Online Analysis of Coal Ash Content on a Moving Conveyor Belt by Machine Vision. Issue 2 (4th March 2017)

• Record Type: Journal Article
• Authors: Zhang, Zelin; Yang, Jianguo
• Is Part Of: International journal of coal preparation and utilization. Volume 37:Issue 2(2017)
• Page Start: 100
• Page End: 111
• Publication Date: 2017-03-04
• Subjects: Machine vision -- coal ash content -- RBF -- LIB-SVM -- MLR; Coal preparation -- Periodicals; Coal -- Research -- Periodicals; 662.62
• DOI: 10.1080/19392699.2016.1140650
• Languages: English
• ISSNs: 1939-2699

Abstract

ABSTRACT: An online machine vision system to predict the ash content of a fine cleaned-coal product on a moving conveyor belt was established and tested. Observing the changes of different coal images and their corresponding ash content, six statistical features from a gray histogram and one feature reflecting the light-spot ratio were extracted. Three methods of different types— radial basis function (RBF) neural network, support vector machine (LIB-SVM), and multiple linear regressions (MLR)—were carried out to predict the ash content and to compare their prediction accuracies. RBF and LIB-SVM also considered the performance of original values and normalized values. Through comparing the fitting effect and generalization ability by RMSE and R 2, RBF had the best fitting effect, but its generalization ability was the worst. The fitting effect and generalization ability of models by LIB-SVM and MLR were all good, while the normalized LIB-SVM model is the best one of the above three methods.