面向地理国情监测的CRF遥感影像分类

doi:10.3969/j.issn.0255-8297.2016.04.005

应用科学学报 ›› 2016, Vol. 34 ›› Issue (4): 397-404.doi: 10.3969/j.issn.0255-8297.2016.04.005

面向地理国情监测的CRF遥感影像分类

张春森¹, 冯晨轶¹, 崔卫红², 郑艺惟¹, 孙志伟³

1. 西安科技大学测绘科学与技术学院, 西安 710054;
2. 武汉大学遥感信息工程学院, 武汉 430079;
3. 北京吉威时代软件股份有限公司, 北京 100043

收稿日期:2016-01-17 修回日期:2016-03-22 出版日期:2016-07-30 发布日期:2016-07-30
作者简介:张春森,博士,教授,研究方向:摄影测量与遥感,E-mail:zhchunsen@aliyun.com
基金资助:
国家自然科学基金(No.41101410)资助

Remote Sensing Image Classification for NationalGeomatics Monitoring Based on ConditionalRandom Field

ZHANG Chun-sen¹, FENG Chen-yi¹, CUI Wei-hong², ZHENG Yi-wei¹, SUN Zhi-wei³

1. College of Geomatics, Xi'an University of Science and Technology, Xi'an 710054, China;
2. School of Remote Sensing and Information Engineering, Wuhan University, Wuhan 430079, China;
3. GEOWAY, Beijing 100043, China

Received:2016-01-17 Revised:2016-03-22 Online:2016-07-30 Published:2016-07-30

摘要/Abstract

摘要：

针对地理国情监测中地表覆盖信息的提取，提出了一种基于条件随机场的高分辨率遥感影像自动分类方法.与面向对象的传统分类方法不同，该方法基于概率图模型分别计算像素级和对象级的势函数，以及像素与它所属对象之间的层间势函数，将所得势函数统一到一个CRF模型中进行图割求解.该方法较充分地表达了像素与对象之间的关系，从而降低了对象分割误差传递对影像分类结果的影响.以“高分1号”遥感影像为实验数据，借鉴地理国情普查中地表覆盖分类体系进行实验验证.分类总体精度和平均精度分别达到91.08%和86.95%，远高于基于面向对象的分类结果.

关键词: 影像分类, 条件随机场, 地表覆盖信息提取, 地理国情监测

Abstract:

A method of high-resolution remote sensing image classification for national geomatics monitoring based on conditional random field (CRF) is proposed. Different from object-based classification, the proposed method is based on probabilistic graphical models that calculate an energy function composed of potentials. The potentials are defined on the pixel-level and the object-level, and the level between pixels and objects. The energy function can be computed by solving a minimum s-t cut problem using a move making algorithm based on powerful graph cut. Since the classification method fully expresses relationship between pixels and objects, classification errors caused by segmentation is effectively reduced. High-resolution remote sensing images of GW-1 are used in the experiment. The overall classification accuracy and average classification accuracy reached 91.08% and 86.95% respectively, much higher than the results of object-based classification.

Key words: national geomatics monitoring, conditional random field (CRF), image classification, land cover information extraction

中图分类号:

TN911.73

张春森, 冯晨轶, 崔卫红, 郑艺惟, 孙志伟. 面向地理国情监测的CRF遥感影像分类[J]. 应用科学学报, 2016, 34(4): 397-404.

ZHANG Chun-sen, FENG Chen-yi, CUI Wei-hong, ZHENG Yi-wei, SUN Zhi-wei. Remote Sensing Image Classification for NationalGeomatics Monitoring Based on ConditionalRandom Field[J]. Journal of Applied Sciences, 2016, 34(4): 397-404.

参考文献

[1] 胥海威, 杨敏华, 韩瑞梅, 王振兴. 用随机决策树群算法进行高光谱遥感影像分类[J]. 应用科学学报, 2011, 29(6): 598-604. Xu H W, Yang M H, Han R M, Wang Z X. Hyperspectral remote sensing image classification with extremely randomized clustering forests [J]. Journal of Applied Sciences, 2011, 29(6): 598-604.(in Chinese)
[2] 曾波, 赵展. 地理国情普查中高分辨率遥感影像自动分类技术研究[J]. 测绘通报, 2015(1): 95-98. Zeng B, Zhao Z. Research on high spatial resolution remote sensing image classification for geographic national conditions investigation [J]. Bulletin of Surveying and Mapping, 2015(1): 95-98.(in Chinese)
[3] 白晓燕, 陈晓宏, 王兆礼. 基于面向对象分类的土地利用信息提取及其时空变化研究[J]. 遥感技术与应用, 2015, 30(4): 798-809. Bai X Y, Chen X H, Wang Z L. A study on land use information extraction based on object-oriented classification technology and the temporal-spatial variation [J]. Remote Sensing Technology and Application, 2015, 30(4): 798-809.(in Chinese)
[4] 张俊, 于庆国, 侯家槐. 面向对象的高分辨率影像分类与信息提取[J]. 遥感技术与应用, 2010, 25(1): 112-117. Zhang J, Yu Q G, Hou J H. Object-oriented classification and information extraction based on high spatial resolution remote sensing image [J]. Remote Sensing Technology and Application, 2010, 25(1): 112-117.(in Chinese)
[5] 朱俊杰, 杜小平, 范湘涛, 郭华东. GIS数据约束的海岸带SAR图像多尺度分割[J]. 应用科学学报, 2013, 31(1): 79-83. Zhu J J, Du X P, Fan X T, Guo H D. Gis-constrained multi-scale coastal SAR image segmentation [J]. Journal of Applied Sciences, 2013, 31(1): 79-83.(in Chinese)
[6] Lafferty J D, Mccallum A, Fernando C N P. Conditional random fields: probabilistic models for segmenting and labeling sequence data [C]//18th International Conference on Machine Learning (ICML 2001), 2001: 282-289.
[7] Rother C, Kolmogorov V, Blake A. "GrabCut": interactive foreground extraction using iterated graph cuts [J]. Acm Transactions on Graphics, 2004, 23(3): 307-312.
[8] Ladicky L, Russell C, Kohli P, Torr P H S. Associative hierarchical random fields [J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 36(6): 1-1.
[9] Shotton J, Winn J, Rother C, Criminisi A. TextonBoost: joint appearance, shape and context modeling for multi-class object recognition and segmentation [M]//Computer Vision-ECCV 2006. Berlin Heidelberg: Springer, 2015: 1-15.
[10] Wang C, Komodakis N, Paragios N. Markov random field modeling, inference & learning in computer vision & image understanding: A survey [J]. Computer Vision and Image Understanding, 2013, 117(11): 1610-1627.
[11] Kitamura Y, Li Y, Ito W, Ishikawa H. Data-dependent higher-order clique selection for artery-vein segmentation by energy minimization [J]. International Journal of Computer Vision, 2015: 1-17.
[12] Kohli P, Torr P H S. Robust higher order potentials for enforcing label consistency [J]. International Journal of Computer Vision, 2009, 82(3): 302-324.
[13] Kohli P, Kumar M P, Torr P H S. P3 & beyond: solving energies with higher order cliques[C]//2007 IEEE Conference on Computer Vision and Pattern Recognition, Minneapolis, USA: 2007: 1-8.
[14] Comaniciu D, Meer P. Mean shift: a robust approach toward feature space analysis [J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2002, 24(5): 603-619.
[15] Boykov Y, Veksler O, Zabih R. Fast approximate energy minimization via graph cut [J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2001, 23(11): 1222-1239.
[16] Julesz B. Textons, the elements of texture perception, and their interactions [J]. Nature, 1981, 290(5802): 91-97.
[17] 金晶, 邹峥嵘, 陶超. 高分辨率遥感影像的压缩纹理元分类[J]. 测绘学报, 2014, 43(5): 493-499. Jin J, Zou Z R, Tao C.Compressed texton based high resolution remote sensing image classification [J].Acta Geodaetica et Cartographica Sinica, 2014, 43(5): 493-499.(in Chinese)
[18] Ojala T, Pietikainen M, Maenpaa T. Multiresolution gray-scale and rotation invariant texture classification with local binary patterns [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(7): 971-987.
[19] Lowe D G. Distinctive image features from scale-invariant keypoints [C]//International Journal of Computer Vision, 2004: 91-110.
[20] Ihler A T, Iii J, Willsky A S. Loopy belief propagation: convergence and effects of message errors [J]. Journal of Machine Learning Research, 2005, 6(5): 905-936.
[21] Boykov Y, Kolmogorov V. An experimental comparison of MIN-cut/MAX-flow algorithms for energy minimization in vision [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2004, 26(9): 1124-1137.
[22] Orlin J B. A faster strongly polynomial time algorithm for submodular function minimization[J]. Mathematical Programming, 2009, 118(2): 237-251.

面向地理国情监测的CRF遥感影像分类

Remote Sensing Image Classification for NationalGeomatics Monitoring Based on ConditionalRandom Field

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 2

编辑推荐

Metrics

本文评价

[1]	吴韵清, 吴鹏, 陈北京, 鞠兴旺, 高野. 基于残差全卷积网络的图像拼接定位算法[J]. 应用科学学报, 2019, 37(5): 651-662.
[2]	付仲良，张文元，孟庆祥. 灰度和纹理特征组合的SAR影像SVM分类[J]. 应用科学学报, 2012, 30(5): 498-504.