Analysis Of Thermal Coal Property Based On Image Analysis

The domestic coal washing and processing industry is currently at a low level of automation and intelligence.The detection of coal quality also depends on the fast ash fast floating experiment.This experiment measures the coal material information feedback period is long,and the result is not time-sensitive.It is impossible to monitor the coal quality in real time,which seriously limits the efficiency of washing and processing.In the early stage,the research group conducted a certain study on the analysis of coal composition of Taixi anthracite coal based on machine vision.Based on the previous research,this paper uses the thermal coal as the experimental coal sample and improves the image analysis algorithm for the thermal coal quality analysis.The paper preprocesses the image through multiple image techniques,including histogram equalization to enhance the contrast of the image,and compares the filtering effect of median filtering and mean filtering.The median filtering has better denoising effect on the image.The coal particle region in the image is segmented and extracted using the improved watershed algorithm.The paper improved the particle size composition analysis method of coal and improved the calculation speed.Five shape parameters of coal particles were extracted,and the correct rate of coal particle size was compared with each parameter,and the optimal particle size characterization parameters were selected.Observing the shape characteristics of the coal particles,the shape of the coal particles is an irregular spherical shape.According to the formula corresponding to the volume V-area S of the sphere and the cube,the coal particle area S is calculated by the number of pixels of the coal particle region in the image.40 sets of V-S data were selected and the V-S model of the coal particles was determined by polynomial fitting.The experiment proves that the model is suitable for the volume prediction of coal particles,which effectively improves the running speed of the algorithm.According to the color and texture features of the image,43 feature parameters are selected,and the feature data is normalized.The ant colony algorithm is used to filter the feature parameters,and the linear correlation parameters are eliminated to reduce the feature dimension.Finally,BP neural network is used to predict the density level of coal particles.Compared with the previous research of the research group,this method effectively improves the prediction accuracy of the density of small-grain coal particles(3-6mm),and the prediction effect is better.