Lerma, Jose Luis
APPLICATION OF SPECTRAL AND TEXTURAL CLASSIFICATIONS TO RECOGNIZE
MATERIALS AND DAMAGES ON HISTORIC BUILDING FACADES
José Luis LERMA, Luis Ángel RUIZ, Fernando BUCHÓN
Polytechnic University of Valencia, Spain
Department of Cartographic Engineering, Geodesy and Photogrammetry
illerma@cgf.upv.es, laruiz@cef.upv.es, fbuchon@upvnet.upv.es
KEY WORDS: Pattern Recognition, Spectral and Textural Classification, Close-range Photogrammetry
ABSTRACT
Spectral and textural classifications were evaluated for the identification, detection and recognition of outer materials
and alterations in an architectural monument. A corpus of seven bands, including the spectral blue, green, red and near-
infrared bands and further three red-textural bands was used in different classifications in order to evaluate the
profitability of combination of both methods, spectral and textural. This study showed either that multispectral
classification with visible and near-infrared bands increased the degree of recognition among different materials and
damages and that the addition of textural bands to the multispectral classification allowed to discriminate similar
materials, as rocks, on a historic facade.
1 INTRODUCTION
The study and evolution of historical monuments usually involves rather different kind of technicians and disciplines, in
order to know and understand their reactions, and the actions required just to preserve or restore them for further
generations, as part of the humans legacy.
Actually, the recognition of materials, their spatial distribution, shape and area recovered in our architectonic facades is
one of the most important and tedious tasks in the documentation, conservation and restoration of monumental
buildings. Until recent years, the conservation and restoration works have been done by visual and direct inspection in
the field, drawing and sketching the materials or features identified directly on the facade. Furthermore, these tasks
were somehow time consuming and required previous to the conservative treatments. Once the restoration work had
begun, special care was fixed on the specific deterioration. On the contrary, little attention has been drawn as regards
automation in surveying, recording, documentation and monitoring the external features on Cultural Heritage by
multidisciplinary conservation teams, although last decade it was a new goal for a technical working group within CIPA
(Waldhàusl, 1997).
Studies regarding the application of spectral classifications for the automatic plotting of materials and damages on
masonry walls have already been done by Grunicke et al (1990), Herráez et al (1997) and Lerma et al (1999). All of
them used reflected-light imagery, i.e. visible and near-infrared wavelength energy, meanwhile the first one
recommended the application of unsupervised classification, the formers required supervised classification due to the
higher accuracy reached.
Anyway, problems arise in segmenting images either when the pixels which comprise individual objects are spectrally
heterogeneous (Ryherd et al, 1996) and when identical spectral features have got different image texture. Unfortunately,
both conditions appear quite often in architectural facades, and they should be taken in mind before attempting the
classification.
The benefits of applying independently spectral and textural classifications have been appointed by the authors in
previous studies: Lerma and Ruiz (1999), Lerma and Ruiz (2000). Herein, several classifications have been carried out
to examine the materials and their alterations applying successively visible spectral classification with three bands (B,
480 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B5. Amsterdam 2000.
G, R
(witl
com
(50x
The
lim
and
dan
tor
dar