boztoprak, halime akdeniz üniversitesi akseki myo bilgisayar programcılığı - akseki myo - bilgisayar programcılığı, Turkey

SEGMENTATION OF FLOC AND FILAMENTS USING CELLULAR NEURAL NETWORKS AND WAVELET TRANSFORM

35853

Examination of morphological characteristics of flocs and filaments plays an important role for activated sludge.An examination should be conducted especially on textural features and free filament amounts. This is why the segmentation stage has a particular importance and there are many studies in this regard. In this study, cellular neural networks (CNN) were used. A constant template was used and it was only the iteration value that was updated according to image. Wavelet method was employed to determine the iteration value. Second level decomposition was made with Haar wavelet filter. Iteration value was calculated with spatial frequency values of subbands acquired from decomposition process. The free filament amount in sludge is substantial in terms of activated sludge features. Filaments and flocs in images may appear one within the other. Therefore, filaments that are present free or in contact with flocs should be degraded from the image. Hence, a series of morphological processes were applied on the image after CNN process. A comparison was made between the image acquired from CNN segmentation process and images acquired from edge extraction and top-hat transform. Consequently, flocs and filaments were segmented parallel to features of the image and aim of the study.

38

Segmentation , Wavelet Transform , Floc , Filament , Cellular Neural Networks , Morphological processing

Sdu International Technologic Science

49