Research And Development Of Image Detection System For Prebaked Anode Carbon Block

In recent years,the world aluminium industry has developed rapidly,and the capacity of domestic electrolytic aluminium industry has increased rapidly.In the production process of primary aluminium,a complete industrial chain which consists of four links,including aluminium ore mining,alumina production,electrolytic aluminium smelting and aluminium processing and production.Among them,in the process of aluminium electrolysis,it is necessary to use petroleum coke and pitch coke as aggregate,coal tar pitch as binder,and carbon blocks with stable geometry structure as pre-baked aluminium cell anode,also known as baked anode.During the production of pre-baked anode carbon blocks,due to the structural properties of raw materials,composition ratio,heat treatment process,heating and cooling,cracks will occur on the surface of carbon blocks.Cracks in carbon blocks can adversely affect the yield of carbon anodes and even the production efficiency of aluminium industry.Cracks can lead to local non-conductivity or uneven current distribution in the process of carbon block electrolysis,thus affecting the operation of the electrolytic cell,resulting in lower electrolytic efficiency.If the crack is found in the early stage,the carbon block can be replaced in time.Traditionally,the surface crack detection of carbon blocks is carried out by using the naked eye of the quality inspector.This method has lower recognition efficiency,higher labor cost and higher false detection rate.This paper presents a method of surface crack identification of pre-baked anode carbon blocks based on image processing,which can extract the geometric parameters of cracks,and according to these parameters and thresholds to distinguish,so as to judge whether the carbon blocks are qualified products or not.Firstly,this paper introduces the traditional crack recognition method based on image processing.Aiming at the occurrence of edge blurring which caused by the traditional median filtering method,a direction-based median filtering algorithm is proposed.In the process of traditional Gaussian filtering,sigma parameters can not be adaptively set,thus resulting in poor image smoothing effect.In this paper,an improved Gaussian filter is proposed in the century,which can adjust the scale adaptively.In view of the poor segmentation accuracy caused by manual experience when using traditional Canny operator for image segmentation,this paper uses the method of maximum inter-class variance to dynamically obtain high thresholds,and calculates the value of low thresholds according to the multiplier relationship between high and low thresholds.In addition,the algorithm of calculating the magnitude and direction of image pixel gradient is improved in Canny operator.The original 2*2 template is replaced by 3*3 template,which adds the influence of the pixel factor of the diagonal direction of the central pixel point.It can achieve good noise suppression effect and locate the crack of carbon block more accurately.In this paper,an image processing-based crack identification system for anode carbon blocks is designed.The core of the image processing module is located in the background system of the host computer.In addition,the lower computer system is responsible for obtaining the image information of carbon block and transmitting it to the upper computer system.Android users can access the processed image information in the background system of the host computer through APP to improve the real-time and reliability of the carbon crack detection system.