A methodology for quantifying and improving pavement condition estimation and forecasting by integrating smartphone and 3D laser data

Abstract

This thesis aims to combine data from accurate but expensive 3D laser scanners and inexpensive smartphones for improving pavement condition estimation and forecasting using both technologies. This thesis presents 1) A methodology for registering both 3D laser data and smartphone sensor data onto a common GIS model of the road network; 2) A methodology for single-run pavement condition estimation using smartphone data trained using labeled data from a 3D laser scanner; 3) A methodology for combining multiple-run pavement condition estimates from both smartphones and 3D laser scanners with an associated confidence level; and 4) A methodology to improve pavement condition forecasting using 3D laser data by combining updated evidence obtained from crowdsourced smartphone data. Data was collected on a test route consisting of diverse pavement conditions and road classifications to test and validate the proposed methodologies. The data registration methodology was validated by observing a large peak in cross-correlation between pavement distress metrics calculated from multiple runs registered using the proposed methodology. The methodologies for pavement condition estimation and forecasting were validated by observing low median error for International Roughness Index (IRI) estimation and forecasting. The proposed method for associating a confidence level with each estimate was able to successfully separate low and high error estimates. This research will strongly improve the utility of both 3D laser technologies and smartphones for pavement condition assessment.