郑爽, 梁云浩, 武俊峰, 乔壮, 刘付刚. 基于AlexNet-SN网络的煤与煤矸石分类方法[J]. 中国矿业, 2022, 31(6): 79-85. DOI: 10.12075/j.issn.1004-4051.2022.06.007
    引用本文: 郑爽, 梁云浩, 武俊峰, 乔壮, 刘付刚. 基于AlexNet-SN网络的煤与煤矸石分类方法[J]. 中国矿业, 2022, 31(6): 79-85. DOI: 10.12075/j.issn.1004-4051.2022.06.007
    ZHENG Shuang, LIANG Yunhao, WU Junfeng, QIAO Zhuang, LIU Fugang. Classification method of coal and coal gangue based on AlexNet-SN network[J]. CHINA MINING MAGAZINE, 2022, 31(6): 79-85. DOI: 10.12075/j.issn.1004-4051.2022.06.007
    Citation: ZHENG Shuang, LIANG Yunhao, WU Junfeng, QIAO Zhuang, LIU Fugang. Classification method of coal and coal gangue based on AlexNet-SN network[J]. CHINA MINING MAGAZINE, 2022, 31(6): 79-85. DOI: 10.12075/j.issn.1004-4051.2022.06.007

    基于AlexNet-SN网络的煤与煤矸石分类方法

    Classification method of coal and coal gangue based on AlexNet-SN network

    • 摘要: 现有煤矸石分选方法主要依据人工设计特征对煤矸石进行识别,但特征提取过程复杂,准确率也较低。随着人工智能技术的快速发展,智能选矸成为解决煤矸石分拣问题的重要研究方向。为提高煤与煤矸石分类准确率,本文提出了一种基于AlexNet网络和风格迁移技术改进的煤矸石分拣方法。选用3×3的卷积核代替原AlexNet网络前几层中较大的卷积核,利用BN层代替LRN层和Dropout,并采用风格迁移数据增强法提高煤与煤矸石数据集的多样性。研究结果表明,与原始的AlexNet网络相比,该方法的准确率提高了1.8%,损失率下降了2.0%。此方法不仅能够满足煤与煤矸石实时检测的要求,而且具有更高的识别精度,能有效应用于煤矸石识别。

       

      Abstract: The existing coal gangue separation methods mainly identify coal gangue based on manual design features, but the feature extraction process is complex and the accuracy is low.With the rapid development of artificial intelligence technology, intelligent gangue separation has become an important research direction to solve the problem of coal gangue sorting.In order to improve the classification accuracy of coal and coal gangue, an improved coal gangue sorting method based on AlexNet network and style migration technology is proposed in this paper.Selecting 3×3 instead of the larger convolution kernel in the first few layers of the original AlexNet network, and BN layer is used to replace LRN and Dropout, and using the style migration data enhancement method to improve the diversity of coal and coal gangue data sets.The results show that compared with the original AlexNet network, the accuracy of this method is improved by 1.8% and the loss rate is reduced by 2.0%.This method can not only meet the requirements of real-time detection of coal and coal gangue, but also has higher recognition accuracy, and can be effectively applied to coal gangue recognition.

       

    /

    返回文章
    返回