There has been increasing interest in the use of robotic machines, and several prototype robotic machines have been developed to automate the grit-blasting process for the purpose of steel bridge maintenance. To utilize such a robotic grit-blast machine effectively, the first consideration is automating the determination of the rust surface area to blast, on the basis of standards of practice, in a rapid manner. This study aims to propose a method to rapidly and accurately determine the rust surface area on steel bridges to blast, with consideration for the standards of practice, from images acquired via a blasting machine. The first step is to perform a color space conversion to transform the input image from a red/green/blue (RGB) color space to a hue/saturation/intensity (HSI) color space. The next step is to detect the presence of rust, using pixel-level classification via the C4.5 decision tree algorithm. Then it is necessary to confirm the blasting area by verifying whether the rust detection result satisfies the specified criteria on the basis of standards of practice. The proposed method was validated on 39 test images with various characteristics with respect to the degree of rusting and rust distribution type. The experimental results showed that the average accuracy rate of rust area classification was about 97.63%, and the success rate of the final decision of blasting area determination was 100.00% for 39 test examples. The whole processing time took an average of only 0.86 seconds per image. The preliminary results demonstrated that the proposed method not only determined whether rust was present in an image and the amount of rust but also indicated whether blasting was necessary, and, if necessary, it rapidly specified the rust surface area that should be blasted on the basis of standards of practice. The proposed method could be successfully incorporated into a robotic grit-blast machine.
Keywords: steel; Steels; steel; blasting; Distribution; Robotics; classification; test;
Full Access to Technical Paper
PDF version for $20.00
Other papers from ISARC