Road damage detection over road scanner images using deep convolutional neural network
- Authors
- Jo, H.; Kim, D.; Pak, K.-W.; Kim, M.
- Issue Date
- Oct-2020
- Publisher
- ICIC International
- Keywords
- Convolutional neural network; Deep learning; Image classification; Pavement distress detection; Road crack detection
- Citation
- ICIC Express Letters, v.14, no.10, pp.1001 - 1008
- Journal Title
- ICIC Express Letters
- Volume
- 14
- Number
- 10
- Start Page
- 1001
- End Page
- 1008
- URI
- http://scholarworks.bwise.kr/ssu/handle/2018.sw.ssu/39772
- DOI
- 10.24507/icicel.14.10.1001
- ISSN
- 1881-803X
- Abstract
- This paper deals with the automatic detection and classification of road cracks. Determining the type and severity of a crack under the conventional PMS (Pavement Management System) is a challenging task in terms of process efficiency and consistency since it relies on human intervention of many trained workers. Many studies that reduce this human reliance have been conducted for the automation of crack detection, and thus improve the whole pavement management process significantly. Recently, neural networks have been actively tried for road damage detection and have achieved remarkable results. However, since many researches have been done on the basis of limited road images of a few countries or well-cleansed example data, more adaptive research is needed to obtain effective results for road conditions of individual countries. In this paper, 3 CNN-based models for detecting road cracks are tested using actual data from Korea. By doing multiple experiments for 6 types of cracks, several approaches to better detect cracks in Korean road conditions are examined. Based on the result of this study, many researches will be able to further enhance their studies while adapting their models to better fit to their respective domestic road conditions. ICIC International © 2020
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - College of Engineering > ETC > 1. Journal Articles
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.