Remote sensing image change detection based on cyclic convolution projection and EM algorithm
-
摘要: 不同时态的遥感影像变化检测已成为遥感快速灾害评估中具有挑战性的研究课题. 针对如何确定不同时态遥感影像中像素差异影像合适的变化阈值问题，提出一种基于循环卷积投影与最大期望算法(EM)的遥感影像变化检测方法. 首先，基于循环卷积将两个时态的影像互相投影到各自的成像模态，使得投影后的影像与目标影像具有相同统计量，逐像素相减可得两个影像之间的像素差异影像；其次，基于贝叶斯最小错误率理论和像素差异影像直方图的分布特点，假设像素差异影像服从广义高斯分布；最后，采用EM迭代求取广义高斯分布的各项参数，当变化类像素与未变化类像素的条件概率相等即可得到最佳变化阈值，完成自适应的变化检测过程. 采用不同成像方式、分辨率和不同变化类型下采集的多对遥感影像进行实验. 结果表明：本文方法在实验数据集上的检测精度较对比方法平均提升2.55%，FM综合评价指标较对比方法平均提升0.086，可以更有效地检测出遥感影像的变化区域，具有较强的鲁棒性.
-
关键词:
- 遥感影像 /
- 循环卷积投影 /
- 变化检测 /
- 最大期望算法(EM) /
- 差异影像
Abstract: The detection of changes in remote sensing images in different tenses has become a challenging research topic in rapid remote sensing damage assessment. Aiming at the problem of how to determine the appropriate change threshold of pixel difference images in different temporal remote sensing images, this paper proposes a remote sensing image change detection method based on cyclic convolution projection and Expectation-Maximization algorithm (EM). First, the two temporal images are projected into their respective imaging modalities based on cyclic convolution, so that the projected image and the target image have the same statistics, and the pixel difference image between the two images can be obtained by subtracting pixel by pixel; secondly, based on the Bayesian minimum error rate theory and the distribution characteristics of the pixel difference image histogram, it is assumed that the pixel difference image obeys the generalized Gaussian distribution; finally, the maximum expectation algorithm is used to iteratively obtain the various parameters of the generalized Gaussian distribution. If the conditional probability of the pixel and the unchanged pixel are equal, the optimal change threshold can be obtained, and the adaptive change detection process can be completed. Experiments were performed on multiple pairs of remote sensing images collected under different imaging methods, resolutions and different types of changes. The results show that the detection accuracy of this method on the experimental data set is 2.55% higher than the comparison method on average, and the FM comprehensive evaluation index is increased by 0.086 on the average compared with the comparison method. It can detect the change area more accurately and has strong robustness. -
表 1 实验数据集
数据集 拍摄日期 地点 影像大小 事件(分辨率) 传感器 Dataset#1 1995-09—1996-07 Sardinia，It 412×300 湖泊溢出(30 m) Landsat-5/Optical Dataset#2 2006-07—2007-07 Gloucester，UK 2325×4135 洪水(0.65 m) TerraSAR-X/QuickBird 2 Dataset#3 2012-05—2013-07 Toulouse，Fr 2000×2000 建设用地(0.52 m) Pleiades/WorldView2 
下载: 导出CSV
表 3 不同方法在Dataset#1上的检测精度
数据集 方法 ACC/% $ \mathrm{F}\mathrm{M} $ Dataset#1 方法一 93.31 0.4988 方法二 93.95 0.5003 方法三 94.20 0.5136 本文方法 96.02 0.5478
下载: 导出CSV
表 4 不同方法在Dataset#2上的检测精度
数据集 方法 ACC/% $ \mathrm{F}\mathrm{M} $ Dataset#2 方法一 88.21 0.4228 方法二 88.98 0.4302 方法三 89.60 0.4444 本文方法 91.97 0.4749
下载: 导出CSV
表 5 不同方法在Dataset#3上的检测精度
数据集 方法 ACC/% $ {F}_{m} $ Dataset#3 方法一 94.03 0.5015 方法二 94.36 0.5135 方法三 94.90 0.5188 本文方法 96.28 0.5498
下载: 导出CSV
-
[1] LONGBOTHAM N, PACIFICI F, GLENN T, et al. Multi-modal change detection, application to the detection of flooded areas: outcome of the 2009–2010 data fusion contest[J]. IEEE journal of selected topics in applied earth observations and remote sensing, 2012, 5(1): 331-342. DOI: 10.1109/JSTARS.2011.2179638 [2] 叶沅鑫, 孙苗苗, 王蒙蒙, 等. 结合邻域信息和结构特征的遥感影像变化检测[J]. 测绘学报, 2021, 50(10): 1349-1357. DOI: 10.11947/j.AGCS.2021.20200130 [3] AGRAWAL D G, JANGALE P M. Dynamic texture feature extraction using weber local descriptor[J/OL]. [2022-01-05]. Journal of engineering research and applications, 2014, 4(3): 502-506. http://ijera.com/papers/Vol4_issue3/Version%201/CL4301502506.pdf [4] TOUATI R, MIGNOTTE M, DAHMANE M. A new change detector in heterogeneous remote sensing imagery[C]//The Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA), 2017. DOI: 10.1109/IPTA.2017.8310138 [5] YANG X J, LO C P. Relative radiometric normalization performance for change detection from multi-date satellite images[J]. Photogrammetric engineering and remote sensing, 2000, 66(8): 967-980. DOI: 10.1080/01431160050030628 [6] ZONG K B, SOWMYA A, TRINDER J. Kernel partial least squares based hierarchical building change detection using high resolution aerial images and lidar data[C]//International Conference on Digital Image Computing: Techniques and Applications (DICTA), 2013. DOI: 10.1109/DICTA.2013.6691502 [7] TOUATI R, MIGONTTE M. A multidimensional scaling optimization and fusion approach for the unsupervised change detection problem in remote sensing images[C]//The Sixth International Conference on Image Processing Theory, Tools and Applications (IPTA), 2016. DOI: 10.1109/IPTA.2016.7821021 [8] TOUATI R, MIGNOTTE M. An energy-based model encoding nonlocal pairwise pixel interactions for multisensor change detection[J]. IEEE transactions on geoscience and remote sensing, 2018, 56(2): 1046-1058. DOI: 10.1109/TGRS.2017.2758359 [9] WEN X P, YANG X F. Change detection from remote sensing imageries using spectral change vector analysis[C]//2009 Asia-Pacific Conference on Information Processing, 2009. DOI: 10.1109/APCIP.2009.183 [10] SONG X L, CHENG B. Change detection using change vector analysis from Landsat TM images in Wuhan[J]. Procedia environmental sciences, 2011(11): 238-244. [11] HAO M, ZHOU M C, JIN J, et al. An advanced superpixel-based markov random field model for unsupervised change detection[J]. IEEE geoscience and remote sensing letters, 2020, 17(8): 1401-1405. DOI: 10.1109/LGRS.2019.2948660 [12] TOUATI R, MIGNOTTE M, DAHMANE M. Multimodal change detection in remote sensing images using an unsupervised pixel pairwise-based Markov random field model[J]. IEEE transactions on image processing, 2019(29): 757-767. DOI: 10.1109/TIP.2019.2933747 [13] HE P, SHI W, ZHANG H, et al. A novel dynamic threshold method for unsupervised change detection from remotely sensed images[J]. Remote sensing letters, 2014, 5(4): 396-403. DOI: 10.1080/2150704X.2014.912766 [14] 吴柯, 何坦, 杨叶涛. 基于混合像元分解与EM算法的中低分辨率遥感影像变化检测[J]. 武汉大学学报(信息科学版), 2019, 44(4): 555-562. [15] TOUATI R, MIGNOTTE M, DAHMANE M. A circular invariant convolution model-based mapping for multimodal change detection[J]. Advances in science, technology and engineering systems journal, 2020, 5(5): 1288-1298. DOI: 10.25046/AJ0505155 [16] LU D, BATISTELLA M, MORAN E. Multitemporal spectral mixture analysis for amazonian land-cover change detection[J]. Canadian journal of remote sensing, 2004, 30(1): 87-100. DOI: 10.5589/m03-055 [17] HOLLWEG J V, ISENBERG P A. Reflection of Alfvén waves in the corona and solar wind: an impulse function approach[J]. Journal of geophysical research:space physics, 2007, 112(A8): A08102-A08102-13. DOI: 10.1029/2007ja012253 [18] 李雪, 舒宁, 王琰, 等. 利用土地利用状态转移分析的变化检测[J]. 武汉大学学报(信息科学版), 2011, 36(8): 952-955. -
点击查看大图
图(3) / 表(5)
计量
- 文章访问数: 71
- HTML全文浏览量: 31
- PDF下载量: 14
- 被引次数: 0
