Publications / 2011 Proceedings of the 28th ISARC, Seoul, Korea

A Rust Intensity Recognition Approach for Steel Bridge Surface Coating Images Based on Artificial Neural Networks

Po-Han Chen, Heng-Kuang Shen, Luh-Maan Chang
Pages 996-1001 (2011 Proceedings of the 28th ISARC, Seoul, Korea, ISBN 978-89-954572-4-5, ISSN 2413-5844)
Abstract:

One of the traditional methods to assess the quality of the coating on the steel bridge surface is calculating the ratio of rust to the total area of coated surface. In North America, digital image processing has been applied on quality assessment of steel bridge coating since late 1990s due to the difficulty of rust area quantification by human vision. Most of the previously developed systems segment images into two groups, rust and background, before assess the rust ratio. These systems are usually lack of the ability to recognize the intensities of rust. In order to provide a flexible assessment, one pioneer method called adaptive ellipse approach (AEA) was proposed in 2009 to classify the images into three groups, background, rust, and mild-rust. However, even the three-group system is still weak at describing gradually changed colors and thus unable to classify those colors automatically. In this paper, we segmented the image into several clusters based on the similarity of colors, and used the root-mean-square standard deviation (RMSSTD) to measure the similarity of each cluster. Artificial Neural Networks was applied to choose the best number of clusters based on the values of RMSSTD. This paper models a rust intensity recognition approach to choose the most suitable number of clusters for describing rust intensities, trying to overcome a hurdle that two- or three-group systems could not resolve.

Keywords: Coating Defect Recognition, Color Image Processing, K-Means, Artificial Neural Network, Root-mean-square Standard Deviation (RMSSTD)