In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
i 
► 
575 
Satellite remote sensing has revolutionized modem 
oceanography and coastal applications, providing frequent 
synoptic-scale information that can be used to deduce 
ocean/coastal processes. However, it is often difficult to extract 
interpretable patterns from satellite images, as data sets are large 
and often non-linear. 
A technique that has been applied with success to extract 
interpretable patterns from different types of data sets is the 
artificial neural network (ANN). The ANNs have a number of 
advantages over traditional statistical methods (Wassermann, 
1989). Firstly, they can solve non-linear problems of almost 
infinite complexity (Dayhoff, 1990). Secondly, they are more 
robust in handling noisy and missing data than traditional 
methods. This is especially desirable for satellite data from 
visible and infrared sensors that often have a considerable 
portion of the image not visible because of clouds. Finally, they 
do not require prior knowledge and assumptions about the data, 
such as normality or equality of variances (Chen and Ware, 
1999). 
There are many examples of the use of ANNs in remote sensing, 
including many studies that illustrate the ability of ANNs to 
generalize (Lippmann, 1987; Atkinson and Tatnall, 1997; 
Wilkinson, 1997). The use of ANNs to classify remotely sensed 
data has often resulted in a higher or equal mapping accuracy 
than that achieved with traditional classification methodologies 
or mixture modeling (Benediktsson et al., 1990; Atkinson et al., 
1997; Teodoro et al., 2007). There is considerable interest in 
the development of a classifier that can be applied to images of 
coastal areas not only for quantifying water quality parameters 
(Teodoro et al, 2007) or monitoring protected coastal 
environments (Palandro et al., 2008), but in many other 
applications as beach features/pattems classification. 
This study focuses on the application of one ANN to an 
IKONOS image in order to classify the beach features/pattems, 
in a stretch of the northwest coast of Portugal. 
2. METHODOLOGY 
The methodology adopted in this work consists in the 
application of an ANN to an IKONOS image in order to classify 
the beach features/pattems. 
Pais-Barbosa et al., (2007 and 2009) and Teodoro et al., (2009) 
developed methodologies to identify, to measure and to classify 
hydroforms and hydromorphologies, based on the visual 
analysis of vertical aerial photographs datasets in a GIS 
environment and in image classification techniques, 
respectively. The hydroforms were identified based on several 
criteria (location, spectral differences between morphological 
elements, shape and tide) presented in APPENDIX A. Based on 
the knowledge of the coastal features, five classes were defined: 
Sea (S), Suspended-Sediments (SS), Breaking-Zone (BZ), 
Beachface (BF) and Beach (B). 
2.1 Study area 
A stretch of the northwest coast of Portugal was chosen as the 
study area (Figure 2), limited to the north by the Douro River 
mouth (Vila Nova de Gaia city) and to the south by a small 
fisherman village (Aguda), with an extension of approximately 
9.5 km. This coastal stretch represents a dynamic environment, 
which is constantly changing in response to natural processes 
and human activities. 
Over the last few years, some coastal erosion in this particular 
area had been reported (e.g. Southern of Aguda breakwater). 
The main causes of this serious environmental problem have 
been identified as a coastal response to the construction of 
Aguda breakwater, weakening of the river basin sediment 
sources and river-sediment transport (Teodoro et al., 2007). 
The study area is a rocky coast, with highly dynamic beaches 
presenting coastal patterns/forms that change continuously. It is 
composed by very dynamic beach systems, adjusting to wave 
climate and tide range. The tide regime is semidiurnal (period 
or cycle of approximately one-half of a tidal day), reaching up 
to 4.0 m for spring tides (mesotidal coast). The littoral drift has 
a dominant north-south direction. 
Figure 2. Study area (Google Earth®) 
2.2 IKONOS data 
The IKONOS image (2005/09/18) was acquired under the 
scope of an ESA funded research project. The IKONOS image 
was already geometrically corrected. The image bands were 
calibrated for radiance values (Lx), through equation (1), 
\0 4 *DN, 
L À = 
CalCoef x * Bandwidth l ^ j ^ 
where DNx is the digital value for spectral band X, Lx is the 
radiance for spectral band X at the sensor’s aperture 
(W/m 2 /|j.m/sr), CalCoefx is the radiometric calibration 
coefficient [DN/(mW/cm2-sr)] and Bandwidthx is the band 
width of spectral band X (nm). 
The surface reflectance (Rx) was obtained through equation (2), 
R = X*L**d 2 ( 2 ) 
 ESUN Z * cos 6 e