基于深度学习的煤粉颗粒CT图像分割方法  

Coal particle CT image segmentation method based on deep learning

作　　者：李颖 李秀宇 卢兆林 李世银 LI Ying;LI Xiu-yu;LU Zhao-lin;LI Shi-yin(School of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221116,China)

机构地区：[1]中国矿业大学信息与控制工程学院,江苏徐州221116

出　　处：《计算机工程与设计》2022年第8期2252-2259,共8页Computer Engineering and Design

基　　金：国家自然科学基金青年科学基金项目(51604271)。

摘　　要：为准确获取煤粉颗粒粒度信息,提出一种基于深度学习的煤粉颗粒CT图像分割方法。通过在深度网络中添加注意力机制增强特征的通道信息和语义信息,可更准确地定位小颗粒,减少漏分割问题;针对煤粉颗粒形态的不规则性,重新设计分割分支,将连续卷积层的输出进行叠加获取新的特征,得到针对煤粉颗粒的精细化分割模型。实验结果表明了该算法在煤粉颗粒CT图像分割上的有效性。To obtain the particle size information of pulverized coal particles accurately,an image segmentation method based on deep learning was proposed to address the problem of CT image segmentation.The channel information and semantic information of the features were enhanced by adding an attention mechanism to the deep network,which more accurately located small par-ticles and reduced missing segmentation phenomemon.In view of the irregularity of the pulverized coal particle shape,the segmentation branch was redesigned and the output of the continuous convolutional layer was superimposed to obtain new features.The fine segmentation model for coal particle image was obtained.Experimental results verify the effectiveness of the proposed method in CT image segmentation of coal particles.

关 键 词：煤粉颗粒 颗粒识别 图像分割 深度学习 注意力机制 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]