly applied in the 
documentation of built structures and archaeological sites. The 
documentation process may sometimes involve making an 
inventory of bricks in walls. This inventory is typically done 
manually and is time consuming. Because of the large size of 
laser scans an automation of this inventory is desirable. A first 
step in automating the inventory is to automate the detection of 
bricks in the point cloud of scanned walls. This paper proposes 
an automatic/semi-automatic segmentation method for detecting 
bricks in walls, in particular masonry walls. 
The paper is divided into four parts. In the first part previous 
work done is discussed. In the second the problems with 
detecting bricks in a 3D point cloud are examined. In the third 
part an outline of the proposed detection algorithm is outlined. 
Finally, in the fourth part early results are presented and 
discussed. 
2. PREVIOUS WORK 
The documentation of walls may be required for such purposes 
as the generation of façade drawings (Henze et. al., 2005), 
detecting structural damage (Arias et. al., 2007, Hemmleb et. al. 
2006, Lerma et. al. 2000), or the quality control of wall finishes. 
Previous work done in the detection of bricks in walls is limited 
and mostly done by means of image processing. Typically, the 
objective is to detect general structural damage rather then the 
bricks themselves. But the classification methods used have the 
potential to detect the bricks as well. 
Hemmleb et. al. 2006 use a multi-spectral (four bands) laser 
scanner system to investigate damage on walls from moisture 
and biological covering. The output from the multi-spectral 
laser scanner is two or more images in the infrared spectrum. 
An image contains the reflectivity at a given wavelength. 
Quantitative damage is assessed based on the differencing of 
the images. 
Further digital cameras are also used. The images from the 
cameras are classified using cluster analysis, maximum- 
likelihood and object oriented techniques. Digital infrared 
cameras can also be used. However, Hemmleb et. al. 2006, 
report unsatisfying classification results from this, because of 
the high correlation between the infrared channels. Some of the 
results obtained by Hemmleb et. al. 2006 suggest that provided 
the surface of mortar has not suffered serious damage it is 
possible to detect bricks in a wall through image segmentation. 
Lerma et. al. 2000, Lerma 2001 and Strackenbrock et. al. 1990 
also use imagery and image processing to identify structural 
damage in walls. 
As more 3D laser scanner data becomes available it becomes 
topical to study the detection of bricks in the 3D scan of a wall. 
3. DETECTING BRICKS IN WALLS 
Detection of bricks in laser scans of walls involves determining 
a discriminating measure between bricks and mortar, and later, 
extracting (segmenting) the individual bricks. Discriminating 
between bricks and mortar assumes that the two possess very 
different radiometric and geometric surface characteristics. 
This section looks at the difficulty associated with detecting 
bricks in a wall in terrestrial laser scanner data. The problems 
discussed below primarily hamper the distinction between 
mortar and brick, and in consequence the segmentation of 
bricks. 
3.1 Similar surface textures 
Segmentation based on surface texture would be possible if the 
mortar in walls had a different texture from those of bricks. 
However, this is not always the case, as in face brick walls. 
This problem can be further complicated if the scanned points 
have a low accuracy or if the average point spacing ofthe point 
cloud is too large.