Skip to content

Godsihyeong/Development-of-a-real-time-sewer-pipe-defect-detection-algorithm-using-YOLOv8

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 

History

24 Commits
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Development of a real-time sewer pipe defect detection algorithm using YOLO

Abstract

Sewer pipelines play a vital role in maintaining urban sanitation and public health, contributing to sustainable development. Early detection and maintenance of physical defects can reduce management costs and improve performance. Traditional CCTV inspections, consisting of two main stages: on-site video collection and office evaluation, face limitations such as technician fatigue and time consumption. To address these issues, a YOLO-based deep learning model for automated defect detection has been proposed. Using sewer pipeline interior images provided by AI Hub, this study classified eight defect types, including joint offsets, connector protrusions, and sediment deposits. The model achieved over 90% mAP accuracy (based on an IOU of 0.5) regardless of lighting and background noise. This technology is expected to enhance the efficiency of sewer maintenance, contributing to safer and more sustainable urban environments.

Introduction

ν•˜μˆ˜λ„ μ‹œμŠ€ν…œμ€ λ„μ‹œν™” 및 인ꡬ λ°€μ§‘ μ§€μ—­μ˜ 지속 κ°€λŠ₯ν•œ λ°œμ „μ— μžˆμ–΄ μ€‘λŒ€ν•œ 역할을 μˆ˜ν–‰ν•΄μ™”λ‹€. 이와 같은 ν•˜μˆ˜λ„ μ‹œμŠ€ν…œμ˜ μ€‘μš”μ„±μ„ μΈμ§€ν•˜κ³ , μ§€λ°© μ •λΆ€ 및 자치 λ‹¨μ²΄λŠ” ν•˜μˆ˜ μ‹œμŠ€ν…œμ˜ 효율적 관리와 μœ μ§€λ₯Ό μœ„ν•˜μ—¬ 2024년에 2μ‘° 7,692μ–΅ μ›μ˜ μ˜ˆμ‚°μ„ μ±…μ •ν•˜μ˜€κ³  2023λ…„ λŒ€λΉ„ 5,567μ–΅ 원(25%) μ¦κ°€ν•˜μ˜€λ‹€(ν™˜κ²½λΆ€, 2024). ν•˜μˆ˜λ„ μ‹œμŠ€ν…œμ˜ 진단과 μœ μ§€ λ³΄μˆ˜λŠ” ꡬ쑰물의 문제λ₯Ό μ‹ μ†νžˆ λ°œκ²¬Β·ν•΄κ²°ν•˜μ—¬ μ„±λŠ₯을 κ°œμ„ ν•˜κ³  관리 λΉ„μš©μ„ μ ˆκ°ν•˜λ©°, μ•ˆμ •μ„±κ³Ό 졜적의 μ„œλΉ„μŠ€ μ œκ³΅μ„ λͺ©μ μœΌλ‘œ ν•œλ‹€.

ν˜„μž¬ CCTV κ²€μ‚¬λŠ” ν•˜μˆ˜λ„ μ‹œμŠ€ν…œμ˜ 진단 및 μœ μ§€ 보수 κ³Όμ •μ—μ„œ ν•„μˆ˜μ μΈ λ„κ΅¬λ‘œ μΈμ‹λ˜κ³  μžˆλ‹€. λ ˆμ΄μ € 기반 μ‹œμŠ€ν…œ, 초음파 μ„Όμ„œ, 적외선 열상 카메라 λ“± λ‹€λ₯Έ 기술적 접근법과 λΉ„κ΅ν•˜μ˜€μ„ λ•Œ, CCTV κ²€μ‚¬λŠ” 뢄석에 μš©μ΄ν•œ μ‹œκ°μ  자료λ₯Ό μ œκ³΅ν•œλ‹€λŠ” μž₯점을 κ°–λŠ”λ‹€. CCTV 검사 과정은 λΉ„λ””μ˜€ μˆ˜μ§‘κ³Ό μ‚¬λ¬΄μ‹€μ—μ„œ ν›ˆλ ¨λœ κΈ°μˆ μžμ— μ˜ν•œ λΉ„λ””μ˜€ λΆ„μ„μœΌλ‘œ, 두 μ£Όμš” λ‹¨κ³„λ‘œ κ΅¬μ„±λœλ‹€.

2022λ…„ κΈ°μ€€ 총길이 16만 8,786km인 ν•˜μˆ˜κ΄€λ‘œμ˜ 결함 λ°œκ²¬μ„ μœ„ν•΄ μ†Œμš”λ˜λŠ” μ‹œκ°„κ³Ό 기술자의 ν”Όλ‘œλŠ” ν‰κ°€μ˜ νš¨μœ¨μ„±μ— 뢀정적인 영ν–₯을 λ―ΈμΉ  수 μžˆλ‹€.(ν™˜κ²½λΆ€, 2024). 졜근 λ”₯λŸ¬λ‹ 기술의 적용이 ν™•λŒ€λ¨μ— 따라, μ΄λŸ¬ν•œ 문제λ₯Ό ν•΄κ²°ν•˜κΈ° μœ„ν•΄ λ§Žμ€ μ—°κ΅¬μžλ“€μ΄ 컴퓨터 λΉ„μ „ κΈ°μˆ μ„ μ΄μš©ν•œ μžλ™ν™” μ ‘κ·Ό 방식을 λͺ¨μƒ‰ν•˜κ³  μžˆλ‹€(μž„μˆ˜ν˜„ et al., 2018), (Cheng et al., 2018). ν•˜μ§€λ§Œ μ„ ν–‰ μ—°κ΅¬μ˜ κ²½μš°λŠ” Two stage detector둜 μΆ”λ‘  속도가 느렀 μ‹€μ‹œκ°„ 탐지가 λΆˆκ°€λŠ₯ν•˜λ‹€λŠ” ν•œκ³„κ°€ μ‘΄μž¬ν•œλ‹€.

특히, β€˜You Only Look Once (YOLO)’ μ•Œκ³ λ¦¬μ¦˜μ€ 개체 감지 기술 쀑 ν•˜λ‚˜λ‘œ, 고속 μ²˜λ¦¬μ™€ 높은 정확도λ₯Ό κ²ΈλΉ„ν•΄ μ‹€μ‹œκ°„ ν•˜μˆ˜κ΄€λ‘œ 결함 탐지에 특히 μ ν•©ν•˜λ‹€(Redmon, J. et al., 2016). λ”°λΌμ„œ, λ³Έ λ…Όλ¬Έμ—μ„œλŠ” YOLOv8을 ν™œμš©ν•˜μ—¬ ν•˜μˆ˜κ΄€λ‘œμ˜ 결함을 νƒμ§€ν•˜λŠ” μ‹œμŠ€ν…œμ„ μ œμ•ˆν•œλ‹€. λ³Έ μ‹œμŠ€ν…œμ€ 이미지λ₯Ό λ‹€μš΄μ‚¬μ΄μ§•ν•˜μ—¬ 기쑴보닀 ν•™μŠ΅ 및 μΆ”λ‘  속도λ₯Ό μ¦κ°€μ‹œν‚€λ©°, μ‹€μ‹œκ°„ 탐지λ₯Ό 톡해 기쑴의 μˆ˜λ™ 검사 λ°©μ‹μ˜ ν•œκ³„λ₯Ό κ·Ήλ³΅ν•˜κ³ , ν•˜μˆ˜λ„ κ΄€λ¦¬μ˜ νš¨μœ¨μ„±κ³Ό 정확성을 κ°œμ„ ν•˜λŠ” ν˜μ‹ μ μΈ μ ‘κ·Ό 방식을 μ œμ‹œν•œλ‹€.

Workflow

Dataset

전체 이미지 λ‹€μš΄λ‘œλ“œ 클릭

Sewerpipe
β”œβ”€β”€ Dataset
    β”œβ”€β”€ train
        β”œβ”€β”€ CC
        β”œβ”€β”€ CL
        β”œβ”€β”€ ...
        └── ...
    β”œβ”€β”€ val
    └── test
β”œβ”€β”€ Dataset(+Background)
    β”œβ”€β”€ train
    β”œβ”€β”€ ...
    └── test

Result

mAP50 Recall Inference Time
0.7474 0.7001 11.1ms

Optimal Parameters

Batch size Optimizer Cos_lr Pretrained Background Label Smoothing
128 Adam False True X 0.3

Information

@inproceedings{title={Development of a real-time sewer pipe defect detection algorithm using YOLO},
author={Jo, Yeongjoo and Yeo, Sihyeong and Kim, Changwoo},
booktitle={2024 κ³΅λ™ν•™μˆ λ°œν‘œνšŒ}
year={2024},
month={March},
organization={Korean Society on Water Environment (ν•œκ΅­λ¬Όν™˜κ²½ν•™νšŒ), Korean Society of Water & Wastewater (λŒ€ν•œμƒν•˜μˆ˜λ„ν•™νšŒ)},
}

License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

About

🌊 A real-time object detection model for sewer pipe defect

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages