Cross-Bore Case Study: 70% Reduction in Review Time & 100% Accuracy
SewerAI analyzed 1,040 lateral inspection videos (64 hours of footage, 25,000 linear feet) to detect cross-bore conflicts in gas distribution lines. AutoCode reduced review time by 70% — from 108 hours to 32 hours — and achieved 97.5% initial accuracy, reaching 100% accuracy after NASSCO PACP-certified review.
The Situation: Automating Cross-Bore Analysis
To mitigate the risks of cross-bore incidents that create accidental conflicts and damage to utilities — especially side sewer laterals — most US gas utilities have invested significant resources in CCTV sewer inspection programs. These programs capture video of sewer collection pipe networks to identify previous cross-bores and support safer construction practices in new Horizontal Directional Drilling (HDD) gas installations.
The intended result: capture data on millions of miles of sewer pipes to protect communities and infrastructure. The unintended result: workflow bottlenecks and errors in data review.
The Problem: Too Much Tedious Work, Too Many Errors
Utilities task team members with watching thousands of hours of video inspections captured in the field. In addition to taking up valuable time and creating backlogs of work, this manual effort yields errors in the identification of pipe defects. Up to 20% of pipe facts are not reported with manual analysis.
The Solution: Automating Success with AutoCode
AutoCode sewer assessment AI leverages artificial intelligence and machine learning algorithms to rapidly and accurately analyze sewer system data, including CCTV footage and sensor readings. By automating the assessment process, AI identifies structural defects, predicts maintenance needs, and prioritizes rehabilitation efforts. This automatic processing is followed by manual review of AutoCode's predictions by NASSCO PACP-certified reviewers to delete false positives and enable the entry of descriptors, modifiers, clock positions, and pipe cross section percentages.
Study Specifics
- 1,040 lateral inspection videos selected — 40 with known cross-bores, 1,000 without
- Approximately 25,000 linear feet of pipe data
- Total playback time: 64 hours
Process
- All videos processed using SewerAI's proprietary AI model, specially calibrated to identify cross-bores.
- AI predictions aggregated into video clips (Positive Windows).
- Positive Windows manually reviewed by NASSCO PACP-certified technicians to identify cross-bores.
The Results: Faster. Better. More Powerful.
SewerAI's AutoCode delivered dramatic improvements in both speed and accuracy across every dimension of the cross-bore detection program:
- SewerAI narrowed down the cross-bore search to 19 of the original 64 hours of footage — just 30% of the original volume. Manual review of this 19 hours took an additional 13 hours.
- Time needed to review videos without AI: 108 hours. With SewerAI: 32 hours total.
- Review time reduced by 70% — from 108 hours to 32 hours.
- SewerAI initially correctly identified 39 out of 40 cross-bores (97.5% accuracy), increasing to 100% recall for the AI-assisted process in its entirety — a nearly 20% improvement over manual methods (80% accuracy).
- The single mis-identified cross-bore was an anomaly covered with deposits. By continuing to train the AI with additional images, errors will not be repeated — the AI gets smarter with every use.
Read the Full Case Study
Download the full SewerAI Cross-Bore Case Study to learn more about how AutoCode AI and the PIONEER platform help utilities automate lateral inspection programs, detect cross-bores, and reduce the cost and risk of lateral assessments.
