Automatic Mapping of irrigation grid of Cavour Channel through aerial multispectral data

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. Following the classification step , the results are simplified to some mathematical models, in order to pick the best one that simulates the results of classification. The models are trained based on histogram analysis of each band of data. The purpose of this step is to validate the process of classification in order to assess their capability of propagation for other study areas. The detected water pixels were then used to generate polygons, representing the irrigation channels within the study area. Buffer zones around channel features are created to ensure that polygons accurately reflect channel segments while excluding misclassified pixels of shadows or roads. A model using the tools of ArcGIS Pro is introduced to generate polygons of channels from their scattered pixels. Ultimately Digital Terrain Model (DTM) together with line features of the channels are used to determine the flow directions in each channel segment. This step is performed based on elevation differences between the two endpoints of each channel, so that the flow across the irrigation network may be visualized.