Research And Design Of Monitor System For Coke State In Power Plant Boiler

It is a well known problem to monitor accurately and in time if there is some coke at the bottom of a boiler in a power plant .The coke is very harmful .It may lead to the decline of output power and heating power efficiency. Even it will lead to such accidents as explosion which are very dangerous and expensive .The digital image processing and pattern recognition technology are feasible to overcome the coke monitor problem.In this thesis a monitor system for coke states is designed .The system is composed of two parts .One is image processing which is main and the other is image recognition .At the part of image processing in this paper ,after getting the coke image accurately and in time ,the coke image is processed by a series of steps .First the image is grayed and then smoothed .Because the high temperature coke is very similar to the background , it is very difficult to segment by a threshold .Edge detection must be used .In the paper several usual used operators are compared and analyzed .After segmenting the image by a threshold,open operation is used to smooth image.Lastly,in order to smooth the region outside the coke box ,at the same time the information inside the box isn't changed ,here a means named model match is introduced .That is a model image is operated with a coke image to recognized which has been open operated.In this thesis the coke image is processed by using Wavelet Transform .The processing result is compared with the result from LoG-Laplacian operator.At the part of image recognition , effective characteristics are chosen to design classifier . Lots of samples are used to test the system ,The result show that the recognition ratio of this system is high and the speed of recognition is fast.