Abyss – Machine Learning Algorithm

by Josh Prigge
Wave Blue
Dec 31, 2019

The Issue at Urban Utilities

Automated closed-circuit television (CCTV) defect identification. QUU conducts regular inspections of its sewerage mains via CCTV. The subsequent review of footage and condition assessment is undertaken manually, is resource-intensive and time-consuming. Seeking an automated solution, which can code and score asset defects, and points of interest, in accordance with the Water Services Association of Australia?s (WSAA) WSA 05-2008 Conduit Inspection Reporting Code of Australia.

The condition assessment of sewers and maintenance structures must comply with the Water Services Association of Australia?s (WSAA) WSA 05-2008 Conduit Inspection Reporting Code of Australia. The Code includes codes to describe defects, a scoring system that recognizes different relevances of particular defects on rigid and flexible pipes, revised structural and service grading thresholds, and acceptance limits for defects for newly constructed sewers.

Urban Utilities conducts regular inspections of its sewer mains via CCTV. The subsequent review of footage and condition assessment is undertaken manually, is resource-intensive and time-consuming.

The Solution

Trial and assess Abyss Solutions’ machine learning algorithm, and its ability to reduce the time required to review footage and its potential to automatically identify and code defects.

The Pilot

The trial was conducted in two phases:

Phase 1 – Urban Utilities contributed 3 kilometers (1.86 miles) of CCTV footage of AC and VC sewer reticulation network. Abyss applied their machine-learning algorithm to this footage to produce a condensed view of only the segments containing defects.

Phase 2 – The ability of the machine learning algorithm to automatically classify/code defects.

Pilot Results

Urban Utilities invested AUD101,000 (USD70,000) of which 50% was funded under arrangements delivered in the Grant Deed between the State of Queensland and State of Nevada. Phase 1 at AUD15,200 (USD11,000) and Phase 2 at AUD86,000 (USD59,000).

PHASE 1

Urban Utilities contributed 3 kilometers (1.86 miles) (64 inspection videos, 12 hours) of CCTV footage of our AC and VC sewer reticulation network. Abyss applied their machine-learning algorithm to this footage to produce a condensed view of only the segments containing defects. The automated algorithm was used to produce fault detections. The faults were classified by a human auditor, a civil engineering expert from Aurecon, who was new to the Abyss tool. Commentary provided by Aurecon has been used as a benchmark for performance evaluation.

Phase 1 – time taken assessment. The manual review 3 hours of footage took 2.5 hours to annotate (83% of the video time), while the Abyss extract tool 6.5 hours of footage took 2 hours to annotate (31% of the video time). The time taken comparison for Phase 1 shows an efficiency gain of 63%.

Phase 1 – accuracy assessment. 3 hours of footage set aside for evaluation. The algorithm detected 26 faults and missed 14 faults. It was noted that 50% of the videos in the validation batch were heavily overexposed due to the unusually bright light on the camera. It is acknowledged that the over-exposed footage did impact accuracy.

Based on the outcome and feedback, Urban Utilities directed Abyss Solutions to proceed with Phase 2. Urban Utilities also agreed on an amendment to Phase 2, asking Abyss Solutions to focus resources on further training of the algorithm to improve the accuracy in detection and ability to automatically code some of the more common defects.

PHASE 2

By the end of Stage 4, the Abyss extract tool was able to accurately identify 87% of all defects contained in the CCTV footage and was found to enable the user to complete the task of assessing the CCTV footage 2.3 times quicker for a novice user and 3-5 times quicker for more experienced users than the current fully manual method, providing higher audit consistency and reducing fatigue. Abyss Solutions also has demonstrated the fault classification on the examples of joint displacement and root detection with the respective accuracies of 82% and 85%.

Further Adoption

The adoption of this technology as a business-as-usual solution and integration in the Urban Utilities workflow is under consideration.

About the Technology

Abyss Solutions, located in Sydney, Australia was founded by four scientists and engineers from The University of Sydney in 2014. Abyss is a robotics company that combines the latest innovations in ROUVs with state of the art data analytics to provide a safer, easier, and more comprehensive underwater inspection, allowing for correct asset management decisions.

http://abysssolutions.co

 

Translate »