| 453 | 1 | 177 |
| 下载次数 | 被引频次 | 阅读次数 |
在复杂的地质勘探条件下准确完成矿物识别是一项重要的任务。基于数据驱动的深度学习模型能精确识别各类岩石矿物，但需要构建庞大且完备的数据集，在实际情况下难以应用。针对此问题，结合小样本学习、度量学习以及元学习训练策略，使用EMD距离度量计算图像之间的结构距离，构建一种适应于小样本矿物的图像分类模型。核心思想在于利用图块级别度量并引入交叉参考权重机制，有效减少同类差异大和背景杂乱带来的影响，优于图与图判定分类的模型。在mini-ImageNet数据集上，5-way 1-shot和5-way 5-shot设置的分类准确率分别提高至55.91%、67.58%;将算法应用于小样本黏土矿物数据集上，5-way 5-shot设置的分类准确率为92.65%。实验结果表明，利用度量学习方法的分类精度高于其他小样本学习方法。
Abstract：Mineral identification is an important task of geological survey, which could be a big challenge in complex geological conditions. Data-driven deep learning model could accurately identify all kinds of rocks and minerals, but large and complete datasets should be constructed, and it was difficult in practical situations. Aiming at the problems, herein, an image classification model adapted to small sample minerals was proposed. It adopted the earth mover′s distance(EMD) as a metric to calculate the structural distance between images, which combined the small sample learning, the metric learning and meta-learning training strategy. The core idea of the method was to use the measure of graph block level and introduce the cross-reference weight mechanism, which could effectively reduce the influence caused by the large difference of the same class and the clutter of the background, which was better than the model of graph to graph direct decision classification. The results of 5-way 1-shot and 5-way 5-shot classification experiment on mini-ImageNet showed that the classification accuracy of 5-way 1-shot and 5-way 5-shot settings was improved to 55.91% and 67.58%. When the algorithm was applied to a Few-shot clay mineral data-set, the classification accuracy in 5-way 5-shot setting reached 92.65%. The experimental results showed that the classification accuracy of metric learning method was higher than other few-shot methods.
[1] 洪汉烈.黏土矿物古气候意义研究的现状与展望[J].地质科技情报，2010,29(1):1-8.HONG H L.A review on paleoclimate interpretation of clay minerals[J].Geological science and technology information,2010,29(1):1-8.
[2] 汤艳杰，贾建业，谢先德.黏土矿物的环境意义[J].地学前缘，2002,9(2):337-344.TANG Y J,JIA J Y,XIE X D.Environment significance of clay minerals[J].Earth science frontiers,2002,9(2):337-344.
[3] SINGER A.The paleoclimatic interpretation of clay minerals in sediments—a review[J].Earth-science reviews,1984,21(4):251-293.
[4] LIU Z F,COLIN C,TRENTESAUX A,et al.Erosional history of the eastern Tibetan Plateau since 190 kyr ago:clay mineralogical and geochemical investigations from the southwestern South China Sea[J].Marine geology,2004,209(1/2/3/4):1-18.
[5] LIU X K,SONG H J.Automatic identification of fossils and abiotic grains during carbonate microfacies analysis using deep convolutional neural networks[J].Sedimentary geology,2020,410:105790.
[6] LI C X,WANG D M,KONG L Y.Application of machine learning techniques in mineral classification for scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDS) images[J].Journal of petroleum science and engineering,2021,200:108178.
[7] KOCH G R,ZEMEL R,SALAKHUTDI-NOV R.Siamese neural networks for one-shot image recognition[C]//Proceedings of the 32nd International Conference on Machine Learning.Piscataway:IEEE Press,2015:103-123.
[8] SNELL J,SWERSKY K,ZEMEL R.Prototypical networks for few-shot learning[C]//Proceeding of Advances in Neural Information Processing Systems.Barcelona:NIPS,2017:4077-4087.
[9] WANG Y Q,YAO Q M,KWOK J T,et al.Generalizing from a few examples[J].ACM computing surveys,2021,53(3):1-34.
[10] 刘颖，雷研博，范九伦，等.基于小样本学习的图像分类技术综述[J].自动化学报，2021,47(2):297-315.LIU Y,LEI Y B,FAN J L,et al.Survey on image classification technology based on small sample learning[J].Acta automatica sinica,2021,47(2):297-315.
[11] FINN C,ABBEEL P,LEVINE S.Model-agnostic meta-learning for fast adaptation of deep networks[C]//Proceedings of the 34th International Conference on Machine Learning.New York:ACM Press,2017:1126-1135.
[12] SUNG F,YANG Y X,ZHANG L,et al.Learning to compare:relation network for few-shot learning[C]//Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press:1199-1208.
[13] VINYALS O,BLUNDELL C,LILLICRAP T,et al.Matching networks for one shot learning[C]//Proceeding of Advances in Neural Information Processing Systems.Barcelona:NIPS,2016:3630-3638.
[14] 陈珂，梁斌，左敬龙，等.一种用于中文微博情感分析的多粒度门控卷积神经网络[J].郑州大学学报(理学版),2020,52(3):21-26,33.CHEN K,LIANG B,ZUO J L,et al.Multiple grains-gated convolutional neural networks for Chinese microblog sentiment analysis[J].Journal of Zhengzhou university (natural science edition),2020,52(3):21-26,33.
[15] 王知人，谷昊晟，任福全，等.基于深度卷积残差学习的图像超分辨[J].郑州大学学报(理学版),2020,52(3):42-48.WANG Z R,GU H S,REN F Q,et al.Residual learning of deep CNN for image super-resolution[J].Journal of Zhengzhou university (natural science edition),2020,52(3):42-48.
[16] YANG Q S,YAN P K,ZHANG Y B,et al.Low-dose CT image denoising using a generative adversarial network with Wasserstein distance and perceptual loss[J].IEEE transactions on medical imaging,2018,37(6):1348-1357.
[17] KULIS B.Metric learning:a survey[J].Foundations and trends? in machine learning,2013,5(4):287-364.
[18] 祝钧桃，姚光乐，张葛祥，等.深度神经网络的小样本学习综述[J].计算机工程与应用，2021,57(7):22-33.ZHU J T,YAO G L,ZHANG G X,et al.Survey of few shot learning of deep neural network[J].Computer engineering and applications,2021,57(7):22-33.
[19] THRUN S,PRATT L.Learning to Learn[M].Boston:Springer,1998.
[20] MARTIN A,SOUMITH C,LéON B.Wasserstein generative adversarial netWorks[C]//International Conference on Machine Learning.Sydney:JMLR,2017:214-223.
[21] SUN Q R,LIU Y Y,CHUA T S,et al.Meta-transfer learning for few-shot learning[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2019:403-412.
[22] RUBNER Y,TOMASI C,GUIBAS L J.The earth mover′s distance as a metric for image retrieval[J].International journal of computer vision,2000,40(2):99-121.
[23] ZHANG C,CAI Y J,LIN G S,et al.DeepEMD:few-shot image classification with differentiable earth mover′s distance and structured classifiers[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE Press,2020:12200-12210.
[24] CHEN W Y,LIU Y C,KIRA,et al.A closer look at few-shot classification[EB/OL].[2022-02-15].https://www.xueshufan.com/publication/2911906319.
[25] RUSSAKOVSKY O,DENG J,SU H,et al.ImageNet large scale visual recognition challenge[J].International journal of computer vision,2015,115(3):211-252.
基本信息:
DOI：10.13705/j.issn.1671-6841.2023176
中图分类号:P575;TP391.41
引用信息:
[1]杜睿山,张轶楠,孟令东,等.基于EMD距离度量的小样本矿物图像分类[J],2023,55(06):63-70.DOI:10.13705/j.issn.1671-6841.2023176.
基金信息:
国家自然科学基金联合基金项目(U20A2093);; 东北石油大学引导性创新基金项目(2020YDL-04)
2022-06-25
2022
2023-08-21
2023
2