This article describes a fully automatic image analysis procedure for fast and reliable characterization of the activated sludge composition, that is, the floc and filament features. The algorithms developed for each of the analysis steps, that is, segmentation, object recognition, and characterization, are described in detail. Although the application range of the recognition method is a priori expanded by introducing a number of control parameters, the procedure proves to be intrinsically robust as it produces satisfactory results for a fixed set of parameter values for a wide variety of image types.
