Peer-Reviewed Articles
33 A Synergistic Automatic Clustering Technique
(SYNERACT) for Multispectral Image Analysis
Kai-Yi Huang
Abstract Download
Full Article
The Iterative Self-Organizing Data Analysis Technique (ISODATA) has been widely
used in unsupervised and supervised classification. However, ISODATA suffers
from several limitations. The user often spends much analyst time on specifying
input parameters by trial and error, particularly initial cluster centers.
Of more importance, an inappropriate choice of initial clusters may cause
poor classification results. ISODATA is computationally intensive because
of its iterative process. This study aimed to develop a synergistic automatic
clustering technique (SYNERACT) that combined the hierarchical descending
and ISODATA clustering procedures to avoid those limitations. The two methods
were compared using multispectral digitized video images. An inappropriate
choice of initial seeds for ISODATA was shown to reduce accuracies significantly.
In contrast, SYNERACT was capable of determining the suitable locations for
the initial clusters automatically from the data, thereby avoiding those
limitations. Owing to this capability, SYNERACT was not so heavily dependent
on the iterative process as was ISODATA , and thus was much faster than ISODATA.
SYNERACT also matched ISODATA in accuracy. Accordingly, SYNERACT could serve
as an alternative to ISODATA for multispectral image analysis.
41 Fuzzy Objects: Their Changes and Uncertainties
Tao Cheng
Abstract Download
Full Article
Although most geographical entities represented as objects in a GIS have crisp
boundaries, in reality they have indeterminate boundaries and fuzzy spatial
extent. This is due to the fact that they are distributed continuously in
space and time. Furthermore, measurement procedures generally produce data
with a limited accuracy, which lead to the uncertain description of geographical
entities. Because of these facts, uncertainties exist in the crisp object
description of geographical entities. When temporal information is applied
to analyze their change, these uncertainties will influence the final results
of mapping. Therefore, the effects of the uncertainties on monitoring geographical
entities should be studied, in order to provide accurate information to decision
makers. This paper discusses the indeterminate nature of geographic entities
and its effect on change detection when they are monitored through time.
The concepts of fuzzy objects and fuzzy change detection are applied to describe
heterogeneous objects or objects with indeterminate boundaries. Several measures
for fuzzy change detection are presented to evaluate the change processes
of the objects. A practical example of the dynamics of sediments along the
Dutch coast is elaborated to demonstrate the approach and the concepts proposed.
51 Scale and Texture in Digital Image Classification
Christopher J.S. Ferro and Timothy Warner
Abstract Download
Full Article
Classification errors using texture are most likely associated with class edges,
but investigators often avoid edges when evaluating texture for classification.
The large windows needed to produce a stable texture measure produce large
edge effects. Small windows minimize edge effects, but often do not provide
stable texture measures. Simulated data experiments showed that class separability
increased when texture was used in addition to spectral information. Texture
separability improved with larger windows. This improvement was over estimated
when pixels were chosen away from class edges. Airborne Data Acquisition
and Registration (ADAR) data showed that separability of class interiors
improved with the addition of texture, but that, for the whole class, separability
fell. Maximum-likelihood classification of the ADAR data demonstrated the
effect of edges and multiple scales in reducing the accuracy of classification
incorporating texture.
65 Impacts of Patch Size and Land-Cover Heterogeneity on Thematic Image Classification
Accuracy
Jonathan H. Smith, James D. Wickham, Steven V. Stehman, and Limin Yang
Abstract Download
Full Article
Landscape characteristics such as small patch size and land-cover heterogeneity
have been hypothesized to increase the likelihood of mis-classifying pixels
during thematic image classification. However, there has been a lack of empirical
evidence to support these hypotheses. This study utilizes data gathered as
part of the accuracy assessment of the 1992 National Land Cover Data (NLCD)
set to identify and quantify the impacts of land-cover heterogeneity and
patch size on classification accuracy. Logistic regression is employed to
assess the impacts of these variables, as well as the impact of land-cover
class information. The results reveal that accuracy decreases as land-cover
heterogeneity increases and as patch size decreases. These landscape variables
remain significant factors in explaining classification accuracy even when
adjusted for their confounding association with land-cover class information.
71 Drought Monitoring with NDVI-Based Standardized Vegetation Index
Albert J. Peters, Elizabeth A. Walter-Shea, Lei Ji, Andrés Viña,
Michael Hayes, and Mark D. Svoboda
Abstract Download
Full Article
Drought is one of the major natural hazards affecting the environment and economy
of countries worldwide. Reliance on weather data alone is not sufficient
to monitor areas of drought, particularly when these data can be untimely,
sparse, and incomplete. Augmenting weather data with satellite images to
identify the location and severity of droughts is a must for complete, up-to-date,
and comprehensive coverage of current drought conditions. The objective of
this research was to standardize, by time of year, the Normalized Difference
Vegetation Index (NDVI) to augment drought-monitoring techniques. The Standardized
Vegetation Index (SVI) describes the probability of vegetation condition
deviation from "normal," based on calculations from weekly NDVI values. The
study was conducted with 12 years (1989-2000) of Advanced Very High-Resolution
Radiometer (AVHRR) satellite images. Z-scores of the NDVI distribution are
used to estimate the probability of occurrence of the present vegetation
condition at a given location relative to the possible range of vegetative
vigor, historically. The SVI can be interpreted as vegetation condition based
on the fact that vegetation is an efficient integrator of climatic and anthropogenic
impacts in the boundary layer of the atmosphere. It thereby provides a spatially
and temporally continuous short-term indicator of climatic conditions. Findings
indicate that the SVI, along with other drought monitoring tools, is useful
for assessing the extent and severity of drought at a spatial resolution
of 1 km. The SVI is capable of providing a near-real-time indicator of vegetation
condition within drought regions, and more specifically areas of varying
drought conditions.
77 Subpixel Classification of Alder Trees Using Multitemporal Landsat Thematic
Mapper Imagery
Kazuo Oki, Hiroyuki Oguma, and Mikio Sugita
Abstract Download
Full Article
An unmixing method was used to estimate the area of alder trees from multitemporal
Thematic Mapper (TM) imagery of the Kushiro mire, an important wetland area
in Japan. Because the spectral radiance (endmember) of alder coverage within
a pixel must be precisely known before the unmixing method is applied, we
first collected ground truth data for the canopy area of alder using field
survey methods, and aerial photographs. The alder endmember was calculated
by linking TM imagery and the ground truth data. Alder coverage was estimated
more precisely than it was with classification by the maximum-likelihood
method. Furthermore, the alder area could be evaluated within a pixel, which
cannot be classified as alder by the maximum-likelihood method.
83 Use of Contour-Based DEMs for Deriving and Mapping Topographic Attributes
Hiroko Mizukoshi and Masamu Aniya
Abstract Download
Full Article
Algorithms using contour-based DEMs to calculate slope gradient and aspect,
and to classify and map slope profile and plan forms, were developed, which
we call C-BATM (Contour-Based Automatic Terrain Mapping). These are important
topographic attributes for various analyses and terrain-related hazard mapping.
As a first step, a fall line, the line of the maximum gradients or a flow
path, was generated from all data points of all contours. Then, using the
fall line segment (between adjacent contours), slope gradient and aspect
were calculated. The change in slope gradients of three consecutive contours
are the basis for classifying slope profile form into concave, planar, and
convex. Slope plan form was classified from the contour crenulation, using
a point of inflection by examining the directional change of contour segments
(line between two consecutive data points). In the process of classifying
slope morphologies, the profile and plan curvatures were also calculated
and recorded. Test results of these methods in mountainous areas indicate
the advantage of using contour-based DEMs over the use of the grid-based
DEM or TIN. These methods have a potential of much wider applications to
terrain analyses and hydrological modeling.
Top | Home |