Research On Sintering Condition Of Rotary Kiln Based On Dynamic Image Video Features

Rotary kilns are widely used in chemical industry,metallurgy,cement and other important chemical industries.In these industries,improving energy efficiency and reducing exhaust emissions are of great significance for maintaining the sustainable development of the national economy.Therefore,monitoring the sintering process of the rotary kiln,identifying the combustion conditions,and maintaining the normal and stable combustion state are essential for effective production,reducing energy consumption and exhaust emissions.At present,most methods for automatically identifying combustion conditions based on computer vision are to extract static features of images from single-frame images for analysis.There are some problems with the method based on static image analysis: On the one hand,due to the serious interference of dust and smoke from industrial sites,the collected images are blurred and difficult to segment,resulting in poor robustness of the features extracted from the region of interest;On the other hand,the static feature only considers the statistical characteristics of a single frame of flame image,and ignores the flashing dynamic information of the coal-fired flame,so the recognition accuracy of the combustion conditions,especially the changing conditions,is not high.In view of the above problems,paper takes the flame video of the sintering process of the rotary kiln as the research object,and studies a method of combustion condition recognition based on the spatiotemporal characteristics of the flame video.The main work of the paper is as follows:(1)In order to quickly extract the dynamic texture information of the video,a new 3D texture feature descriptor(Three Dimensional Brief Local binary pattern,3DBLBP)was constructed to characterize the texture features of the flame video.3DBLBP does not need to segment images to extract 3D texture features of flame.It can accurately and effectively describe dynamic texture features of video sequences with simple implementation and strong robustness.(2)The dynamic characteristics of flame video under different combustion conditions are theoretically analyzed.For the first time,it is proposed to mark the transition period from normal working conditions to abnormal working conditions as unstable working conditions.According to the specific characteristics of this working condition,it is proposed to use the phase information in the video image signal todescribe its morphological structure information.Based on the principle of phase congurency,a new three-dimensional structure descriptor(the Histogram of Phase Congruency from Three Orthogonal Planes,HOPC-TOP)was constructed to represent the dynamic structure information of flame video,which can better achieve the robust identification of unstable conditions.(3)Based on 3DBLBP and HOPC-TOP,a method for detecting the burning condition of rotary kiln based on the three-dimensional time-space feature extracted from the flame video was designed.This method captures the spatiotemporal shape and structure information of the flame image sequence and realizes the recognition of different burning conditions.The 3D features extracted by 3DBLBP and HOPC-TOP are combined to form the spatiotemporal features of the rotary kiln flame video,which can well characterize the spatiotemporal shape and texture of the flame video sequence.Combined with SVM,the recognition accuracy of 96.76% of normal conditions,92.63% of under-chilled conditions,and 86.72% of unstable conditions was achieved,which verified the effectiveness of the method proposed in the paper for the identification of sintering working conditions of rotary kiln.