Proceedings of COBEM 2005 18th International Congress of Mechanical Engineering Copyright © 2005 by ABCM November 6-11, 2005, Ouro Preto, MG AUTOMATIC MONITORING OF POLYMERS FLUX USING IMAGE PROCESSING André Borin Soares LAPSI – DELET – UFRGS, Av. Osvaldo Aranha 103, 206B borin@eletro.ufrgs.br Juan Carlos Sandoval Ortiz LABMECFRAC / GPFAI – DEMEC – UFRGS, Rua Sarmento Leite, 425 juanca@mecanica.ufrgs.br Thiago Rosa Figueiró LAPSI – DELET – UFRGS, Av. Osvaldo Aranha 103, 206B figueiro@eletro.ufrgs.br Leticia Vieira Guimarães LAPSI – DELET – UFRGS, Av. Osvaldo Aranha 103, 206B leticia@eletro.ufrgs.br Altamiro Amadeu Susin LAPSI – DELET – UFRGS, Av. Osvaldo Aranha 103, 206B susin@eletro.ufrgs.br Abstract. A polymer drying system requires constant monitoring due to the possible system fails, specially related to the material particles flux when they leave the conveyor belt to pass throw the funnel and arrive at the fuse. The problem consists on the natural tendency of the particles to aggregate to the fuse, obstructing the conduction duct. Nowadays, the monitoring is made by an operator that constantly observes the images deriving from a camera installed at the top of the funnel, which acquire images of transportation fuse. The problems of this method are the tiredness of the operators and the many attributions they have, observing many factory spots, which makes it difficult to quickly response at the beginning of the duct obstruction. The proposed system uses the camera signal already installed at the top of the funnel to make image processing in order to monitor the flux. In the image acquired is selected a Region of Interest (R.O.I.) and on it the system verify the variation of luminance due to the polymer pass. If the variation became smaller than a threshold and the quantity of polymer on the R.O.I. became greater then a second threshold, then is defined and obstruction. The system performs an automatic estimate the quantity of polymer that may characterize a obstruction and then generate an alarm sound to an operator fast enough to be able to control the flux. Keywords: Flux monitoring, automatic supervision, industrial automation, image processing and motion detection. 1. Introduction Generally, industrial process control needs monitoring several sensors in order to take decisions to control several points of the process. Lately, image sensors have been used for surveillance with manual monitoring and/or manual alarm. Also, in automatic security systems applying image processing as face, finger print or car plate recognition and identification, for example. Generally, industrial environment is very noisy for usual sensors as temperature, pressure that generate low energy signals that are very sensitive to noise. The usages of image sensors in industry have been encouraged by the improvement of their technology, that permit to use cameras in noisy environment as so as the decreasing of their size and cost allow to spread their application even in small places. Moreover, the image processing tasks already used in security systems can be adapted to industrial application successfully. This work proposes an image processing technique to monitoring a material transport line of a polymers industry. Polymers particles are heated and pressed in order to take de water of dry the particle, after the washing process. The transport line is located between the heating and pressing process. The critical points of the material transport line to be monitored are the funnel that conducts the polymer particles from the heating location to the pressing location. The pressing is performed by a fuse that rotates inside of a pipe while the particles are located between the fuse and the pipe wall, so the particles ate pressed and transported to the next point of the process, at the same time. The pipe is often obstructing by the increasing of the number of particles transported by the line. Therefore, the flux of polymers must be controlled. Nowadays, the flux of polymers is controlled manually. A technician monitors the transport line and the funnel state by images transmitted to the control section of the industry. The manual monitoring is not a trustworthy method because