Segmentation of Activated Sludge Filaments using Phase Contrast Images

Abstract

Segmentation algorithms play an important role in image processing and analysis. The identification of objects and process monitoring strongly depends on the accuracy of the segmentation algorithms. Waste water treatment plants are used to treat wastewater from municipal and industrial plants. Activated sludge process is used in wastewater treatment plants to biodegrade the organic constituents present in waste water. This biodegradation is done with the help of microorganisms and bacteria. There are two important types of microscopic organisms present in the activated sludge plants, named as flocs as filaments, which are visible under microscope. In this paper we study the microscopic images of wastewater using phase contrast microscopy. The images are acquired from wastewater sample using a microscope. The samples of wastewater are collected from domestic wastewater treatment plant aeration tank. Our main aim is to segment threadlike organisms knows as filaments. Several segmentation algorithms (such as edge based algorithm, k-means algorithm, texture based algorithm, and watershed algorithm) will be explored and their performance will be compared using gold approximations of the images. The performance of the algorithms are evaluated using different performance metrics, such as Rand Index, specificity, variation of information, and accuracy. We have found that edge based segmentation works well for phase contrast microscopic images of activated sludge wastewater.