Delineation of Water Bodies from Satellite Images Using MATLAB

Abstract

This project aims to extract Water Bodies from high resolution satellite images. The hyperspectral images contain data about all the features. The objective is to eliminate everything but the water body pixels present in the imagery. When we are discussing about extraction, quality of the output depends mostly on the input image. Several other factors like, clarity of the image, weather of the area being photographed, clouds present in the atmosphere, the time of day, etc. come into play. Different extraction techniques are used for the purpose. Starting with analyzing the Digital Number values for getting the basic idea of the image, this thesis moves towards Thresholding which is considered to be the first step for classification. Analyzing the images is done by calculating their Radiance and Top of Atmosphere Spectral Reflectance values. Different types of Indices like Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Automated Water Extraction Index (AWEI) with shadow (‘sh’) and non-shadow (‘nsh’) variants are used for the extraction. These indices provide information about the water body pixels and they become the judging criteria for them. As we move on to different images, we put in the method of Adaptive Thresholding to them. This brings in more accuracy as the threshold values are decided according to the images. For further improvement in the extraction, method of Clustering is also applied