Flow-Path Fitting from Images with Support Vector Regression for River Health Assessment

This study proposes an accurate river flow-path estimation method, which is pivotal for effective environmental monitoring and managing flood risks. Despite the advantages with a previous Fourier basis (FB) approach in capturing river flow-path from images, challenges persist in noise handling, outlier management, and ensuring computational efficiency in complex environmental contexts. This study introduces support vector regression (SVR) with a radial basis function (RBF) kernel as a robust alternative, renowned for its adeptness in regression tasks, particularly in managing nonlinear data relationships and its resilience against noise and outliers. Integrating SVR within a cyber-physical system (CPS) for river health assessment, this research aims to surmount the limitations of FB, thereby enhancing the accuracy and adaptability of flow-path estimation. Through a meticulous comparative analysis against FB method the SVR's superior performance is demonstrated in flow-path estimation accuracy and computational efficiency. This underscores SVR's potential to improve river monitoring methodologies and foster the development of more intelligent, adaptive flood management strategies.