现代电子技术2024，Vol.47Issue(24)：131-136,6.DOI:10.16652/j.issn.1004-373x.2024.24.020
基于改进ConvNeXt模型的轻量化岩屑岩性识别方法
Method of lightweight rock debris lithology identification based on improved ConvNeXt model
摘要
Abstract
Rock debris logging plays an important role in geological analysis and oil and gas exploration,among which rock debris lithology identification is a key content.However,current identification methods are difficult to balance accuracy and lightweight.In order to improve the identification efficiency and ensure efficient logging work,an identification method based on improved lightweight ConvNeXt model is proposed.By adjusting the network layers of the ConvNeXt model and replacing most of the traditional convolutions with Ghost convolutions,the redundant parameters of the model are reduced.The SimAM parameterless attention mechanism is integrated to improve the model performance without increasing model parameters.The experimental results show that the improved ConvNeXt model has a parameter count and weight file size of only 15.44×106 and 15.5 MB,respectively,with a floating-point operation of 0.69×109.The parameter count and operation count are much lower than most network models,making it easy to embed into various mobile devices.Moreover,the comprehensive recognition accuracy on the test set can reach 99.64%,slightly higher than the ConvNeXt-S model,meeting the reliability requirements of recognition.The proposed model combines the characteristics of accuracy and lightweight,and can effectively meet the rock debris identification tasks in practical logging work.关键词
岩屑岩性识别/ConvNeXt模型/Ghost卷积/SimAM注意力机制/轻量化/消融实验Key words
rock debris lithology identification/ConvNeXt model/Ghost convolution/SimAM attention mechanism/lightweight/ablation experiment分类
信息技术与安全科学引用本文复制引用
杨宇轩,钟宝荣..基于改进ConvNeXt模型的轻量化岩屑岩性识别方法[J].现代电子技术,2024,47(24):131-136,6.基金项目
国家自然科学基金项目(62006028) （62006028）
湖北省自然科学基金项目(2023AFB909) （2023AFB909）
