Combining the FCM Classifier with Various Kernels to Handle Non-linearity of Class Boundaries
- PDF / 1,857,027 Bytes
- 8 Pages / 595.276 x 790.866 pts Page_size
- 108 Downloads / 168 Views
RESEARCH ARTICLE
Combining the FCM Classifier with Various Kernels to Handle Nonlinearity of Class Boundaries Akshara Preethy Byju1
•
Anil Kumar2 • Alfred Stein3 • A. Senthil Kumar4
Received: 10 December 2017 / Accepted: 16 July 2018 Ó Indian Society of Remote Sensing 2018
Abstract This article presents the use of kernel functions in fuzzy classifiers for an efficient land use/land cover mapping. It focuses on handling mixed pixels obtained from a remote sensing image by considering non-linearity between class boundaries. It uses kernel functions combined with the conventional fuzzy c-means (FCM) classifier. Kernel-based fuzzy c-mean classifiers were applied to classify AWiFS and LISS-III images from Resourcesat-1 and Resourcesat-2 satellites. Optimal kernels were obtained from eight single kernel functions. Fractional images generated from high resolution LISS-IV image were used as reference data. Classification accuracy of the FCM classifier increased with 12.93%. Improvement in overall accuracy shows that non-linearity in the dataset was handled adequately. The inverse multiquadratic kernel and the Gaussian kernel with the Euclidean norm were identified as optimal kernels. The study showed that overall classification accuracy of the FCM classifier improved if kernel functions were included. Keywords Classification Fuzzy set Kernels Kernel based fuzzy clustering Feature space Fuzzy error matrix
Introduction Continuous improvement in satellite technology and the increasing number of satellites have led to a big amount of remote sensing (RS) data. Efficient processing such a huge amount of data and the extraction and analysis of the information from is a challenge. For instance, classifying images based upon user-specific needs is currently a major research focus. Thus, efficient classification methods have to be developed. Such classifiers should be able to deal
& Akshara Preethy Byju [email protected] 1
Department of Geo-Informatics, Indian Institute of Remote Sensing, 4 Kalidas Road, Dehra Dun, Uttarakhand 248001, India
2
Photogrammetry and Remote Sensing Department, Indian Institute of Remote Sensing, 4 Kalidas Road, Dehra Dun, Uttarakhand 248001, India
3
Department of Earth Observation Science, Faculty of GeoInformation Science and Earth Observation (ITC), University of Twente, Enschede, The Netherlands
4
Indian Institute of Remote Sensing (IIRS), Dehra Dun, Uttarakhand 248001, India
mixed pixels as in remote sensing studies, the presence of mixed pixels interferes with an accurate classification of satellite images (Choodarathnakara et al. 2012). Causes for mixed pixels can be manifold ranging from gradual change of one land cover class to another and use of a sensor with coarser resolution (Lillesand and Kiefer 2014). To solve this issue, Bezdek et al. (1984) where a particular pixel is assigned to more than one cluster using membership grades between 0 and 1. Also, in pattern analysis, class boundaries between samples of different classes may appear non-linear in nature due t
Data Loading...
