Identification of Land Area Using Image Processing Technique

Abstract

Segmentation and classification of high resolution satellite images is a difficult task due to the fact that it is no useful meaning to carry out this task on a pixel-by-pixel method. The fine spatial resolution infers that each object is an accumulation of various pixels in spatial closeness , and exact categorizing requires that this aspect be easily considered. K-means clustering algorithm is considered as a superior method for classifying high resolution satellite images. k- means clustering or Lloyd’s algorithm, is a repetitive, data-partitioning algorithm that allocates n observations to precisely one of clusters determined by centroids, where k is set to initial value before the algorithm starts. The obtained regions are categorized using a minimum distance decision rule. This method fundamentally minimizes the mixed pixel problem suffered by pixel based methods. In this work, we used K-means and fuzzy C means clustering algorithm to classify satellite imagery into specific objects within its boundaries and environmental planning purposes. The various pixels are divided into clusters. A centre point is found for all the centroids. The entire clusters are classified based on this centroid. In our project we are using k-means clustering algorithm and Fuzzy c means to separate the pixels of the satellite images and processing it with the help of matlab software and naming the clusters according to their color configurations. Using this clustering method for satellite images we can name the places in the image based on color. For example green color indicates forest area, blue indicates water mass, brown indicates muddy area and more. Fuzzy C-means is independent of initial clusters and better clustering results are obtained. There are other types of clustering methods like partitioning method, hierarchical method, density-based clustering and model-based clustering.