D6.3 Methodology for routine CH monitoring with multi-type remote sensing

The deliverable 6.3 focuses on the description of methods and tools that have been developed for the routine monitoring of the CH assets and broad areas as well, in the framework of the HYPERION project. The developed methods, apart from those related to deterioration mapping and material loss estimation, were applied on all the pilot sites (city of Rhodes, Granada, Venice and Tønsberg) and CH assets, and their products serve as reference data for the vulnerability models and post disaster damage assessment. In chapter 2, satellite remote sensing methods developed within the HYPERION project are documented. Advanced methodologies using PS and SBAS functionalities on Sentinel-1 time-series satellite data have been developed for producing long-term ground deformation maps. Additionally, several Convolutional Neural Network architectures have been applied on very high-resolution Quickbird and WV2/3 images and evaluated for their ability to produce reliable land cover change detection maps with emphasis on the increase of the impervious materials. In chapter 3, 3D modelling methods and analysis techniques were explored for the CH assets of the four pilot sites. Data acquired by cameras mounted on drones and mobile laser scanners have been used for the 3D representation of the CH assets. The produced point clouds, light models, texture models and sections provide a detailed 3D documentation for the assets and serve as input data for the SG simulator that is developed within the HYPERION project. Additionally, they serve as the basis for material loss estimation and will be used as reference data for creating 4D representations of important infrastructures over different time instances and detect structural deformations and other types of alterations. Finally, in chapter 4, an integrated methodology based on 3D models and hyperspectral data has been developed for providing deterioration and material loss maps. Data acquired by mobile hyperspectral sensors were used in order to build a Spectral Library for the CH asset materials, identify their deviations from normal situations and produce deterioration maps. SAGA-GIS functionalities have been used to provide material loss maps. Additionally, deep learning methods have been used to detect flaws and defects such as cracks on the infrastructure by analysing the data received.