ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Publications Copernicus
Download
Citation
Articles | Volume V-4-2020
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., V-4-2020, 19–26, 2020
https://doi.org/10.5194/isprs-annals-V-4-2020-19-2020
ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., V-4-2020, 19–26, 2020
https://doi.org/10.5194/isprs-annals-V-4-2020-19-2020

  03 Aug 2020

03 Aug 2020

ASSESSING RESILIENCE OF INFRASTRUCTURES TOWARDS EXOGENOUS EVENTS BY USING PS-INSAR-BASED SURFACE MOTION ESTIMATES AND MACHINE LEARNING REGRESSION TECHNIQUES

N. Fiorentini1, M. Maboudi2, M. Losa1, and M. Gerke2 N. Fiorentini et al.
  • 1Department of Civil and Industrial Engineering, Engineering School of the University of Pisa, Largo Lucio Lazzarino 1, 56122, Pisa, Italy
  • 2Institute of Geodesy and Photogrammetry, Technische Universität Braunschweig, Bienroder Weg 81, 38106, Braunschweig, Germany

Keywords: Resilience of Infrastructures, Infrastructure Maintenance, PS-InSAR, Machine Learning, Regression Models, Regression Tree, Subsidence, Surface Motion

Abstract. Technologically advanced strategies in infrastructural maintenance are increasingly required in countries such as Italy, where recovery and rehabilitation interventions are preferred to new works. For this purpose, Interferometric Synthetic Aperture Radar (InSAR) techniques have been employed in recent years, achieving reliable outcomes in the identification of infrastructural instabilities. Nevertheless, using the InSAR survey exclusively, it is not feasible to recognize the reasons for such vulnerabilities, and further in-depth investigations are essential.

The primary purpose of this paper is to predict infrastructural displacements connected to surface motion and the related causes by combining InSAR techniques and Machine Learning algorithms. The development and application of a Regression Tree-based algorithm have been carried out for estimating the displacement of road pavement structures detected by the Persistent Scatterer InSAR technique.

The study area is located in the province of Pistoia, Tuscany, Italy. Sentinel-1 images from 2014 to 2019 were used for the interferometric process, and a set of 29 environmental parameters was collected in a GIS platform. The database is randomly split into a Training (70%) and Test sets (30%). With the Training set, through a 10-Fold Cross-Validation, the model is trained, validated, and the Goodness-of-Fit is evaluated. Also, with the Test set, the Predictive Performance of the model is assessed. Lastly, we applied the model onto a stretch of a two-lane rural road that crosses the area. Results show that the suggested procedure can be used for supporting decision-making processes on planning road maintenance by National Road Authorities.