Automated Image Learning to Improve Sewer Pipeline Inspections

3. Underground City Oy, Finland

Abstract: The need for faster, more reliable and more objective need for sewer inspections and their analyses have been recognized. In Finland, only much less than 5% of the sewer pipelines are inspected annually. Of the inspected pipelines, approx. 60% has been found to be in good condition, and 8% of the inspected pipelines need some actual means to repair or replace in near future. In this project, a method to use machine learning technique to inspect and analyze the sewer inspection filming was piloted by two consultants. Both pilot projects proved that this technique is auspicious and feasible method to help and accelerate the speed of pipeline inspections, when the CCTV inspection is made with a digital filming and scanning camera. DigiSewer® was used in this piloting project. Further development steps will include the use of a screening method with a cloud-based data transfer, improving the process of sewer inspections further including the information of operational condition of the pipelines.