现代电子技术2025，Vol.48Issue(23)：25-34,10.DOI:10.16652/j.issn.1004-373x.2025.23.004
基于特征描述和注意力机制的层理构造分类
Bedding structure classification method based on feature description and attention mechanism
摘要
Abstract
The recognition of sedimentary rock bedding structure images is the focus of attention in the field of petroleum geological exploration.Due to the complexity and specialization of bedding features,a bedding structure recognition model LC-ResNeXt based on texture feature description and parallel attention mechanism is proposed.The ResNeXt50 is selected as the backbone network,and the improved LBP(local binary pattern)feature description operator is used to describe the RGB image to highlight the texture features of the bedding structure images.The LBP image is input into the feature extraction network.The SoftPool is introduced to improve the pooling layer of the network and retain the key information in the feature map.The improved parallel CBAM(convolutional block attention module)is integrated into the ResNeXt50 network to improve the model's learning effect and fusion ability for key bedding features.The experimental results show that the LBP images contain richer texture features in comparison with the RGB images.In addition,the LC-ResNeXt model can further improve the accuracy rate of recognition and classification of sedimentary rock bedding structure images in comparison with the other models,so it is helpful for related research in the field of petroleum geology.关键词
层理构造/局部二值模式/注意力机制/深度学习/特征提取/图像分类Key words
bedding structure/LBP/attention mechanism/deep learning/feature extraction/image classification分类
电子信息工程引用本文复制引用
刘晨,沈疆海..基于特征描述和注意力机制的层理构造分类[J].现代电子技术,2025,48(23):25-34,10.基金项目
中国高校产学研创新基金(2021ALA01004) （2021ALA01004）
新疆自治区创新人才建设专项自然科学计划(2020D01A132) （2020D01A132）
