Study Of Furnace Flame Image Enhancement And Edge Detection Algorithm

With the persistently increasing capacity of thermal power generating units, its furnace structure and the complexity of environment also is continuously strengthening, the system is put forward higher requirements for monitoring the furnace flame real-timely and efficiently. In order to improve the coal powders burnt steadily and reduce the emission of pollutants, what's more, the furnace unsafe accidents will be effectively prevented. In this paper, we study the furnace image of thermal power plant firing coal. To improve the effectiveness and speed of the furnace flame image processing algorithms, and on the basis of the research results about furnace flame image processing algorithm, the de-noising,enhancement, and edge processing technology to the furnace flame image are studied in this paper.First, in order to obtain the information of furnace flame image, and improve arithmetic speed on de-noising algorithms, an furnace flame image de-noising method based on the Dual-tree Complex Wavelet Transform(DTCWT)and the Hidden Markov Tree(HMT) model and is proposed.In the algorithm, the HMT model parameters are making an estimate by EM algorithms. Because the Dual-tree Complex Wavelet Transform has shift stability and excellent directional selectivity, the dual-tree complex wavelet coefficients are respectively modeled according to the HMT model, which includes the advantages of model precision. This article de-noising algorithms result is obvious. Noise ratio is drastically improved by de-noising algorithms. Matlab simulation results show that the information of furnace flame image are obtained,and arithmetic speed is improved by de-noising algorithm, and because it helps to reduce the noise more effectively.Secondly, to effectively improve the brightness of furnace flame image and furnace background contrast ratio, an adaptive multi-scale Retinex furnace flame image enhancement method is proposed. The algorithm adopts the guided filter to smoothing filter a furnace flame image, then the flame illumination image is obtain. In time domain flame image reflection components are gained by using subtraction, and reflection components make color correction, which avoids the furnace flame color distortion. The error probability and the minimum threshold is solved by using iterative calculation update threshold, then it realizes the adaptive optimal threshold segmentation. Finally, That the furnace flame image can arise some possible problems, such as overall dim and local details, will be eliminated. Matlab simulation results show that adaptive multi-scale Retinex furnace flame image enhancement method can efficiently strengthen the furnace flame images, and the brightness of furnace flame image and furnace background contrast ratio is improved.Finally, in order to effectively extract the edge information of furnace flame image, and overcome poor adaptability defects to the traditional Canny edge detection algorithm. In this paper, based on the research of the traditional Canny operator, an gray image segmentation algorithm based on analysis of histogram concavity is presented. The image segmentation is the key to select threshold. The algorithm will analysis of histogram concavity apply to the double threshold Th or Ti. Matlab simulation results show that the algorithm can effectively extract the edge information of furnace flame image. compared with the traditional Canny edge detection algorithm, the algorithm can present the edge information of furnace flame image, and furnace flame image can effectively be recovered.