SCISYS developed a solution to automate the analysis of remote sensing tunnel inspections.
Strategically important tunnels require regular inspections to determine and monitor the condition and identify the requirements for repairs.
This is often carried out by experienced civil engineers. This can, however, introduce some significant risks and the results of the inspection can be prone to significant variation due to their subjective nature.
To improve data gathering remote sensing is becoming more widespread – such as the use of LIDAR and photogrammetry. Classifying the condition of the structure or monitoring changes remains a manual job requiring a high level of knowledge and experience.
SCISYS developed a solution to automate the analysis of such remote sensing tunnel inspections.
It was able to do so by leveraging its experience in machine learning and artificial intelligence developed over the last 20 years through its work for the European Space Agency to support Mars exploration.
SCISYS developed a portal where tunnel inspection data could be uploaded, and training and labelling of datasets could be performed. Once the model was trained, an automated report was generated displaying an in-depth analysis of the tunnel’s condition, locating and identifying all the defects within the tunnel.
Scale and Consistency
The project demonstrated that it was possible to use deep learning based artificial networks to accurately classify defects in remote sensing data of a tunnel.
It was shown that this could be achieved at scale and much faster than a human counterpart. This opens exciting possibilities of how tunnel inspection could become more automated, produce more consistent results and performed on a much greater scale than currently undertaken.
ELAN VALLEY AQUEDUCT ROBOTICS SURVEYING