摘要
将Dense Net卷积神经网络模型与迁移学习技术相结合,应用于岩石岩性识别.传统岩石识别方法依赖经验且耗时耗力,易受主观因素影响,而深度学习的卷积神经网络能自动学习和提取特征.Dense Net模型连接紧密,增强特征重用性,提高信息传递效率.迁移学习可将知识和经验迁移到新任务,改善性能.实验选取石灰岩、大理石、石英岩和砂岩四类岩石图像进行测试,训练准确率趋于100%,测试准确率基本稳定在80%左右,最高预测准确率83.2%,表明模型训练效果理想,鲁棒性和泛化能力较强.未来可进一步收集更丰富专业的数据集并优化模型以提高准确率.
This study aims to combine the Dense Net convolutional neural network model with transfer learning techniques for rock lithology recognition.Traditional rock identification methods rely on experience and are time-consuming and labor-intensive,and are susceptible to subjective factors;while deep learning convolutional neural networks can automatically learn and extract features.The Dense Net model has tight connections,enhancing feature reusability and improving information transmission efficiency.Transfer learning can transfer knowledge and experience to new tasks,improving performance.Four types of rock images,namely limestone,marble,quartzite,and sandstone,were selected for testing in the experiment.The training accuracy tended to reach 100%,and the testing accuracy remained stable at around 80%,with the highest prediction accuracy reaching 83.2%.This indicates that the model training effect is ideal,with strong robustness and generalization ability.In the future,more diverse and professional datasets can be further collected and models can be optimized to improve accuracy.
作者
杨建松
曹成
YANG Jiansong;CAO Cheng(School of Mathematics and Computer Science,Shaanxi University of Technology,Hanzhong 723001,China)
出处
《西安文理学院学报(自然科学版)》
2025年第3期61-67,共7页
Journal of Xi’an University(Natural Science Edition)
