Abstract— Unsupervised classification creates clusters by
grouping pixels based on the reflectance properties of pixels. This
paper presents a new approach to classify multi-spectral
remotely sensed image using pixels’ density in N-dimensional
scatterplot. It first finds the densely populated clusters in N-
dimensional scatter plot and then finds gravity centers of these
densely populated clusters. Later, multispectral image is
classified using minimum distance to gravity classifier. At the
beginning of classification, this approach neither makes extreme
assumption of considering each pixel as a different cluster nor
goes to the other extreme by considering all the pixels in a single
cluster. It follows the middle path by making assumption of some
pixels as gravity centers of different clusters before classifying
the image. It creates clusters of equal size in N-dimensional
scatter plot and picks up the densely populated clusters. All these
clusters are recursively iterated for self-adjustment of gravity
centers using mathematical algorithm. The approach uses
gravitational force for merging two nearby clusters. Here, gravity
center of a cluster is calculated by summing up the spectral
bands’ values and dividing it by number of pixels within that
cluster. When two nearby clusters are merged, their gravity
centers are also adjusted accordingly. This provides the most
densely populated clusters and their gravity centers. Now, these
gravity centers can be used to classify the remotely sensed image
using minimum distance to gravity center classifier.
Index Terms— Clustering, Multi-Spectral, Remote Sensing,
Scatter Plot, Unsupervised Classification,
I. INTRODUCTION
he classification is the grouping of pixels with common
characteristics. It divides the feature space into several
classes. The remotely sensed images can be classified using
unsupervised classifier, supervised classifier or object-based
classifier. Supervised classification needs training of data,
whereas unsupervised classifier needs no training of data.
Unsupervised classifier uses mathematical algorithm for
classifying the data. The object-based classification is based
on set of similar pixels called objects.
The reflectance signals are captured as analog signals and
converted to digital numbers. These digital number (DN)
values are also called grey-level values. The number of grey
This paper was submitted on 3
rd
August 2015. This work was not supported
in part or in full from anywhere.
The author is working as Senior Consultant, National Institute for Smart
Government. Author is with Department of Information Technology, Block-
24, Kasumpti, Shimla-171009 HP India (adarsh_khare2@yahoo.com).
levels that can be recorded for a given pixel is called
radiometric resolution. A radiometric resolution of 8 bit means
pixel value ranges from 0-255 (0 – (2
8
-1)). Here, sample
image used in this research paper has radiometric resolution of
8 bit. So, scatter plot uses DN values ranging from 0 to 255.
The classification is generally carried out on the basis of
features like pixels density, texture, etc. in the image. In this
approach, the classification is carried out on the basis of
pixels’ density in N-dimensional scatter plot. This approach
first finds out the gravity centers of pixels on the basis of
pixels density in scatter plot. Later it classifies the image using
minimum Euclidian distance to gravity centers classifier.
Following is the process flow diagram for finding gravity
centers in N-dimensional scatter plot and classifying the
multispectral image using minimum distance to gravity center
classifier.
Fig. 1. Process flow diagram of unsupervised classification using N-
dimensional scatter plot.
In clustering technique like Adaptive Resonance Theory-2,
pixels are grouped based upon their minimum distance
decision rule and also mean of the cluster is adjusted
accordingly. But in the following approach, clusters in N-
Unsupervised Classification
Using Gravity Centers from Scatter Plot
Adarsh Kumar Khare
T
2016 Second International Conference on Computational Intelligence & Communication Technology
978-1-5090-0210-8/16 $31.00 © 2016 IEEE
DOI 10.1109/CICT.2016.36
140
2016 Second International Conference on Computational Intelligence & Communication Technology
978-1-5090-0210-8/16 $31.00 © 2016 IEEE
DOI 10.1109/CICT.2016.36
140