"Flocculation and Erosion analysis using Image Processing and EPS analysis"There is a rising interest in how sediment particles behave, due to the importance to accurately address sediment transport models. Understanding how biologically changed flocs impact further transport; helps not only to calculate mass balances of particles in a watershed, but also to predict how associated contaminants might be moving. For evaluating floc behaviour, sediments with biofilm are eroded in a microcosm and the entrained flocs are transferred to a sedimentation column for recording. The main objective of this study is then to automatize the image processing by creating a MATLAB program named FloculamMazza, to enable the calculation of characteristics such as floc velocity, size, area and acceleration that are important for further transport and deposition of resuspended sediments.
Applying FloculaMazza the separation of foreground from background is vital in order to recognize flocs as independent objects and has been done by one of the thresholding methods, the adaptive method (subtraction of a background picture) to give a more accurate representation on the present flocs, and their features. One of the challenges for elaborating FloculaMazza is then to track the movements of flocs within the pictures, despite the fact that flocs show unpredictable behaviour such as further aggregation, occlusion and/or rotation during the recording that might hamper floc identification. In the trade-off between the reliability of results and the wish to track as much of the present flocs as possible, it has been decided to rather limit the huge floc population by a search window (in which flocs are compared using the area and CSF) to secure a proper recording of the remaining ones.
The validation of FloculaMazza has been performed by analysing results of eroded sediment flocs with a biofilm growth history. By analysing 7270 photos, 309258 flocs could be followed altogether. Horizontal and vertical displacements have been evaluated, giving a closer look at those flocs that perform inversely to the “normal course” of the group. It has been noticed that the flocs with “extreme” horizontal velocities were tracked correctly, concluding that the values correctly calculated for the corresponding floc. In the case of the vertical velocity it is identified that values out of the expected range are due to miscalculations by FloculaMazza, for which possible enhancements have been proposed. These recommendations include additionally considering also a minimum floc displacement for the flocs tracking, improving the thresholding by eliminating those flocs that are out of focus, processing the pictures backwards as well as forwards to find the best fitting floc for tracking, and only considered flocs that are tracked in at least three pictures.