Goud, Rishabh Sanjay
DISTRESS DETECTION OF ROAD IMAGES USING DEEP LEARNING
1 online resource (83 pages) : PDF
University of North Carolina at Charlotte
The NCDOT road survey involves capturing images of highway and later processing to determine an overall pavement performance. Terabytes of data is captured even over a small highway section, therefore, there is a need to find better solutions to process these images. This study aims to label and detect distresses on the collected images using deep learning. Previous studies have manually labeled images to develop a labeled dataset, however, there is no documented record of this process. Therefore, this research enumerates the steps required to develop a labeled dataset from the raw images and utilize these images to develop several Mask-RCNN models. In addition, previous studies have utilized deep learning models to classify road survey images, however, this study utilizes Mask-RCNN to not only detect the different distresses present on an image but also to classify its type. This study identified Adobe Photoshop as the ideal labelling software for survey images. Mask-RCNN models were developed using python with Tensorflow, Keras and opencv2 libraries. Several parameters were changed and a total of five models were tested on training and test road survey images. The results were promising, and further studies need to develop more accurate models for conclusive results. The findings of this study can be applied by other transportation agencies on other road survey images.
Construction & Facilities Mgmt
Smithwick, JakeSherlock, Barry
Thesis (M.S.)--University of North Carolina at Charlotte, 2019.
This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s). For additional information, see http://rightsstatements.org/page/InC/1.0/.
Copyright is held by the author unless otherwise indicated.