Maceral Classification Of Exinite In Coal Based On Contourlet Transform

With the increasingly shortage of high-quality coal resources and the serious of environmental pollution caused by coal processing,the clean and efficient utilization of coal resources becomes a hot topic of widespread concern.According to its original stroma,forming conditions,microscopic features and properties,macerals of coal are divided into groups of vitrinite,inertinite and exinite,and each group contains some macerals and submacerals.The constitution of macerals has an important effect on properties of coal,such as cokeability,cohesive property,thermal fragmentation performance and adsorptive of CO.Therefore,classification and identification of coal macerals are of significance for evaluating coal performance and guiding the coal blending.According to the difference of grayscale distribution,texture feature,contour and its direction of macerals of exinite group,grayscale statistical features based on the Contourlet Transform are extracted,and texture features based on Tamura and Haralick methods are extracted as complements,and different classification schemes are designed to implement the classification of exinite macerals.The main contributions of this dissertation is as follows:(1)On the basis of analyzing the image features of coal macerals and consulting a large number of relevant literatures of domestic and foreign,the image characteristics among different exinite macerals are emphatically analyzed.(2)According to directivity characteristic of texture and area contour of macerals in the exinite group,microscopic images of macerals are decomposed with the Contourlet Transform,grayscale statistical features based on different sub-band are extracted,and gray-scale distribution feature,Tamura texture,Haralick texture are also extracted,the distinguishability of these features to maceral classification is analyzed.(3)Based on the theory of support vector machine(SVM),two classification schemes,"Voting" method and "Decision Tree" method,are constructed,and the effective feature sets are selected respectively for training classification to implement the classification of 7 macerals in exinite group.(4)For contrast,according to the theory of extreme learning machine(ELM),two classification schemes,named direct method and decision tree method,are also constructed,and 7 of macerals in the exinite group are classified.(5)All algorithms of feature extraction and classification are implemented on the platforms Visual 2013 and Matlab software by programming,and results from different methods are comparied analyzed.The innovations and specialties of this dissertation is as follows: macerals of exinite are classified by means of image analysis,which achieve the purpose of auto-classification with machine;information of direction is introduced into the feature set by Contourlet Transform,which brings about a good result of classification on this topic.