This study presents a comprehensive approach to the automatic mapping and analysis of the irrigation grid of the Cavour Canal, located in Piemonte, Italy, addressing the unique challenges of applying the methodology to larger irrigation systems. The study focuses on identifying irrigation channels using aerial multispectral data to create a polygon feature of channels and determine flow directions in the channels. The research uses supervised pixel-based classification methods on CIR (Color-Infrared) and RGB orthophotos for the detection of water bodies, using four classification algorithms: K-Nearest Neighbor, Random Trees, Support Vector Machine, and Maximum Likelihood. The classification performance and effectiveness of each algorithm are assessed by comparing the results of each classifier, with the aim of improving the water detection process for these narrow, complex irrigation channels.