Α well-developed road network is essential for the efficient movement of people and goods. Therefore, maintenance of road networks is paramount to ensure their continued functionality and the economic and social wellbeing they support. When it comes to road maintenance, a necessary condition for optimal resource allocation is the mapping of the road condition.
In this work, an integrated system for road surface damage detection is presented. Τhe detection of anomalies is based on the vibrations received by a moving vehicle the moment it passes on top of a defect. The vibrations are sensed by an inertia measurement unit (IMU), while the geographic location of the damage is determined using a GPS sensor.
The system is based on an embedded device designed as part of this diploma thesis which was used in real-life conditions to collect data and conduct experiments. For the development of the related software, the ROS2 framework was used.
The implementation addresses several challenges of a real-world scenario, such as the existence of noise in the sensor signals and the random orientation in which the device is placed on the vehicles. A heuristic algorithmic approach was developed to identify road imperfections through the IMU measurements. The analysis on hand-labeled data showed that a F1-Measure score of ≈ 0.83 is achieved when it comes to identifying anomalies in the signal.
Finally, a logic for storing road anomalies in a database is presented that clusters its entries to extract a better estimate of the characteristics of a road defect that is met more than once in the data.
Original text (in Greek) can be found here.
A dataset was created with synchronized IMU, GPS and Camera data. This can be found here.
Folder ros_packages contains ROS2 packages meant for use in a ROS2 workspace. Check README.md inside directory for more info
However, folder also contains scripts for pushing these packages to device developed and used during thesis. These scripts can be found under sync_rpi4.
Furthermore, genesis contains a systemd service that automatically starts ROS2 ecosystem on device.
Folder python_files contains files related to data handling and data analysis. Check README.md inside this directory for more info.
A linux distro is recommended for using this repo. Ubuntu 22.04 was the OS used during this project.
You need to install ROS2. Version used was ROS2 Humble, you can find its installation instructions here.
Python requirements file can be found inside python_files folder. You also need to include its subdirectories to PYTHONPATH in order for imports to work properly.