dc.description.abstract |
This study aimed to explore on image processing techniques to determine the
shoreline at the sub-pixel level by using simulated coarse satellite sensor
imagery and to examine the potential of soft classification on shoreline mapping
of Kuala Terengganu coastal area. In general, to map shoreline require two
types of imagery; fine and coarse spatial resolution which taken on the same
day, date, time and area. However due to the lack of data obtained, this study
used simulated 10 m and 20 m coarse spatial resolution imagery from 2.5 m
fine spatial resolution imagery. The performances of three soft classifications on
simulated coarse satellite imagery were evaluated mainly Fuzzy Set Theory
classification, Bayesian classification and Demspter-shafer classification. Unlike
conventional classification, in soft classification the composition information in a
pixel can be extracted which allows the shoreline to be mapped within image
pixels producing an accurate and realistic prediction of the shoreline. This
method was applied to different shape of shoreline extracts; Area I (linear or cross to lying along pixel orientation), Area II (shoreline orientation changes
abruptly with respect to pixel grid) and Area III (linear shoreline) for 10 m and 20
m imagery. From the result, it showed that the accuracy of shoreline prediction
varied according to shoreline orientation and shape. Shoreline generated from
contouring soft classification for linear shoreline area (Area III) produced the
most accurate predicting shoreline with RMSE value less than 1.4 m from Fuzzy
Sigmoidal classification. When the shoreline was aligned exact parallel to the
column of the pixel grid, the accuracy was increased. However, in certain area,
Bayesian showed the potential on producing shoreline with RMSE lower than 2
m. Overall, the effects of spatial resolution were very similar. The coarser the
spatial resolution, the accuracy decreased. From the results, Fuzzy Sigmoidal
though showed less effect on spatial resolution. |
en_US |