Enhancing visual-based bridge condition assessment for concrete crack evaluation using image processing techniques

Abstract

Condition assessment is one of the most essential practices in bridge asset management to maintain the safety and durability of structures. Routine bridge inspection, a visual-based method, is regularly performed by qualified inspectors to determine the condition of individual bridge elements manually using bridge inspection standards. However, the quality of a visual-based condition assessment relies heavily on the inspector's knowledge and experience. The research presented here focuses on the development of an enhanced method to minimise the shortcomings of visual-based inspection. In this paper, we investigate the performance of RBF-kernel support vector machines (SVMs), a supervised machine learning technique, to increase the reliability of visual-based bridge inspection. The results of this study can contribute to minimising the shortcomings of current visual-based bridge inspection practices.