Article Preview
Top1. Introduction
Researchers have been continuously searching for new techniques that can extract maximum information from the remotely sensed image (Long, W., et al., 2004; Demetrics et al., 2006)). Various swarm intelligence techniques such as the ACO (Dorigo, 2004; S. Parpinelli et al., 2002; Xiaoping Liu et al., 2008), PSO (Bratton, J. Kennedy et al., 2007; Omkar, S.N., 2007; Wang, Dong et al., 2008) and hybrid ACO2/PSO optimization (Omkar, S.N., 2007; Bansal, Shelly, et al., 2009) have been used for solving the problem of satellite image classification using the concepts of information sharing as demonstrated in our paper (Goel, L., et al., 2011b). In order to choose the best classifier for extracting land cover features, one should study the strengths and weaknesses of these algorithms and hence conducting a performance analysis of these algorithms is very important (Lillesand, et al., 2008).
In our recent work, we had adapted the original BBO algorithm (Simon, D., 2008) and applied it for landuse / landcover feature extraction from satellite images (Panchal, V.K., et al., 2009; Goel, L., et al., 2010a; Goel, L., et al., 2012). Enriched with our experience of the above mentioned works (Panchal, V.K., et al., 2009; Goel, L., et al., 2010b), we concluded that the main characteristic of the BBO technique for satellite image classification is that this technique is flexible to classify the desirable features more efficiently than the other features and hence it shows a wide range of efficiencies in classifying different features of an image. In this paper, we present a mathematical framework for the behavioral analysis of the BBO technique over natural terrain features for satellite image classification which shows that the BBO based classifier classifies the homogeneous regions more efficiently than the heterogeneous regions of the image. Below, we summarize the main contributions of the paper: