Seoul to Utilize AI for 'Sewer Pipeline Defect Detection'... Expansion to All Districts Next Year

AI Automatically Identifies Defect Types and Locations for Notification
Improved Detection Accuracy and Reduced Reading Time Enable Proactive Accident Prevention and Enhanced Administrative Efficiency

[Asia Economy Reporter Lim Cheol-young] The Seoul Metropolitan Government is utilizing artificial intelligence (AI) technology for defect detection tasks such as perforations and collapses in sewer pipelines, which were previously identified by human visual inspection.

On the 23rd, the Seoul Digital Foundation announced that it developed last year an "AI-based Sewer Pipeline Defect Detection System" that automatically identifies defects in sewer pipeline CCTV footage using computer vision technology, which enables computers to recognize videos and images.

Previously, defect detection in sewer pipelines involved inspectors manually reviewing CCTV footage of circular sewer pipelines spanning a total length of 9,000 km across Seoul. This video analysis required considerable time and cost, and the inspection results varied depending on the inspector’s expertise.

Going forward, AI will analyze CCTV footage to automatically identify the types and locations of defects in sewer pipelines and display the results on the inspector’s computer. To this end, the foundation identified 10 types of sewer pipeline defects last year and trained the AI with 5,000 CCTV video data samples.

The Seoul Metropolitan Government expects that with the introduction of the "AI-based Sewer Pipeline Defect Detection System," the accuracy of defect detection will improve, and the analysis time will be shortened, enabling more proactive maintenance measures to prevent accidents. It will also reduce the fatigue of field personnel and significantly enhance administrative efficiency.

The foundation conducted a pilot application of the system in Gwangjin-gu and Dongdaemun-gu districts of Seoul at the end of April this year. Simultaneously, it is conducting additional research to expand the defect types to 19 and improve detection accuracy. Based on the pilot project and research results, the system will be expanded to all districts of Seoul next year. Seoul plans to build a cloud-based defect detection system environment for convenient use by districts and to expand AI-based municipal services.

Additionally, the Seoul Digital Foundation will launch a project in June called "Aerial Image Urban Building Change Detection Data," which detects new buildings, demolished buildings, and changes due to reconstruction or extension using AI.

This project involves building AI training data from 146,000 buildings (covering an area of 160 km²) and a total of 50,000 aerial images owned by Seoul, and developing an AI model that automatically detects changes in urban buildings. The foundation will receive high-quality aerial images from Seoul, perform corrections for shadows and building distortions, and maximize the identification capability of the image maps to build training data.

With the introduction of AI technology, tasks that previously required a few experts to visually interpret building changes will be automated, speeding up processing. The accuracy of interpretation will also improve, enabling rapid identification of unauthorized buildings and allowing prompt responses to collapse or fire accidents caused by such unauthorized constructions.

The foundation plans to expand services beyond building changes to include roads, algae management, and biotopes (biological habitats), and link these with S-MAP, a 3D virtual reality representation of the entire Seoul area, to support Seoul in providing systematic urban change detection services.

Lee Won-mok, Acting Chairman of the Seoul Digital Foundation, stated, "The Seoul Digital Foundation aims to fully introduce innovative administrative systems applying AI to enhance urban administration efficiency and prevent safety accidents. We will improve the completeness of AI-based innovative administrative services by advancing the technology to a level applicable in the field, reflecting on-site demands."