CNN을 이용한 딥러닝 기반 하수관 손상 탐지 분류 시스템 = Damage Detection and Classification System for Sewer Inspection using Convolutional Neural Networks based on Deep Learning

국문 초록 (Abstract)

본 연구는 인공지능 분야의 딥러닝 기술을 기반으로 한 하수관 손상의 자동 탐지 분류 시스템을 제안한다. 성능의최적화를 위하여 DB 획득 시 발생된 조도 및 그림자 변화와 같은 다양한 환경변화에 강인한 시스템을 구현하였다.
제안된 시스템에서는 Convolutional Neural Network(CNN) 기반의 균열 탐지 및 손상 분류 기법을 구현하였다. 최적의 결과를 위하여 256 x 256 픽셀 해상도의 CCTV 영상 9,941개를 이용하여 CNN모델을 적용하여 손상부위에 대한딥러닝을 수행하였고 그 결과 98.76 %의 인식률을 획득하였다. 기계학습을 통한 딥러닝 모델을 기반으로 다양한 환경의 하수도 DB에서 720 x 480 픽셀 해상도의 646개의 이미지를 추출하여 성능 평가를 수행 하였다. 본 시스템은 다양한 환경에서 구축된 하수관 데이터베이스 에서 손상 유형의 자동 탐지 및 분류에 최적화된 인식률을 제시한다.

다국어 초록 (Multilingual Abstract)

We propose an automatic detection and classification system of sewer damage database based on artificial intelligenceand deep learning. In order to optimize the performance, we implemented a robust system against various environmentalvariations such as illumination and shadow changes. In our proposed system, a crack detection and damage classificationmethod using a deep learning based Convolutional Neural Network (CNN) is implemented. For optimal results, 9,941CCTV images with 256 x 256 pixel resolution were used for machine learning on the damaged area based on the CNNmodel. As a result, the recognition rate of 98.76% was obtained. Total of 646 images of 720 x 480 pixel resolution wereextracted from various sewage DB for performance evaluation. Proposed system presents the optimal recognition rate forthe automatic detection and classification of damage in the sewer DB constructed in various environments.

참고문헌 (Reference)

1 Z. Lei, "Road crack detection using deep convolutional neural network" 3708-3712, 2016

2 P. Soille, "Morphological image analysis: principles and applications, 2nd ed" Springer Science & Business Media 2013

3 D. H. Koo, "Innovative method for assessment of underground sewer pipe condition" 15 (15): 479-488, 2006

4 O. Russakovsky, "Imagenet large scale visual recognition challenge" 115 (115): 211-252, 2015

5 A. Krizhevsky, "Imagenet classification with deep convolutional neural networks" 1097-1105, 2012

6 N. Ejaz, "Efficient visual attention based framework for extracting key frames from videos" 28 (28): 34-44, 2013

7 M. K.Vairalkar, "Edge detection of images using sobel operator" 2 (2): 291-293, 2012

8 Y. J. Cha, "Deep Learning-Based Crack Damage Detection Using Convolutional Neural Networks" 32 (32): 361-378, 2017

9 M. Osama, "Classification of defects in sewer pipes using neural networks" 6 (6): 97-104, 2000

10 W. Zhang, "Automatic crack detection and classification method for subway tunnel safety monitoring" 14 (14): 19307-19328, 2014

11 M. D. Yang, "Automated diagnosis of sewer pipe defects based on machine learning approaches" 35 (35): 1327-1337, 2008

12 T. C. Su, "Application of morphological segmentation to leaking defect detection in sewer pipelines" 14 (14): 8686-8704, 2014

13 I. Khalifa, "A New Image Model for Predicting Cracks in Sewer Pipes based on Time" 87 (87): 25-32, 2014