A Pavement Disease Detection Method based on the Improved Mask R-CNN
Authors:
- 1 November 2020
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 619-623
Abstract
Accurate detection and location of pavement disease has attracted more and more research interest. Existing approaches depend heavily on manual works and expert experience, which are time-consuming and cost-effective. The paper presents an automated pavement damage detection and location method by adopting Mask R-CNN algorithm and active contour model. Mask R-CNN algorithm selects ResNet as the backbone network because of its strong feature extraction ability, and generates candidate target regions through 12 different anchors to further improve the accuracy of detection. Then, it is sent to the sub-network for classification and positioning. And contour of the diseased region is generated by Mask R-CNN. The region of the disease can be a preparatory knowledge for later road maintenance. But Mask R-CNN leads to the missing area due to inherent defects, the paper adopts active contour model to address the problem. The initial curve of the active contour model comes from the result of Mask R-CNN model and evolves under the action of internal energy and external energy. When the energy functional reaches the minimum, the curve converges to the edge of the object. The experimental result for pavement disease image shows the proposed method has desirable performance.Keywords
Funding Information
- SDUST (BJRC20180501)
- National Natural Science Foundation of China (61976125)
This publication has 17 references indexed in Scilit:
- Asphalt surfaced pavement cracks detection based on histograms of oriented gradients2015
- Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation2014
- Selective Search for Object RecognitionInternational Journal of Computer Vision, 2013
- Stochastic Gradient Descent Tricks2012
- Automatic Pavement Crack Detection Using Texture and Shape DescriptorsIETE Technical Review, 2010
- LabelMe: A Database and Web-Based Tool for Image AnnotationInternational Journal of Computer Vision, 2007
- Fast Global Minimization of the Active Contour/Snake ModelJournal of Mathematical Imaging and Vision, 2007
- A geometric snake model for segmentation of medical imageryIEEE Transactions on Medical Imaging, 1997
- On active contour models and balloonsCVGIP: Image Understanding, 1991
- Snakes: Active contour modelsInternational Journal of Computer Vision, 1988