融合图像外特征的岩屑岩性深度学习识别方法
DOI:
作者:
作者单位:

1.长江大学计算机科学学院 湖北 荆州;2.中国科学院大学计算机科学与技术学院;3.长江大学计算机科学学院

作者简介:

通讯作者:

中图分类号:

TP391

基金项目:

2020年新疆自治区创新人才建设专项自然科学计划(自然科学基金)(2020D01A132);湖北省科技示范项目(2019ZYYD016);长江大学(教育部、湖北省)非常规油气合作创新中心(UOG2020-10)


Deep learning method for rock debris lithology recognition by fusing external features of image
Author:
Affiliation:

1.College of Computer Science,Yangtze University,HuBei Jingzhou;2.School of Computer Science and Technology, University of Chinese Academy of Science;3.College of Computer Science, Yangtze University

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 文章评论
    摘要:

    岩屑的岩性识别是地质工作中的一项重要内容。为解决传统人工鉴别岩性的低效问题和传统机器识别的低可靠性问题,提出一种融合图像特征与图像外特征的岩性识别方法。首先采集岩屑的高分辨率图像,使用Xception特征提取器对图像特征进行提取并降维为一维向量,提高模型抽象特征敏感性并防止网络退化问题。同时量化岩屑的物理化学特征如:与盐酸反应程度、含矿物纯度、元素分析结果、硬度等,构建图像外特征向量。融合图像特征向量与图像外特征向量为总特征向量,构建神经网络与分类器进行训练,产生岩性识别模型。该模型相较于仅图像训练模型,在高质量岩屑图像数据集上提高3.45个百分点,在低质量岩屑图像数据集上提高20.92个百分点。该模型结合了传统录井与机器学习的优势,为建立可靠岩性剖面与实现数字化岩屑录井提供了更为高效的方法。

    Abstract:

    The lithology identification of rock debris is an important part of geological work. In order to solve the low efficiency problem of traditional artificial lithology identification and the low reliability problem of traditional machine recognition, a lithology identification method based on fusion of image features and external features is proposed. Firstly, the high-resolution images of rock debris are collected, and the Xception feature extractor is used to extract the image features and reduce the dimension to one-dimensional vector, so as to improve the sensitivity of model abstract features and prevent network degradation. At the same time, the physical and chemical characteristics of rock debris, such as reaction degree with hydrochloric acid, mineral purity, element analysis results, hardness and so on, are quantified to construct the external feature vector of the image. Combining the image feature vector and the image feature vector as the total feature vector, the neural network and classifier are constructed to train and generate the lithology recognition model. Compared with the only image training model, this model increases by 3.45 percentage points on the high-quality lithic image dataset and 20.92 percentage points on the low-quality lithic image dataset. This model combines the advantages of traditional logging and machine learning, and provides a more efficient method for establishing reliable lithology profile and realizing digital rock debris logging.

    参考文献
    相似文献
    引证文献
引用本文

覃本学,沈疆海,马丙鹏,等. 融合图像外特征的岩屑岩性深度学习识别方法[J]. 科学技术与工程, , ():

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-11-23
  • 最后修改日期:2022-04-24
  • 录用日期:2022-04-30
  • 在线发布日期:
  • 出版日期:
×
关于近期《科学技术与工程》编辑部居家办公的说明
亟待确认的版面费信息