1 minute read

Road maintenance is a crucial aspect of ensuring safe and efficient transportation. In Germany alone, over 10 billion Euros were spent on road maintenance projects in 2015, as reported by the German Federal Statistical Office. Poor road conditions can affect a wide range of factors, from vehicle operating costs and fuel efficiency to driving comfort and, most importantly, road safety. In 2015, over 1200 accidents in Germany were related to road hazards.

Inspecting roads regularly is crucial in reducing the risk of accidents, but in many cases, smaller communities lack the resources to do so effectively. This is where our innovative solution comes in: an automatic road damage detection and labelling system. Our system aims to improve road condition monitoring and reduce the manual and financial effort required for regular inspections.

Our method is based on a low-cost measurement device, consisting of an inertial sensor and GPS sensor, which is placed near the center of gravity of the vehicle. Using a machine learning algorithm and statistics calculated from vibrations and dynamics, the system can classify road infrastructure features and estimate the condition. The system also has the ability to collect the required training data automatically, avoiding the need for a physical model and manual training for each vehicle.

In addition to providing periodic monitoring of roads, our system can also be used to compare output from multiple vehicles, improving the prediction of road conditions and enabling trend recognition. With our proposed method, road safety and quality can be improved with comparatively little expense.

I started on this project in my bachelor’s thesis and continued the work for another year. In that time, we published our findings in four peer-reviewed journals:

  • Learning from the crowd: Road infrastructure monitoring system [Open-Access]
  • Multiple vehicle fusion for a robust road condition estimation based on vehicle sensors and data mining [Open-Access]
  • Characterization of Road Condition with Data Mining Based on Measured Kinematic Vehicle Parameters [Open-Access]
  • A novel approach to label road defects in video data: semi-automated video analysis [Open-Access]