摘要
通过人工观察岩石薄片来进行分类效率低,且容易受主观影响,在如今人工智能的时代背景下,使用计算机视觉技术对其进行智能处理,已经成为一种公认的研究方法。为此本文通过使用卷积神经网络来代替人工进行岩石薄片分类。本实验使用"中国科学数据"中的《部分造岩矿物、典型变质矿物和鲕粒薄片显微图像数据集》部分数据,采用数据增强手段进行处理,使其数据量扩增10倍。采用了ResNet模型,对其进行训练,最终准确率达到了96.8%。结果表明使用卷积神经网络对岩石薄片进行分类可以获得高效、准确的结果。
Classification by manually observing rock slices is inefficient and subject to subjective influence.In today's artificial in⁃telligence era,the use of computer vision technology for intelligent processing has become a recognized research method.To this end,this paper uses convolutional neural networks to replace artificial rock slice classification.This experiment uses part of the da⁃ta from"Partial Rock-forming Minerals,Typical Metamorphic Minerals,and Oolitic Thin Section Microscopic Image Data Set"in"Chinese Scientific Data",and uses data enhancement methods to process the data to increase the amount of data by 10 times.Us⁃ing the ResNet model and training it,the final accuracy rate reached 96.8%.The results show that using convolutional neural net⁃work to classify rock slices can obtain efficient and accurate results.
作者
贾立铭
梁少华
JIA Li-ming;LIANG Shao-hua(College of Computer Science,Yangtze University,Hubei 430000,China)
出处
《电脑知识与技术》
2021年第28期107-109,共3页
Computer Knowledge and Technology
关键词
岩石薄片
图像分类
卷积神经网络
残差网络
rock slice
image classification
convolution neural network
residual network
