基于多特征卷积神经网络的哨兵二号影像地物分类

doi:10.3969/j.issn.0255-8297.2023.05.004

摘要/Abstract

摘要： 空间分辨率为10 m的哨兵二号影像在原始的GoogLeNet中以影像的光谱值作为输入，没有将影像中的地物视为一个整体对象，为了利用影像的面向对象特征，提出了基于多特征的Object-oriented GoogLeNet网络结构。Object-oriented GoogLeNet在原有模型的基础上，引入了面向对象的光谱特征和形状特征，充分利用了不同地物间差异的形状特征进行分类。在武汉市及其周边的无云影像制作的数据集上，Object-oriented GoogLeNet模型的分类结果总体精度在GoogLeNet基础上提升了1.773%。结果表明，引入面向对象的特征模型在哨兵二号遥感影像分类中效果更好。

关键词: 哨兵二号, GoogLeNet, 深度学习, 形状特征, 土地利用分类

Abstract: The 10 m resolution Sentinel-2 images takes the spectral values of the image as input in the original GoogLeNet without treating the ground objects in the image as a whole. To leverage object-oriented features in Sentinel-2 remote sensing image classification, this paper proposes an Object-oriented GoogLeNet network structure based on multiple features. Object-oriented GoogLeNet incorporates object-oriented spectral and shape features, and fully utilizes the shape features of differences between different ground objects for classification. On the data set of cloudless images in Wuhan and its surrounding areas, the overall accuracy of the classification results of Object-oriented GoogLeNet model has increased by 1.773% compared to GoogLeNet. The results show that the model with object-oriented features enhances the classification performance of Sentinel-2 remote sensing images.

Key words: Sentinel-2, GoogLeNet, deep learning, shape features, land use classification

中图分类号:

TP751.1

黄显培, 孟庆祥. 基于多特征卷积神经网络的哨兵二号影像地物分类[J]. 应用科学学报, 2023, 41(5): 766-776.

HUANG Xianpei, MENG Qingxiang. Land Cover Classification of Sentinel-2 Image Based on Multi-feature Convolution Neural Network[J]. Journal of Applied Sciences, 2023, 41(5): 766-776.

参考文献

[1] 张鑫龙, 陈秀万, 李飞, 等. 高分辨率遥感影像的深度学习变化检测方法[J]. 测绘学报, 2017, 46(8):999-1008. Zhang X L, Chen X W, Li F, et al. Change detection method for high resolution remote sensing images using deep learning[J]. Acta Geodaetica et Cartographica Sinica, 2017, 46(8):999-1008. (in Chinese)
[2] 王卷乐, 程凯, 韩雪华, 等. 大数据驱动的资源学科领域数据分析前沿与应用[J]. 数据与计算发展前沿, 2020, 2(2):20-30. Wang J L, Cheng K, Han X H, et al. Big data driven data technology analysis frontier and application in resource discipline[J]. Frontiers of Data & Computing, 2020, 2(2):20-30. (in Chinese)
[3] 李德仁, 童庆禧, 李荣兴, 等. 高分辨率对地观测的若干前沿科学问题[J]. 中国科学:地球科学, 2012, 42(6):805-813. Li D R, Tong Q X, Li R X, et al. Current issues in high-resolution earth observation technology[J]. Scientia Sinica (Terrae), 2012, 42(6):805-813. (in Chinese)
[4] 巫兆聪, 刘培, 巫远. 一种多光谱遥感应用需求综合方法[J]. 应用科学学报, 2017, 35(5):658-666. Wu Z C, Liu P, Wu Y. Synthesis of requirements in applications of multi-spectral remote sensing[J]. Journal of Applied Sciences, 2017, 35(5):658-666. (in Chinese)
[5] 巫兆聪, 项伟, 李俊, 等. 基于满足度评估的遥感应用需求聚类算法[J]. 应用科学学报, 2018, 36(4):635-643. Wu Z C, Xiang W, Li J, et al. Clustering algorithm for remote sensing application requirements based on satisfaction evaluation[J]. Journal of Applied Sciences, 2018, 36(4):635-643. (in Chinese)
[6] 魏香琴, 顾行发, 余涛, 等. 面向应用需求的遥感卫星载荷空间分辨率标准化研究[J]. 光谱学与光谱分析, 2012, 32(3):781-785. Wei X Q, Gu X F, Yu T, et al. Spatial resolution standardization of payload on board of remote sensing satellite based on application requirements[J]. Spectroscopy and Spectral Analysis, 2012, 32(3):781-785. (in Chinese)
[7] 胡杰, 张莹, 谢仕义. 国产遥感影像分类技术应用研究进展综述[J]. 计算机工程与应用, 2021, 57(3):1-13. Hu J, Zhang Y, Xie S Y. Summary of research progress on application of domestic remote sensing image classification technology[J]. Computer Engineering and Applications, 2021, 57(3):1-13. (in Chinese)
[8] 张裕, 杨海涛, 袁春慧. 遥感图像分类方法综述[J]. 兵器装备工程学报, 2018, 39(8):108-112. Zhang Y, Yang H T, Yuan C H. A survey of remote sensing image classification methods[J]. Journal of Sichuan Ordnance, 2018, 39(8):108-112. (in Chinese)
[9] 谢天, 袁正颖, 杨海从, 等. 基于遥感影像的土地利用/覆盖变化分类方法研究综述[J]. 地球科学前沿, 2020, 10(6):500-507. Xie T, Yuan Z Y, Yang H C, et al. Review of land use/cover change classification methods based on remote sensing image[J]. Advances in Geosciences, 2020, 10(6):500-507. (in Chinese)
[10] 赵怡, 许剑辉, 钟凯文, 等. 基于Sentinel-2A和Landsat8的城市不透水面的提取[J]. 国土资源遥感, 2021, 33(2):40-47. Zhao Y, Xu J H, Zhong K W, et al. Impervious surface extraction based on Sentinel-2A and Landsat8[J]. Remote Sensing for Land and Resources, 2021, 33(2):40-47. (in Chinese)
[11] Alex K, Ilya S, Hinton G E. ImageNet classification with deep convolutional neural networks[J]. Communications of the ACM, 2017, 60(6):84-90.
[12] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[C]//International Conference on Learning Representations (ICLR), 2015:1-14.
[13] Szegedy C, Liu W, Jia Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2015:1-9.
[14] 党宇, 张继贤, 邓喀中, 等. 基于深度学习AlexNet的遥感影像地表覆盖分类评价研究[J]. 地球信息科学学报, 2017, 19(11):1530-1537. Dang Y, Zhang J X, Deng K Z, et al. Study on the evaluation of land cover classification using remote sensing images based on AlexNet[J]. Journal of Geo-information Science, 2017, 19(11):1530-1537. (in Chinese)
[15] 阚希, 张永宏, 曹庭, 等. 利用多光谱卫星遥感和深度学习方法进行青藏高原积雪判识[J]. 测绘学报, 2016, 45(10):1210-1221. Kan X, Zhang Y H, Cao T, et al. Snow cover recognition for Qinghai-Tibetan plateau using deep learning and multispectral remote sensing[J]. Acta Geodaetica et Cartographica Sinica, 2016, 45(10):1210-1221. (in Chinese)
[16] 孔令婷, 杨英宝, 章勇. 基于不同分辨率遥感影像的分类方法对比研究[J]. 测绘与空间地理信息, 2015, 38(3):40-43. Kong L T, Yang Y B, Zhang Y. Study on classification methods based on different resolution remote sensing image[J]. Geomatics & Spatial Information Technology, 2015, 38(3):40-43. (in Chinese)
[17] 蒋治浩, 林辉, 张怀清, 等. 面向对象结合卷积神经网络的GF-1影像遥感分类[J]. 中南林业科技大学学报, 2021, 41(8):45-55, 67. Jiang Z H, Lin H, Zhang H Q, et al. Remote sensing image of GF-1 classification using objectoriented method and convolutional neural network[J]. Journal of Central South University of Forestry & Technology, 2021, 41(8):45-55, 67. (in Chinese)
[18] 马长辉, 黄登山. 纹理与几何特征信息在高空间分辨率遥感影像分类中的应用[J]. 测绘地理信息, 2019, 44(6):66-70, 92. Ma C H, Huang D S. Application of texture features and geometric feature information in high spatial resolution remote rensing image classification[J]. Journal of Geomatics, 2019, 44(6):66-70, 92. (in Chinese)
[19] 杨闫君, 田庆久, 占玉林, 等. 空间分辨率与纹理特征对多光谱遥感分类的影响[J]. 地球信息科学学报, 2018, 20(1):99-107. Yang Y J, Tian Q J, Zhan Y L, et al. Effects of spatial resolution and texture features on multi-spectral remote sensing classification[J]. Journal of Geo-information Science, 2018, 20(1):99-107. (in Chinese)
[20] 夏炎, 黄亮, 陈朋弟. 模糊超像素分割算法的无人机影像烟株精细提取[J]. 国土资源遥感, 2021, 33(1):115-122. Xia Y, Huang L, Chen P D. Tobacco fine extraction from UAV image based on fuzzysuperpixel segmentation algorithm[J]. Remote Sensing for Land & Resources, 2021, 33(1):115-122. (in Chinese)
[21] 郭鹏程, 周志易. 面向对象方法支持下的土地利用分类[J]. 北京测绘, 2021, 35(5):616-621. Guo P C, Zhou Z Y. Land use classification supported by object-based methods[J]. Beijing Surveying and Mapping, 2021, 35(5):616-621. (in Chinese)
[22] 陈云浩, 冯通, 史培军, 等. 基于面向对象和规则的遥感影像分类研究[J]. 武汉大学学报(信息科学版), 2006(4):316-320. Chen Y H, Feng T, Shi P J, et al. Classification of remote sensing image based on object oriented and class rules[J]. Geomatics and Information Science of Wuhan University, 2006(4):316-320. (in Chinese)
[23] Robert C, Ian H, Joshua G, et al. Empirical cross sensor comparison of Sentinel-2A and 2B MSI, Landsat-8 OLI, and Landsat-7 ETM+ top of atmosphere spectral characteristics over the conterminous United States[J]. Remote Sensing of Environment, 2019(221):274-285.
[24] 邓书涵, 彭祎, 董可. 基于分形理论的洪山区土地利用变化研究[J]. 科技创新与应用, 2021, 11(23):83-86. Deng S H, Peng W, Dong K. Study on land use change in Hongshan district based on fractal theory[J]. Technology Innovation and Application, 2021, 11(23):83-86. (in Chinese)
[25] Smith L. Cyclical learning rates for training neural networks[C]//2017 IEEE Winter Conference on Applications of Computer Vision (WACV), 2017, 464-472.