The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008
571
Figure 8 Point hue (left), saturation (middle) and values (right)
of wall 1. The masonry wall is made of tightly packed granite
stones, hence the hue and saturation components have almost
uniform shading through out, except for two of the stones. As
expected the value component (representing the distribution of
light on the wall) contains more information.
Point and edge attributes were computed as in wall 1. The
segmentation of the wall using the edge length and value is
shown in Figure 9.
Using edge length Using value
Figure 9 Segmentation using edge length and value as edge
strength criteria
The two results are fairly similar. The most notable aspect of
the result is that there is a lot of under segmentation within and
on the edge of the bricks. It is suspected that these are caused
by depressions/ protrusions on the surface of the bricks that
cause discontinuities or cast shadows. This however needs to be
further tested. Wall 2 shows the difficulty of segmenting tightly
packed masonry walls. Nonetheless, while the segmentation is
not good enough for a brick reconstruction, a visual comparison
with the value image in Figure 8 and the segmentation result in
figure 9 suggests that it is good enough for the purposes of
detection.
5.3 Discussion
From tests carried out, the proposed general algorithm works if
the mortar channel is deep enough to cause a discontinuity
between the surface of the brick and the mortar and the scan is
dense enough to capture the mortar between bricks (at least
three points). Variable resolutions in the scan present the
greatest challenge as they make the selection of a segmentation
threshold difficult. Because of this it is best to detect bricks in
single scans, to avoid the dense overlaps that are caused by the
overlapping of two or more scans.
Experience from several tests indicates that the problems
attendant with every wall is different. Lighting conditions will
vary, the material of both mortar and brick vary, the texture of
the mortar and brick are not the same between walls.
Furthermore the resolution of scans will also vary. Thus, a
semi-automatic detection approach may be better than an
automatic one.
6. CONCLUSIONS
Laser scanning for the documentation of walls is done primarily
for the purposes of generating façade drawings, assessing
structural damage and quality control during and after
construction. However, detecting individual bricks should allow
for greater enhancement and intelligence in these activities. For
example, if the bricks in a wall are of a uniform size then
detected misshapen bricks will hint at defects (such as cracks)
in a wall. Furthermore, the deformation analysis of a wall can
be done based on a brick by brick comparison as opposed to
current techniques that use targets or surface matching.
A semi-automatic algorithm and tool was developed and
experiences during its development showed that the bricks in a
wall can be detect fairly quickly and with a high level of
accuracy. The development of the proposed algorithm is still in
its early stages. Greater testing of suitable point and edge
attributes need to be done. For example a shortcoming of the
algorithm is that it does not use the normal at a point. Using the
normal at a point in the computation of edge attributes will
enhance the detection of discontinuities between bricks and
mortar. As is, the algorithm is still naïve. More testing is
required to improve the algorithm. But the results obtained thus
far are encouraging.
Segmenting embedded walls is complicated because the walls
at different depths will contain points at different resolutions.
To overcome the problem of segmenting embedded walls an
optimum threshold selection technique has been proposed. The
technique has the potential for wider application in segmenting
other types of point clouds.
ACKNOWLEDGEMENTS
The author wishes to thank Prof. Heinz Ruther for providing the
laser scanner data used in the tests.
REFERENCES
Hemmleb, M., Weritz, F., Schiemenz, A., Grote, A., Maierhofer,
C., 2006. Multi-spectral data acquisition and processing
techniques for damage detection on building surfaces.
Proceedings of ISPRS Commission V Symposium, Dresden,
WG 2: Cultural heritage documentation. 6 pages.
Henze, F., Wulf-Rheidt, U., Schneider, D., Bienert, A., 2005.
Photogrammetric and geodetic documentation methods at St.
Petri Cathedral, Bautzen. Proceedings of CIPA 2005 XX
International Symposium, 26 September - 01 October, Torino,
Italy. 6 pages.
Lerma, J. L., Ruiz, L. A., Buchon, F., 2000. Application of
spectral and textural classifications to recognize materials and
damages on historic building facades. IAPRS, vol. XXXIII, part
B5, Amsterdam, pp. 480-483.
Lerma, J. L., 2001. Multiband Versus Multispectral Supervised
Classification of Architectural Images. Photogrammetric
Record, volume 17, issue 97, pp. 89-101.
Strackenbrock, B., Sacher, G., Grunicke, J.-M., 1990. Image
Processing for Mapping Damages to Buildings. In: Proceedings
of the XIII. International CIPA symposium, Cracow, 7 pages.