Swarm Intelligence (SI) refers to a kind of problem-solving ability that emerges by the interaction of simple information-processing units. The overall behaviour of the system results from the interactions of individuals through information sharing with each other and with their environment, i.e., the self-organized group behaviour. The chapter details the theoretical aspects and the mathematical framework of the concept of information sharing in each of the swarm intelligence techniques of Biogeography-Based Optimization (BBO), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), and Bee Colony Optimization (BCO), which are the major constituents of the SI techniques that have been used for land cover feature extraction of multi-spectral satellite images. The authors then demonstrate the results of classification after applying each of the above SI techniques presented in the chapter and calculate the classification accuracy for each in terms of the kappa coefficient generated from the error matrix obtained. For verification, they test their results on two datasets and also calculate the producer’s and the user’s accuracy separately for each land cover feature in order to explore the performance of the technique on different features of the satellite image. From the results, they conclude that the concepts of information sharing can be successfully adapted for the design of efficient algorithms that can be successfully applied for feature extraction of satellite images.
Top1. Introduction
Swarm intelligence is any attempt to design algorithm or distributed problem-solving devices inspired by collective behavior of social insect colonies or other animal societies (Bonabeau, Dorigo, & Theraulaz, 1999). The social insect colony is a distributed system comprising of direct or indirect interactions through information sharing among relatively simple (social) agents that can solve the problems in a very flexible and robust way: flexibility allow adaptation to changing environment, while robustness endows the colony with the ability to function even though some individuals may fail to perform their tasks (Goel, Gupta, & Panchal, 2011). These swarm intelligence techniques can form the basis of building an optimization algorithm which can adapt itself to suit our purpose of natural terrain feature extraction, and prove to be better by giving more accurate results than the other existing optimization techniques for certain specific applications. The swarm intelligence techniques under this category are listed below.
- •
Ant Colony Optimization
- •
Swarm Particle Optimization
- •
Biogeography Based Optimization
- •
Bacterial Foraging Optimization
In remote sensing (Lillesand & Kiefer, 2008), the problem of feature extraction has been solved by using the traditional classical approaches of artificial intelligence like Parallel-o-piped Classification (Long & Srihann, 2004), Minimum Distance to Mean Classification (Long & Srihann, 2004), Maximum Likelihood Classification (Long & Srihann, 2004) etc. A major disadvantage of the above traditional AI techniques of natural terrain feature extraction is that these techniques show limited accuracy in information retrieval and high-resolution satellite image is needed. Also these techniques are insensitive to different degrees of variance in the spectral response data. The computational intelligence techniques like the fuzzy sets based classifier and path planner and the rough set classifier, which have been used recently, are not able to provide good result in case of ambiguity and also result in inaccuracy with low spatial resolution. Also these are not able to handle the continuous and the crisp data separately. Hence we shifted to the fundamentals from Swarm Intelligence, an optimized approach of feature extraction of satellite multi-spectral images. Swarm Intelligence provides a good number of accuracy even with low spatial resolution image. This technique, with lower cost and higher degree of accuracy, will be able to replace high resolution high cost satellite imageries (Lillesand & Kiefer, 2008). Particles involved in these techniques can either interact directly through waggle dancing (BCO), global-best positions (PSO), etc. Or indirectly through the environment via pheromone update (ACO), migration of SIV’s between candidate solutions (BBO) etc. Ant colony optimization was proposed by Marco Dorigo (Dorigo & Stuetzle, 2004; Dorigo, Maniezzo, & Colorni, 1996) in 1992. ACO has since been applied to many search, optimisation and anticipatory problems (Piatrik & Izquierdo, 2006). Similarly, particle swarm optimization algorithm which simulates the social behaviour of bird flocking or fish schooling was introduced by Eberhart and Kennedy (Bratton & Kennedy, 2007) in 1995. PSO is a population based stochastic optimization technique and is well adapted to the optimization of nonlinear functions in multidimensional space (Kennedy & Eberhart, 1995). PSO has been applied to several real-world problems. Likewise, motivated by the foraging behaviour of honeybees, researchers Riley et al., Karaboga and Basturk (2005), proposed artificial bee colony algorithm for solving various optimization problems (Karaboga, 2005; Karaboga & Basturk, 2007; Karaboga & Akay, 2009). This algorithm is easy to implement and robust. Also a novel approach is added in this category called biogeography-based optimization and was recently proposed by Dan Simon (2008).The key to maintaining global, self-organized behaviour is social interaction. The fundamental principle of these techniques is cooperation and sharing of knowledge. The basis of the increased intelligence is the shared information discovered individually and communicated to the swarm by different mechanisms of social interaction. In this way, intelligent solutions to problems naturally emerge from the self-organization and communication among simple individuals.