Condition Recognition Of Coal Flame Images Based On Phase Consistency

Rotary kiln is a kind of typical complex industrial controlled object,and the recognition of sintering condition in kiln is the key to stable control.Because of the particularity of kiln body structure and the on-site dust environment,the sintering temperature cannot be directly measured physically.The coal flame image contains abundant information of sintering conditions,but the poor kiln environment results in uneven illumination of the flame image and blurring of edges.Image segmentation and feature extraction are difficult.On the other hand,image phase information has stronger robustness than gray information,and the phase congruency is also consistent with the human visual mechanism.Based on the above two points,this paper takes the coal flame image in the alumina production process as the specific object,and uses the phase congruency principle to divide the pulverized coal zone to identify the temperature in the kiln,and solves the problem of short service life of equipment,poor product quality,low yield,high energy consumption caused by the current"artificial observation of fire".It is specifically divided into the following aspects:(1)The basic theory of phase consistency and its application in recent years are reviewed.The phase-consistent edge detection and the traditional edge detection algorithm are compared by simulation.The experimental results show that the phase-consistent method has stronger gray-scale invariant characteristics than the spatial gray-scale method;(2)Aiming at the problem of illumination unevenness in coal flame image,an improved light compensation algorithm for Contrast Limited Histogram Equalization(CLAHE)is proposed.The image processed by CLAHE wavelet high-frequency denoising.Then for the dust concentration in the kiln is large,the image edge is blurred,an adaptive threshold fuzzy enhancement algorithm is applied to the demised image to further enhance the enhancement effect of the coal flame image;(3)Proposing a fusion algorithm for flame image coal blending area(commonly known as "black handle")that combines phase consistency and Otsu method.Firstly,edge detection is performed with phase-consistent method.Binary fusion followed by Otsu method is used to segment the complete foreground.This method combines Otsu's method's advantages of center segmentation in the region of interest and edge-segmentation advantage in the same phase to effectively segment the coal powder region;(4)Extracting the morphological characteristics of the five rectangles in the pulverized coal area,and using the neural network to classify the working conditions shows that the method can better identify the sintering conditions in the rotary kiln.It provides an effective condition identification method for the testing of coal-fired process conditions.