Land-Use/Land-Cover Classification Using Elephant Herding Algorithm
- PDF / 6,204,790 Bytes
- 10 Pages / 595.276 x 790.866 pts Page_size
- 101 Downloads / 247 Views
RESEARCH ARTICLE
Land-Use/Land-Cover Classification Using Elephant Herding Algorithm J. Jayanth1
•
V. S. Shalini2 • T. Ashok Kumar3 • Shivaprakash Koliwad4
Received: 8 November 2018 / Accepted: 26 December 2018 Indian Society of Remote Sensing 2019
Abstract In recent years, swarm intelligence algorithms such as particle swarm optimisation, ant colony optimisation, cuckoo search and artificial bee colony algorithm have shown promising results in multispectral image classification. Elephant herding algorithm is one of the newly emerging nature inspired algorithms which can analyse multispectral pixels and determine the information of class via fitness function. When the spectral resolution of the satellite imagery is increased, the higher within-class variability reduces the statistical separability between the LU/LC classes in spectral space and tends to continue diminishing classification accuracy of the traditional classifiers. These are mostly per pixel and parametric in nature. Experimental result has revealed that elephant herding algorithm shows an improvement of 10.7% on Arsikere taluk and 6.63% on NITK campus over support vector machine. Keywords Support vector machine (SVM) Elephant herding (EH) Multispectral (MS) image classification
Introduction Recent advancements in technology and data analysis have made remote sensing a powerful tool for regional mapping of natural resources to extract useful thematic information. Information extraction is based on digital image classification which is viewed as the process of categorising all the pixels in an image automatically into a finite number of landcover classes (Bahmanyar et al. 2015). Main objective of the classification is to exploit spectral and spatial resolution from the data so as to improve classification accuracy. As far as the classifiers are concerned, it is evident from the literature survey that classifiers are broadly grouped & J. Jayanth [email protected] 1
Department of Electronics and Communication Engineering, GSSS Institute of Engineering and Technology for Women, Mysore, Karnataka 570 016, India
2
Department of Electronics and Communication Engineering, ATME College of Engineering, Mysore, Karnataka 570 028, India
3
SDM Institute of Technology, Ujire, Belthangady 574240, India
4
Department of Electronics and Communication Engineering, Malnad College of Engineering, Hassan, Karnataka 573 202, India
into supervised and unsupervised types depending upon the need of training phase. The classifiers are also identified as parametric and nonparametric types employing pixel-based or object-oriented classification approach. The limitations of the parametric classifiers become serious when the availability of the training data is insufficient and unable to satisfy the rule of thumb defined for training data set size. The performance of the statistical classifier worsens further when the ancillary data are integrated into spectral bands where the latter result in a non-Gaussian distribution of the resultant data (Chen and Tian
Data Loading...
