Crack Detection in Concrete Pipes Using Deep Learning Assisted Computer Vision

Water authorities use CCTV technology to monitor reinforced precast concrete pipes for cracking.

But the technology is not accurate enough to distinguish between wide cracks in need of repair and narrow cracks that can self-heal. This leads asset owners to spend millions of dollars unnecessarily repairing, replacing or relining in-service pipes that are still safe and performing to standard.

In this project, researchers from the University of Technology Sydney have partnered with the Concrete Pipe Association of Australasia to develop a better way of monitoring and measuring pipe cracks. They will develop deep learning assisted computer vision that will have far greater resolution than current CCTV detectors and collect more accurate data on crack widths. Their innovation will help asset owners make more informed and cost-effective decisions about which pipes need repairs.

Researchers will also gather data on acceptable crack width limitations under diverse environmental conditions. This data will then go towards a revised and more precise AS/NZS4058.

PROJECT PARTNERS

• University of Technology Sydney
• Concrete Pipe Association of Australasia