基于CNN卷积神经网络的煤矸石自动分选研究  

Study onautomatic separation of coal gangue based on CNN convolutional neural network

作　　者：王莉[1] 于国防[2] 沈慧宇 田波 WANG Li;YU Guofang;SHEN HuiYu;TIAN Bo(School ofInformation and Electronics Engineering,Jiangsu Vocational Institute of Architectural Technology,Xuzhou,Jiangsu 221116,China;School of Electronic and Information Engineering,China University of Mining and Technology,Xuzhou,Jiangsu 221006,China;Department of Information Management,Chienkuo Technology University,Taiwan,China;Shanxi Lanxian Coking Coal Company)

机构地区：[1]江苏建筑学院信电工程学院,江苏徐州221006 [2]中国矿业大学(徐州)电子信息工程学院,江苏徐州221006 [3]台湾建国科技大学资讯与网路通讯系,中国台湾500 [4]山西省岚县社科乡下会村昌恒煤焦有限公司,033599

出　　处：《江苏建筑职业技术学院学报》2019年第4期35-39,共5页Journal Of Jiangsu Vocational Institute of Architectural Technology

摘　　要：采用山西省的焦煤和肥煤作为研究对象,针对目前利用煤矸石灰度信息作为判断二者依据的局限性问题,提出了一种基于CNN卷积神经网络的煤矸石自动分选系统.该系统利用构建的卷积神经网络通过对煤块和矸石图像纹理特征的多层次提取进行结果分类输出.测试结果表明,该方法不受样本数据色差的影响,可以成功的识别检测出煤块和矸石,准确率达到92%.Taking coking coal and fat coal of Shanxi Province as the research object,aiming at the limitation of using the information of coal gangue lime degree as the basis of judging the two,an automatic separation system of coal gangue based on CNN convolution neural network is proposed.The system uses the convolution neural network to extract the texture features of the coal and gangue image and output the classification results.The test results show that the method is not affected by thecolor difference of sample data,and can successfully identify and detect coal and gangue,with an accuracy of 92%.

关 键 词：卷积神经网络 深度学习 煤矸石分选 灰度信息 

分 类 号：TP394[自动化与计算机技术—计算机应用技术] TD948[自动化与计算机技术—计算机科学与技术]