Illumination adjustment for bridge coating images using BEMD-Morphology Approach (BMA)

Abstract

Image recognition methods have been applied to infrastructure surface coating assessment for more than a decade. However, non-uniform illumination, such as shades and highlights, is still a big challenge to date. To tackle the non-uniform illumination problem in infrastructure surface coating assessment, the BEMD-Morphology Approach (BMA) is proposed in this paper. Basically, non-uniform illumination problems can be classified into two categories: shades and highlights. The proposed BEMD-Morphology Approach (BMA) adopts the Bidimensional Empirical Mode Decomposition (BEMD) to mitigate the effect of shades and uses morphological processing to detect highlight spots and replace them with non-highlight neighboring pixels. With BEMD and morphological processing, the non-uniform illumination problem in infrastructure surface coating assessment (e.g., steel bridge coating assessment) could be better handled and more accurate assessment results could be generated. At last, the performance of BMA is evaluated using the K-Means algorithm, one of the most popular image recognition methods, to show the effectiveness of BMA.