INFORMATION EXTRACTION FROM REMOTE SENSING IMAGES USING OBJECT BASED IMAGE ANALYSIS

Abstract

Remote sensing images on-board from a large variety of airborne borne sensors provide valuable data for about the earth's surface. The most common application of high resolution remote sensing images is extraction of land cover information map for the city governments to seek better planning and management practices with increasing urbanization. As the spatial resolution becomes higher, more details within a land parcel are resolved in the images. As a result, both per-pixel and sub-pixel algorithms fail miserably in extracting information from high spatial resolution data, now available at 40cm resolution and at 25cm resolution in future. Therefore algorithms are needed which treat land parcel areas as single objects. This can be possible, if the algorithms are capable of including various other attributes of an object such as texture, context, shape, size etc. along with the spectral attributes of remote sensing data in the classification process. Object Based Image Analysis (OBIA) is an emerging discipline of remote sensing and geosciences which offers huge potential for processing and analysis of high resolution remote sensing data. The basic processing units of OBIA are image objects, which are contiguous pixel groups having real world connotation exhibiting useful attributes like shape, texture which single pixels do not contain. The work presented in this dissertation is aimed at investigating the potential of OBIA for processing and classification of remote sensing data and developing a methodology based on the concept of OBIA for extraction of land cover information map from high resolution multispectral images.