Multiple layer separation and visualisation of inductively scanned images of reinforcing bars in concrete using a polynomial-based separation algorithm

Inductive scanning systems that exploit eddy current effects for imaging steel reinforcing bar mesh within concrete have been developed and reported by the authors in several publications. Images generated in this manner depict the different horizontal and vertical layers of the mesh within a single, 2D plane. Deeper bars appear as much fainter structures than those closer to the surface for two reasons: first, the signals they generate are weaker, and second, the image is linearly normalised with respect to the much stronger signals returned from the upper bars. This makes depth and dimensional analysis of deeper bars a severe problem. Below we describe a new suite of image processing algorithms that enables the original composite image to be visualised as separate, multiple images representing the various bar layers. This technique is termed polynomial-based layer separation (PBLS). The method also makes it possible to perform brightness-compensation of the lower bars and is a precursor to analysis that allows measurement of the diameters, orientations and depths of the bars. This information is critical for civil structures inspection teams. Knowing the spatial location of the bar peak widths in one layer, curve fitting is applied to calculate the baselines of the bars in other layers of the image. For a two-layer image, the peaks in the lower layer are removed leaving an image of the top bars, and vice versa for the bottom bars. Images of steel bar mesh processed using this PBLS system offer significant enhancements to the qualitative and quantitative properties of the original image data, and in tests described below, is significantly more robust than comparable methods of image segmentation.