Int. J. of Oil, Gas and Coal Technology   »   2016 Vol.11, No.3

 

 

Title: Narrow density fraction prediction of coarse coal by image analysis and MIV-SVM

 

Authors: Zelin Zhang; Jianguo Yang

 

Addresses:
School of Resource and Environmental Engineering, Wuhan University of Science and Technology, Wuhan, Hubei, 430080, China
School of Chemical Engineering and Technology, China University of Mining and Technology, Xuzhou, Jiangsu, 221116, China

 

Abstract: Fast estimation of density distribution is important for the gravity coal preparation in coal processing field. A narrow density-fraction prediction method for coarse coal by image analysis and MIV-SVM was proposed for the future estimation of density distribution. A semi-automatic local-segmentation algorithm was proposed to identify the coal particle regions and 38 colour and texture features extracted. Mean impact value (MIV) is considered as one of the best indexes to evaluate features correlation in neural network or classification methods. Support vector machine (SVM) is state-of-the-art large margin classifiers which have recently gained popularity within visual pattern recognition. Results indicated MIV-SVM can effectively select surface features of coal particles and establish the prediction model of density fractions. Classification accuracy (CA) rises gradually with the increase of feature number and the trend is smaller and smaller. The group of 32 features with the CA 84.29% is more suitable to predict the density fraction of coarse coal. [Received: April 26, 2014; Accepted: December 28, 2014]

 

Keywords: density fractions; coarse coal; image segmentation; mean impact value; MIV; support vector machines; SVM; image analysis; colour features; texture features; feature extraction; surface features; coal particles; prediction modelling; classification accuracy.

 

DOI: 10.1504/IJOGCT.2016.074768

 

Int. J. of Oil, Gas and Coal Technology, 2016 Vol.11, No.3, pp.279 - 289

 

Available online: 17 Feb 2016

 

 

Editors Full text accessAccess for SubscribersPurchase this articleComment on this article