基于经典卷积神经网络的智能煤矸识别  

Intelligent coal gangue identification based on classical convolutional neural network

作　　者：张勇 杨鹏[1,3] 王亮 李飞[1,3] ZHANG Yong;YANG Peng;WANG Liang;LI Fei(State Key Laboratory of Mining Response and Disaster Prevention and Control in Deep Coal Mines,Huainan 232001,China;School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan 232001,China;School of Mechanical Engineering,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]深部煤矿采动响应与灾害防控国家重点实验室,安徽淮南232001 [2]安徽理工大学电气与信息工程学院,安徽淮南232001 [3]安徽理工大学机械工程学院,安徽淮南232001

出　　处：《邵阳学院学报（自然科学版）》2021年第1期34-44,共11页Journal of Shaoyang University：Natural Science Edition

基　　金：安徽省科技重大专项资助项目(18030901049)。

摘　　要：为了准确快速地识别原煤中的煤和矸石,基于机器视觉的方式,采取经典卷积神经网络模型对煤和矸石图像进行识别分类;利用在以实验室环境下采集的小批量煤和矸石图像数据,运用数据增强技术扩充数据集,在深度学习框架中搭建各种经典卷积神经网络模型,对采集的数据集进行训练、验证和测试,获得各经典网络的训练准确率和损失函数曲线,并结合训练过程中的训练和验证损失函数曲线,对训练情况进行可视化分析,最后以训练参数量、浮点运算次数和评估损失函数为评价指标,建立综合评价函数E对各经典网络进行评估。结果表明AlexNet,ResNet50和ResNet101网络的训练准确率均在90%以上,且训练状态为完美拟合,评估结果显示,AlexNetj是最适合于煤矸分选的经典卷积神经网络。In order to accurately and quickly identify coal and gangue in raw coal,the classical convolutional neural network model was adopted to identify and classify coal and gangue images based on machine vision.Using the small batch of coal and gangue image data collected in the laboratory environment,the data set was expanded by using data enhancement technology,and various classic convolution neural network models were built in the deep learning framework.The collected data sets were trained,verified and tested,and the training accuracy and loss function trace of each classic network were obtained,combined with the training and verification loss function trace in the training process.Finally,taking the amount of training parameters,floating-point operation times and evaluation loss function as the evaluation index,the comprehensive evaluation function E was established to evaluate the classic networks.The results show that the training accuracy of Alexnet,Resnet50 and Resnet101 networks is more than 90%,and the training state is perfectly fitting.The evaluation results show that Alexnet is the most suitable network and a classical convolution neural network for coal gangue separation.

关 键 词：经典卷积神经网络 煤矸分选 机器视觉 数据增强 模型评价 

分 类 号：TD94[矿业工程—选矿]