11-05-2012, 04:03 PM
Land cover classification using reformed fuzzy C-means
153.pdf (Size: 308.73 KB / Downloads: 32)
Introduction
Land cover refers to features of land surface. These can be natural, semi-natural, managed or
totally man-made. They are directly observable. The main reason for producing land cover maps
is to give us a clear idea of the stock and state of our natural and built resources. A land cover
classification, is an essential component in developing a responsible attitude to environmental
management. Land cover is distinct from land use despite the two terms often being used
interchangeably. Land use is a description of how people utilize the land and socio-economic
activity–urban and agricultural land uses are two of the most commonly recognised high-level
classes of use. At any one point or place, there may be multiple and alternate land uses, the specification
of which may have a political dimension. Land cover classifications are essential inputs
to environmental and land use planning at local, regional, and national levels. This paper uses
segmentation based on unsupervised clustering techniques for classification of land cover.
Related work
Fuzzy C-means is a method of clustering which allows one data to belong to two or more clusters.
This method was developed by Dunn (1973) and improved by Bezdek (1981) and is frequently
used in image segmentation and pattern recognition. The main disadvantage of FCM is that the
sum of membership values of a data point in all the clusters must be one and so the algorithm has
difficulty in handling outlier points (Cox 2005). Frank Klawonn & Annette Keller (1997) have
proposed a modified C-means algorithm with changed distance function which is the dot product
instead of the conventional Euclidean distance. This method is used for identifying clusters of
new shapes. An additional term is injected into the objective function to constrain the behaviour
of membership functions with the neighbourhood effect by Lei Jiang & Wenhui Yang (2003).
Krishnapuram & Keller (1996) have proposed a fuzzy-possibilistic C-means (FPCM) model and
algorithm that generated both membership and typicality values when clustering unlabelled data.
FPCM constrains the typicality values so that the sum of over all data points of typicality to a
cluster is one. The row sum constraint produces unrealistic typicality values for large data sets.
Later they have modified it by a new model called possibilistic-fuzzy C-means (PFCM) model.
PFCM produces memberships and possibilities simultaneously, along with the usual point prototypes
or cluster centers for each cluster (Pal et al 2005). Both FPCM and PFCM require some
boot strap method for initialization of weights.
Fuzzy clustering algorithms and classification method
In real applications very often no sharp boundary between clusters so that fuzzy clustering is
often better suited for the data. Membership degrees between zero and one are used in fuzzy
clustering instead of crisp assignments of the data to clusters.
The resulting data partition improves data understanding and reveals its internal structure.
Partition clustering algorithms divide up a data set into clusters or classes, where similar data
objects are assigned to the same cluster whereas dissimilar data objects should belong to different
clusters.
Areas of application of fuzzy cluster analysis include data analysis, pattern recognition, and
image segmentation. The detection of special geometrical shapes like circles and ellipses can be
achieved by so-called shell clustering algorithms.
Fuzzy C means
The most prominent algorithm is the FCM or fuzzy C means algorithm. The fuzzy C means
algorithm was proposed as an improvement of the classic hard C-means clustering algorithm.
The FCM algorithm receives the data or the sample space, an n×m matrix where n is the number
of data and m is the number of parameters. The number of clusters c, the assumption partition
matrix U, the convergence value E all must be given to the algorithm. The assumption partition
matrix has c number of rows and n number of columns and contains values from 0 to 1.