Adaptive fuzzy c-means through support vector regression for segmentation of calcite deposits on concrete dam walls

Dams are very important economical and social structures that have a great impact on the population living in surrounding area. Dam surveillance is a complex process which involves data acquisition and analysis techniques, implying both measurements from sensors and transducers placed in the dam body and its surroundings, and also visual inspection. In order to enhance the visual inspection process of large concrete dams, we propose a computer vision technique that allows detection and quantification of calcite deposits on dam wall surface. These cal-cite deposits are a clear sign that water infiltrates within the dam body. Further, their intensity and extent could provide valuable information on severity degree of the infiltration. The proposed scheme for identification of calcite / non-calcite areas on the color image of dam wall consists classifying the pixels into three classes, using a modified fuzzy c-means algorithm, which assigns an error penalty factor to membership degree, based on the distance between the classes' centroids and histogram skew. The weight for the calcite class is determined using support vector regression, in order to obtain a numerical mapping for calcite class's weight and histogram skewness.