Unsupervised pixel classification in satellite imagery using a new multiobjective symmetry based clustering approach

An important approach for landcover classification in remote sensing images is by clustering the pixels in the spectral domain into several partitions. In this article the problem of clustering is posed as one of searching for some suitable number of cluster centers so that some measures of validity of the obtained partitions should be optimized. In this paper a new multiobjective clustering technique based on a newly developed point symmetry based distance is used to solve this partitioning problem. The proposed multiobjective clustering technique utilizes a recently developed multiobjective simulated annealing technique, AMOSA (archived multiobjective simulated annealing technique) to optimize two validity measures as two objective functions. Assignment of points to different clusters are done based on the newly developed point symmetry based distance rather than the Euclidean distance. Two validity measures, one measuring the total symmetry present in a partition in terms of the newly defined point symmetry based distance, and the other measuring the total compactness of the partitioning in terms of the popular Euclidean distance are optimized simultaneously to obtain the correct partitioning for a given number of clusters present in an image. Thus the proposed algorithm provides a set of final non-dominated solutions, which the user can judge relatively and pick up the most promising one according to the domain requirement. The effectiveness of this proposed clustering technique in comparison with the existing fuzzy C-means clustering is shown for classifying two remote sensing satellite images of the parts of the cities of Kolkata and Mumbai.