Integration of accelerometer and video data for road surface condition evaluation using deep learning models

Abstract

The International Roughness Index (IRI) is an important index for assessing road surface conditions and planning road maintenance. Typically, IRI is measured using equipment like Laser Profilers, Laser Crack Measurement Systems, and Walking Profilers, which require experts to control and evaluate the results. These devices also have high costs, limiting their availability and affecting the frequency of data collection. This study presents a new method for estimating IRI using data collected from a GoPro camera, which includes acceleration data and images of road surfaces. The images are processed through an object detection process to identify the type of road damage and indicate the road distress status. The road distress data is then combined with the acceleration data and processed through regression models to predict IRI values, which are compared with data collected from survey vehicles. The proposed method demonstrates the potential to use low-cost tools to increase data collection frequency and accurately calculate IRI. This research aims to be useful for agencies or developers who need to collect IRI values but face equipment and budget limitations.