Road Surface Damage Detection Algorithms and its Road Test using Asymmetric Auto-Encoder Deep Neural Network

In modern times, road pavement is mainly made of asphalt. Such roads enable stable high-speed travel of the vehicle due to excellent road surface packaging and provide a good ride for the driver. However, the road surface is repeatedly damaged by various causes. The typical problems are cracks, deformation, Received (December 10, 2018), Review Result(December 26, 2018) Accepted(January 5, 2019), Published(January 31, 2019) Research Specialist, 10223 Korea Institute of Civil Engineering and Building Technology, (Daehwa-Dong)283, Goyangdae-Ro, Ilsanseo-Gu, Goyang-Si, Gyeonggi-Do, Korea email: seungboshim@kict.re.kr, chanjunchun@kict.re.kr, smkang0521kict.re.kr (Corresponding Author) Senior Research Fellow, 10223 Korea Institute of Civil Engineering and Building Technology, (Daehwa-Dong)283, Goyangdae-Ro, Ilsanseo-Gu, Goyang-Si, Gyeonggi-Do, Korea email: skryu@kict.re.kr * 본 연구는 국토교통부 국토교통기술사업화지원사업의 연구비지원(과제번호 18TBIP-C144255-01) 에 의해 수행되었습니다. Road Surface Damage Detection Algorithms and its Road Test using Asymmetric Auto-Encoder Deep Neural Network Copyright c 2019 HSST 2 and road surface flaws, which bring about traffic accidents causing various personal injury. In order to prevent it, an efficient road surface management technique is required and various methods have been developed. Among them, a method using a black box type image acquisition skill has been proposed these days. Although various image processing techniques have existed using this method, image recognition technology based on deep neural network has been most actively studied these days. This paper also introduces the results of the image recognition algorithm using the deep neural network and the its real driving tests. The first is the asymmetric auto-encoded deep neural network for determining the road surface damage location. This deep neural network takes the image as input and determines whether the road surface is damaged or not, and detects the damaged area. The second is the road test that is performed in outdoor road environment using the above and the performance of the image recognition technology is evaluated.

[1]  Kaiming He,et al.  Feature Pyramid Networks for Object Detection , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[2]  Youngtae Jo,et al.  Pothole Detection System Using a Black-box Camera , 2015, Sensors.

[3]  Ian D. Reid,et al.  RefineNet: Multi-path Refinement Networks for High-Resolution Semantic Segmentation , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[4]  Xiaogang Wang,et al.  Pyramid Scene Parsing Network , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[5]  Taehyeong Kim,et al.  System and Method for Detecting Potholes based on Video Data , 2014 .

[6]  Vladlen Koltun,et al.  Multi-Scale Context Aggregation by Dilated Convolutions , 2015, ICLR.

[7]  Ross B. Girshick,et al.  Fast R-CNN , 2015, 1504.08083.

[8]  Roberto Cipolla,et al.  SegNet: A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[9]  Ross B. Girshick,et al.  Mask R-CNN , 2017, 1703.06870.

[10]  Christian Koch,et al.  Pothole detection in asphalt pavement images , 2011, Adv. Eng. Informatics.

[11]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.