基于深度残差网络的煤泥浮选泡沫图像分类方法研究  

Study on Forth Image Classification of Coal Flotation Based on Deep Residual Network

作　　者：杨林顺 刘航涛 YANG Linshun;LIU Hangtao(Tunlan Coal Preparation Plant,Shanxi Coking Coal Xishan Coal and Electricity,Taiyuan 030299,China;School of Chemical and Technology Engineering,China University of Mining and Technology,Xuzhou 221116,China)

机构地区：[1]山西焦煤西山煤电屯兰选煤厂,太原030299 [2]中国矿业大学化工学院,江苏徐州221116

出　　处：《煤炭技术》2023年第7期226-229,共4页Coal Technology

基　　金：国家自然科学基金项目(92062109)。

摘　　要：针对浮选精煤灰分在线检测困难、识别效率低和主观性强等问题,研究了一种基于深度残差网络的浮选精煤泡沫图像分类方法。以6种不同灰分级别的精煤泡沫图像为数据集,对数据集进行了数据增强,采用基于Pytorch框架的ResNet18,ResNet50,ResNet152神经网络模型进行训练分类,通过训练过程的损失值和准确率判断模型的性能。结果表明深度残差网络具有良好的准确度,现场应用效果良好,随着网络层数的加深,模型的分类效果越好。深度残差网络能有效地识别泡沫图像,在煤泥浮选图像分类中具有很大的应用潜力。To address the problems associated with online detection,low recognition efficiency,and strong subjectivity of concentrate ash content of coal flotation,this paper proposes a froth images classification method for coal flotation based on the deep residual network.Using six class images of concentrate ash content interval as the dataset,data augmentation was used for this dataset.ResNet18,ResNet50,and ResNet152 neural network models based on the Pytorch framework were adopted for training classification and used the loss value of the training process and accuracy to determine the performance of the model.The results show that the deep residual network has good performance and good field application effect.The classification accuracy of the model increases as the deepening of the residual network increases.The deep residual network can effectively classify froth images and has great potential in the application of image classification of slime flotation.

关 键 词：煤泥浮选 深度学习 精矿灰分 深度残差网络 

分 类 号：TD94[矿业工程—选矿]