Image Segmentation Algorithms for Industrial Process Tomography

Dominik Sankowski1, Andrzej Plaskowski2, Volodymyr Mosorov1, Krzysztof Strzecha1, Slawomir Jezewski1

1Computer Engineering Department, Technical University of Lodz, al. Politechniki 11, 90-924 Lodz, Poland

dsan@kis.p.lodz.pl, mosorow@kis.p.lodz.pl, strzecha@kis.p.lodz.pl, frank@kis.p.lodz.pl

2Industrial Institute of Organic Chemistry in Warsaw, ul. Annopol 6, 03-236 Warszawa, Poland tomograf@astercity.net

ABSTRACT

Tomography images provide useful information for monitoring and control of the formation and combustion of dust - air clouds during an experiment in a steel non-transparent explosion vessel. The geometry of the flame-front propagation of the combustion process of the dust cloud and its intensity depend on the explosion state and can be used for control and monitoring purposes. Image segmentation of the flame-front propagation process is difficult since the background intensity is sometimes higher than the flame intensity, therefore the use of sophisticated image analysis algorithms is required. This paper describes methods of analysing tomography images for industrial process and characterising the combustion geometry. A segmentation algorithm is proposed to separate the flame from the background using the image segmentation method. Segmentation algorithms (e.g., active contours) usually assume solid objects with sharp boundaries. This is not true in the case of combustion images. The flame is nonhomogeneous and has a fuzzy boundary. To circumvent this difficulty multiple contours are used to characterise the flame geometry. The flame shape is then obtained by robust estimation methods, using a model of the image formation process inside the combustion chamber.

Keywords image processing, image segmentation, combustion tomography