DT-CWT Fusion Detection Method Based On Sparse De-noising For Flame Image

In the sintering production technology, the rotary kiln firing zone temperature is a key technology control indicator during the production process. The traditional method of controlling the kiln firing temperature is liable to the working conditions. To overcome these problems, using the rotary kiln flame image processing to detect the temperature information of the firing zone. The black handle region in flame images carrying very important feature information that can assist to judge the sintered statement when the flame area and material area can not accurately judge. However, due to the flame image’s characteristics and the effect of the interference signals, leaded to the extraction of the black handle regional features is difficult. During the Rotary kiln flame image edge detection, mainly aimed at single image segmentation and processing, some researchers to extract areas of single image then fusion them. But these methods are ineffective for the single image which black handle edge is not complete, so this paper uses the pixel level image fusion method to fuse two single-frame images, can detect integrated black handle region.At present the wavelet transform(DWT) theory is commonly used in image fusion at pixel level method. DWT have some limitations that cannot meet the processing requirements about noise interference serious industrial image The dual-tree complex wavelet transform(DT-CWT) can overcome the deficiency of the DWT. DT-CWT fusion method can get better result than DWT fusion method. In order to suppress the irrelevant noise information should be combined with certain de-noising method to enhance the noise immunity of DT-CWT fusion method. Traditional spatial filtering methods always caused the image edge information blurred that cannot be performance well for image characteristics. Image sparse representation based on complete dictionary satisfy the sparsity, characteristics maintain, divisibility, which can get better performance of the image characteristics, becomes a good method for image de-noising, therefore this paper will use sparse de-noising in rotary kiln flame image processing. This paper based on the characteristics of coal-fired flame image, comprehensive sparse de-noising and DT-CWT fusion method, proposed a DT-CWT fusion detection method based on sparse de-noising for flame image, in order to solve the problem of black handle region’s effective fusion and edge detected for the rotary kiln flame image under noise interference serious condition. The concrete content as follows:Firstly, study the DT-CWT fusion method for flame image The basic idea is:Two single frame flame image do two level DT-CWT decomposition process, the first low-frequency sub-band used for the second decomposition. The high-frequency part of two original image after decomposition use the most supreme absolute value fusion rules to obtain the high-frequency coefficients of fusion image. The low-frequency part of two original image after decomposition use weighted average fusion rules to obtain the low-frequency coefficients of fusion image. Every frequency fusion coefficient do the reverse DT-CWT obtain fusion image from reconstruction.Secondly, during the processing of flame image fusion, the high-frequency sub-band and low-frequency sub-band on the first decomposition level is de-noising by sparse representation. The noise immunity of the fusion method improved effectively.Finally, the simulation used four groups flame image which collected from the rotary kiln industrial site. It was proved that the DT-CWT fusion detection method based on sparse de-noising for flame image is effective by compare the effect parameters and edge image of different fusion detection method. Experimental results show that this method could be applied on the actual situation.