Design and development of a generic framework for surface water delineation and monitoring using a hybrid level set algorithm on landsat multi spectral data

Abstract

Surface water bodies are critical to the existence and sustenance of civilizations. Water bodies in urban cities across the world have undergone drastic decline in quality and quantity. This has been the result of a multitude of reasons like increase in population, urbanization and encroachment. Monitoring changes to water bodies is a newlinenecessary requirement in devising strategies to conserve them. This thesis proposes newlinea generic framework for monitoring and forecasting changes in the surface area of newlinelakes using a hybrid level set algorithm for water body delineation followed by a double exponential smoothing model for forecasting. The proposed hybrid level set algorithm combines the advantages of edge based and region based level sets. An edge detection term is introduced into the formulation which improves the delineation accuracy by forcing the level set evolution to stop at the boundaries of the region of interest. The performance of the algorithm was analyzed using Pearson s Correlation Co-effcient (PCC), Structural Similarity Index (SSIM) and Dice Similarity index and found to have superior performance compared to established methods in the literature. The study uses Landsat multi-spectral data for the last 30 years to build the proposed framework for forecasting the changes in the surface area of water bodies. The experiments were conducted for nine lakes in Bangalore, a fast growing city in India, and a steady decrease in the surface area is observed for most of the lakes that were studied. The city s renovation attempts have also seen that the some of the lakes are sustaining the rapid urbanization. The proposed forecast model has yielded acceptable results with an average error of 0.22% and a correlation coeffcient of 0.94 between the actual surface area and the forecasted surface area. The framework can be customized in the future to study specifc water bodies by plugging in external newlineparameters to improve the forecasting accuracy.