A robust approach for automatic detection and segmentation of cracks in underground pipeline images

Cracks in underground pipeline images are indicative of the condition of buried infrastructures like sewers and water mains. This paper presents a three step method to identify and extract crack-like structures from pipe images whose contrast have been enhanced. The proposed method is based on mathematical morphology and curvature evaluation that detects crack-like patterns in a noisy environment. Careful observation reveals that the cracks resemble a tree-like geometry in most cases which can be a usable feature for registration between successive images of the same region taken from various depths in the thickness of the buried pipe (3D visualization). In this study, segmentation is performed with respect to a precise geometric model to define crack-like patterns. Cracks in pipe images can be defined as clearly visible patterns (darkest in the image), locally linear and branching in a piece-wise fashion. First, the cracks are enhanced by mathematical morphology with respect to their spatial properties. In order to differentiate cracks from analogous background patterns, cross-curvature evaluation followed by linear filtering is performed. We discuss its implementation on 225 pipe images taken from various cities in North America and statistically evaluate its accuracy and robustness with respect to varying pipe background color, crack geometries and background noise.