LAND INFORMATION EXTRACTION BY USING MODIFIED SEGMENTATION TECHNIQUE FOR HIGH RESOLUTION SATELLITE IMAGE

Abstract

The advancement of technology in satellite system has drastically improved the quality of images which we call high resolution images. Today we have many satellites which provide high resolution images such as QUICKBIRD. IKONOS, WORLD-VIEW etc. High resolution provides much greater detail of information such as Buildings or Roads etc. can be seen clearly which helps in detailed mapping. Now the question arises how we can extract these Land Objects which contain various information. Traditionally we use manual digitization in different sothvare like Arc-Map to define the Land Object boundary but this process is quite time taking and not appropriate for the changing Land details. Another problem with this technique is that it requires much human effort and there is no any shorter path available with this method. In this modern world we need to utilize the efficiency of computer based algorithms and some fast techniques which can extract the Land boundaries and the information associated with them such as their area, boundary detail etc. Object Based techniques are one of the methods to extract Land Objects but they are associated with the problem of appropriate Segmentation. Defining the Land boundary is first task which is necessary to extract the information. This Dissertation includes Modified Segmentation Technique which utilizes the Edge Detection followed by Region Growing Technique to preserve the Land boundaries. After defining the boundary of various Land Objects, Object Based Analysis can be done to assign class to different Land Object present in the image. So this Dissertation aims to use a Modified Segmentation Technique and Object Based Analysis to extract the Land Information automatically which reduces the time and human effort and can be used for detailed mapping.