Motivation

Rail transportation plays a crucial role in environmentally friendly transportation of people and goods. In this context, the safety and reliability of rail infrastructure are particularly important and must be ensured through appropriate maintenance work. Rail inspection is currently carried out through regular checks by personnel or through irregular, costly measurement vehicles. The goal of this project is to install an intelligent camera system on regular trains and to examine the rail infrastructure using image-based algorithms. In addition to detecting rail damage, foreign objects on the track are of particular interest.

Challenges and Objectives

To ensure real-time capability of the system, both the camera sensors and the algorithms and computer components used are of great importance. Since rail damage and foreign objects are relatively rare, the algorithms should be able to detect deviations from a learned concept of normalcy. Various machine learning approaches have been tested and analyzed to implement a system with satisfactory accuracy under limited computer resources.

Outlook

The system will be further developed to enable deployment from other perspectives as well. This includes, among other things, investigating the influence of expanded field of view, different resolutions, and environmental influences. The funding for the follow-up project OCTAVE has already been secured.