EN中文

地球信息科学学报 >

2018 , Vol. 20 >Issue 4: 440 - 451

DOI: https://doi.org/10.12082/dqxxkx.2018.170579

全国激光雷达大会特约稿件

适用于自然地表形变反演的小基线集方法

  • 黄俊松 , 1 ,
  • 曾琪明 , 1, * ,
  • 高胜 1, 2 ,
  • 焦健 1 ,
  • 胡乐银 3
展开
  • 1. 北京大学遥感与地理信息系统研究所,北京 100871
  • 2. 中国科学院电子学研究所,北京 100190
  • 3. 北京市地震局,北京 100080
*通讯作者：曾琪明(1964-),男,教授,研究方向为InSAR及其应用。E-mail:

作者简介：黄俊松(1990-),男,博士生,研究方向为InSAR时序分析。E-mail:

收稿日期: 2017-12-04

  要求修回日期: 2018-02-05

  网络出版日期: 2018-04-20

基金资助

国家重点研发计划（2017YFB0502703）

内蒙古自治区科技厅“数字化矿区资源管理与矿区生态环境监测技术与应用”项目（2015-2019）

收起

An Improved Small Baseline Subset Method for Deformation Retrieval of Natural Terrains

  • HUANG Junsong , 1 ,
  • ZENG Qiming , 1, * ,
  • GAO Sheng 1, 2 ,
  • JIAO Jian 1 ,
  • HU Leyin 3
Expand
  • 1. Institute of Remote Sensing and GIS, Peking University, Beijing 100871, China
  • 2. Institute of Electronics, Chinese Academy of Sciences, Beijing 100190, China
  • 3. Beijing Earthquake Agency, Beijing 100080, China
*Corresponding author: ZENG Qiming, E-mail:

Received date: 2017-12-04

  Request revised date: 2018-02-05

  Online published: 2018-04-20

Supported by

National Key R&D Program of China, No.2017YFB0502703

Digital Mining Resource Management and Mining Ecological Environment Monitoring Technology and Application Project of the Science and Technology Department of Inner Mongolia Autonomous Region (2015-2019).

Copyright

《地球信息科学学报》编辑部 所有
Fold

摘要

由于自然地表像元在长期观测中容易发生时空失相干,利用时序InSAR（Synthetic Aperture Radar Interferometry）技术对其开展形变监测会面临可用形变测量点不足的挑战。针对这一问题,提出一种改进的小基线（Small Baseline Subset,SBAS）方法。该方法改进了传统SBAS中初始高相干像元筛选及相位滤波过程：首先利用拟合优度检验,并结合相干性阈值条件来识别同质像元;然后根据同质像元数量将所有像元分成2部分,即PS（Persistent Scatterers）候选点和DS（Distributed Scatterers）候选点;其次分别在这两部分像元中开展初始高相干PS点及DS点筛选;最后对选出的高相干PS点及DS点进行加权相位滤波。利用覆盖北京平原区西北部（含城区及山区）的27景ENVISAT ASAR影像开展的形变监测实验表明：与2个参考方法相比,该方法能够有效扩展形变结果上的测量点数量和覆盖范围,测量点数量分别提高了22.6%及27.6%,且自然地表的形变测量点密度得到了明显提升。同时,研究区形变结果与4个连续GPS站的位移数据有很好的一致性,证明了该方法在地表形变反演中的有效性及优越性。

关键词： SBAS; 形变; 时序InSAR; PS-InSAR; PS; DS

本文引用格式

黄俊松 , 曾琪明 , 高胜 , 焦健 , 胡乐银 . 适用于自然地表形变反演的小基线集方法[J]. 地球信息科学学报, 2018 , 20(4) : 440 -451 . DOI: 10.12082/dqxxkx.2018.170579

Abstract

Owning to that the pixels in natural terrains are prone to spatial-temporal decorrelation during the long-term observation, using time-series InSAR (Synthetic Aperture Interferometry) technique to carry out deformation monitoring of natural terrains will face the challenge of lacking of available deformation measurement points. To solve this problem, an improved Small Baseline Subset (SBAS) method is proposed. It improves the selection process of initial high coherent pixels and phase filtering in conventional SBAS. Firstly, it uses the goodness of fit and the coherence threshold condition to identify statistically homogeneous pixels (SHP). After this, all pixels are divided into two parts base on the number of SHP, i.e. Persistent Scatterers (PS) candidates and Distributed Scatterers (DS) candidates. Then, initial high coherent PS and DS are selected from these two parts respectively. Finally those selected high coherent PS and DS are filtered by a weighted phase filter. The deformation monitoring experiment with 27 ENVISAT ASAR images, acquired over the northwest part of Beijing plain shows that: compared with StaMPS-PS (refers to the PS-InSAR in StaMPS) method and StaMPS-SBAS (refers to the SBAS in StaMPS) method, the improved method can effectively extend the quantity and coverage of deformation measurement points. The quantity of measurement points is increased by 22.6% and 27.6% respectively, and the deformation result of natural terrains is improved effectively. The deformation result of this study area is in good agreement with the displacement of 4 continuous GPS stations. Experimental results prove the effectiveness and superiority of this method in the inversion of ground deformation.

Key words： SBAS; deformation; time-series InSAR; PS-InSAR; PS; DS

1 引言

在时序InSAR中,主要选择相干目标进行分析建模。相干目标是指在长时间序列观测中能够保持稳定相干的目标（通常用相位稳定刻画）。根据目标散射类型,以及其尺寸之于图像分辨率的比例的不同,可将地表相干目标分为点目标及分布式目标2类,二者在长期观测中保持高相干的能力不同。前者占据了分辨单元的散射主导,在长期观测中能保持强而稳定的散射特性,从而保持高相干,是时序InSAR中良好的高相干目标源。根据其特性,点目标也称为永久散射体点,即PS点。人造建筑如房屋、桥梁,地表裸露的岩石,角反射器等是时序InSAR中良好的PS点源。而分布式目标即DS点,其分辨单元内并无主导散射体,在多时相SAR重复观测中,其回波信号（通过复高斯统计模型刻画[1]）的随机性较强,容易发生失相干现象,通常成片分布,低植被覆盖区（如草地）、沙漠、洪积扇等是典型的DS点源。
对自然地表而言,容易失相干的DS点在地表像元中占据绝大多数,而稳定高相干的PS点却极少。自然地表因易失相干造成可用高相干像元的减少,对时序InSAR地表形变监测形成了挑战,失相干严重者甚至会导致无法有效获取形变结果[2]。空间上可用高相干像元的减少主要会带来3方面问题：① 解缠精度相对不可靠;② 通过滤波分离的大气相位相对不可靠;③ 形变结果的空间分辨率低,精度受限[3]。为尽可能提高地表可用高相干像元点数量,SBAS方法通过构建小基线干涉网络来降低像元的失相干程度,并对干涉图进行多视处理以进一步提高像元信噪比,使部分原本低相干的像元被有效利用,较适用于自然地表的形变反演[4,5],特别是在SAR数据较少,时间跨度（Time interval）和时间间隔（Time gap）较大情况下比较适用[6]。干涉图多视处理的缺点是会降低图像分辨率,使原本孤立的高相干PS点被模糊而无法被有效识别,Hooper[7]和Lanari等[8]提出了可在全分辨率条件处理高相干点的SBAS方法[7,8]。龚文瑜等[6]对各类SBAS方法在不同地表环境,不同干涉对数量等条件下的形变监测结果进行了实验评价。
在众多SBAS方法中,高相干像元选取都是一个十分重要的过程,筛选的高相干像元数量及质量对形变结果上测量点覆盖范围（包括数量及密度）和精度都有重要影响。为针对不同地理区域选择合适的高相干像元选取方法,范锐彦等[9]对各类高相干像元选取方法的特点、适用性等进行了对比总结。在包括但不限于文献[9]所列方法中,基本都是采用单阈值的高相干像元选取方法,如时序平均相干性法[10,11,12],幅度差离差（Dispersion of Amplitude Difference,DAD）法[7,13]等,而单阈值方法在实际操作过程中存在一定的缺点,体现在阈值的取值通常偏松[13,14],这虽避免了部分高相干像元的漏选,但同时大量相对低质量的像元也会被选择,降低了时序分析的处理效率。此外,大量低质量的像元也会对高质量的像元在解缠等过程中产生不利影响[15]。比较理想的筛选办法应是能够精准地筛选高相干像元：既要避免漏选高相干像元,同时不必要的低质量像元被排除（即不会被选择）。基于以上,本文对SBAS方法中高相干像元筛选方法进行改进,提出了一种类似于双阈值的筛选方法,该方法借鉴了Ferretti的SqueeSARTM思想[16],但在干涉网络构建,相位滤波,形变反演及整个数据处理流程上则与其完全不同。本改进方法首先利用拟合优度检验,并结合相干性条件来识别同质像元,之后根据同质像元数量（定义见2.1节所述）将所有像元分成2部分,再在这2部分像元中分别开展初始高相干PS点及DS点筛选,最后利用一个加权的相位滤波对选出的高相干PS点及DS点滤波。该改进方法的特点是在高相干像元筛选过程中给像元另赋予了一个维度的信息—同质像元数量信息,另一特点是该加权相位滤波是在同质像元之间进行的。
在本方法的实现过程中,选择开源时序分析工具包StaMPS中的SBAS方法（以下记为StaMPS-SBAS）作为本文的改进和参考对象。这是因为StaMPS-SBAS是一个具有广泛代表性、广泛用户基础的SBAS方法,被认为在自然地表形变反演中性能优异,因此对其进行改进并作为参考方法具有重要的参考价值和示范意义。此外,考虑到实验区除自然地表外（如山区）还存在城区地表,为增强实验的说服力,本文还选用StaMPS中的PS-InSAR方法（以下记为StaMPS-PS）作为参考方法,这是因为StaMPS-PS是基于公共主图像构建干涉网络（与SBAS不同）,不仅适宜处理城区形变反演,在处理自然地表形变反演方面也被认为较优,具有同样广泛的用户基础和代表性。综上,为验证该改进方法的有效性和优越性,实验利用本改进方法及2个参考方法—StaMPS-PS及未做改进的StaMPS-SBAS,对2007-2010年覆盖北京平原区西北部（包含城区及山区）的27景ENVISAT ASAR数据进行了处理,获取了该研究区的形变结果并对形变结果进行了对比和统计分析,最后用4个位于该地区的连续GPS观测站位移数据对形变结果进行了验证。

2 改进的SBAS方法

图1是本文改进的SBAS方法的核心数据处理流程,其中改进步骤包括同质像元识别和高相干像元选取及相位滤波（图1中深色方框所示）。同质像元识别是高相干像元选取及相位滤波的基础,目的是为了更精准地筛选高相干像元,以及在同质像元间进行相位滤波。筛选出来的高相干像元在滤波后被导入到SBAS反演模型中进行形变求解。
Fig. 1 The processing flow of the proposed method

图1 本方法数据处理流程

2.1 同质像元识别

同质像元（Statistically Homogeneous Pixels,SHP）概念是由Ferretti首次提出,它是基于像元间的统计特征而定义的[16]。时间序列上具有相同或相似散射行为,通常由同一类地物构成的像元间可以互称同质像元。在本方法中采用鲁棒性较好的非参数化拟合优度检验—Kolmogorov-Smirnov （K-S）检验[17]来识别同质像元：如果2个像元的时间序列幅度Mp1与Mp2（式（1））被K-S检测为符合相同统计分布,则二者是同质像元,否则不是。式（1）中,mp1,1与mp2,1分别为像元p1和p2在第一景SAR图像上的幅度,其他以此类推。
M p 1 = m | m = m p 2,1 , m p 2,2 , ⋯ , m p 2 , N M p 2 = m | m = m p 2,1 , m p 2,2 , ⋯ , m p 2 , N （1）
同质像元识别过程如下：
（1）对实验中所有要使用的SAR图像进行定标、配准,并计算每个像元的时序平均相干性。
（2）依据植被覆盖区的像元时序平均相干性设定一个相干性阈值;对时序平均相干性低于该阈值的像元不进行以下步骤。
（3）对时序平均相干性高于该相干性阈值的每一个像元P1,以P1为中心选取一定大小的窗口,开展同质像元识别：如果中心像元P1的时间序列幅度分布与窗口中某一像元M的时间序列幅度被K-S检测判定为符合相同统计分布,则像元M是中心像元P1的同质像元;继续检测窗口内的其他像元,直到窗口内每个像元都被遍历过,若窗口内有N个像元被检测为是中心像元P1的同质像元,则记中心像元P1的同质像元数量为N。
（4）将窗口滑动到下一个像元P2并以其为中心,重复步骤（3）完成像元P2的同质像元检测过程;直到所有时序平均相干性高于相干性阈值的像元都被遍历过,则完成了同质像元的识别过程。
在上述同质像元识别过程中,设定相干性阈值的目的是为了在基本不影响最终形变结果的前 提下,对低于相干性阈值的像元略过上述同质像 元识别过程,这可以大大提高数据处理效率。因为,3.3节将证明（前人实验也早已证明）,在特别 低相干的地区,几乎所有的像元点都会在时序分析过程中被剔除,对形变测量点数量的提高基本无 贡献。

2.2 高相干像元选取及相位滤波

高相干像元选取的示意图如图2所示,图中D表示每个像元的同质像元数量,Dthr表示实验确定的同质像元数量阈值。如图2所示,到选出高相干PS点和高相干DS点经过了一个2层“过滤”过程：第1层是同质像元数量的筛选;第2层是DA（Dispersion of Amplitude, DA）[18]指数或DAD指数的筛选。这样做的好处是可以缩小筛选点的空间范围,精准识别最可靠的高相干像元点,效果类似于双阈值筛选高相干点[11,19];而传统高相干PS点或高相干DS点的选取只经历一层“过滤”即从所有像元中直接利用DA指数或DAD指数筛选高相干PS点或DS点,其优点是简单方便,但不利于精准选点,会有很多无用点被选取,而这些点在时序分析中基本都会被淘汰,对结果基本无贡献却会导致处理效率降低[13,14]。
Fig. 2 Selection process of high coherent pixels

图2 高相干像元选取流程

高相干像元选取完成后,需要对高相干DS及高相干PS进行相位滤波以去除相位噪声。考虑到滤波窗口内像元的异质性会对中心像元的相位估计会产生不利影响,本文采用基于同质像元的加权相位滤波,如式（2）所示。该滤波是在同质像元间进行的,也即窗口内参与滤波的像元都是中心像元的同质像元,保证了滤波像元的同质性,有利于噪声的估计[20]。
I ^ p = ∑ j = 1 M P KS ( D p , j ) ⋅ I j ∑ j = 1 M P KS ( D p , j ) Q ^ p = ∑ j = 1 M P KS ( D p , j ) ⋅ Q j ∑ j = 1 M P KS ( D p , j ) （2）
式中： I ^ p 及 Q ^ p 为中心像元p的相位的虚部及实部估计值;M为中心像元的同质像元数量,PKS(Dp,j)为中心像元p与像元j的K-S检测的输出概率,作为滤波的权重。滤波后的实部及虚部需要重新组合成相位。

3 实验分析

3.1 研究区概括及实验数据

图3中黑色方框所示即为本研究区范围,大小约63 km（南北）×46 km（东西）。在研究区分布4个代号分别为CHAO,DSQI,BJSH,CHPN的连续GPS观测站,用来验证形变结果。为便于后续分析,将研究区划大体划分为3类地表：中心城区,郊区以及山区,中心城区的建设趋于稳定,建筑物密集;郊区仍然存在大量的建筑工地、空地及有植被覆盖区等,建筑物相对稀疏;山区除冬天以外其他时间一直有植被覆盖。研究区地处华北平原,地表水资源较为贫乏,早期城市建设、农业灌溉等主要依靠开采地下水资源,随着地下水开采量的增加,地面沉降比较严重[21],近些年由于中心城区限制地下水开采并进行地下水回填,有趋于稳定的趋势[22]。
Fig. 3 The location of study area

图3 研究区地理位置

实验获取了2007年1月到2010年10月覆盖该研究区的27景ENVISAT ASAR 降轨数据,数据信息如表1所示。在参考方法StaMPS-PS中,根据使所构成的干涉图总体失相干程度最小原则[23],2009年8月24日获取的SAR图像被选为公共主图像,其余各从图像与主图像之间的垂直基线见表1。而在本改进方法及StaMPS-SBAS方法中,形变反演模型采用的是SBAS模型,因此需要先通过小基线原则构建干涉网络,设定最小相干性为0.5,最大时间基线为1000 d,最大垂直基线为500 m,27景SAR数据共构建了54个干涉对如图4所示,除图中序号为1的干涉对垂直基线为323 m外,其他干涉对的垂直基线均处于200 m以内,有利于避免解缠等误差的传递[24]。
Tab. 1 The information of interferograms in StaMPS-PS

表1 StaMPS-PS中干涉配对信息

序号 获取时间 垂直基线 B⊥/m 序号 获取时间 垂直基线B⊥/m
1 2007-01-22 586 15 2009-05-11 -165
2 2007-02-26 -46 16 2009-08-24 0
3 2007-05-07 -195 17 2009-09-28 556
4 2007-06-11 -237 18 2009-11-02 -205
5 2007-07-16 -107 19 2010-01-11 -252
6 2007-12-03 263 20 2010-02-15 423
7 2008-01-07 -524 21 2010-05-22 61
8 2008-02-11 274 22 2010-04-26 216
9 2008-03-17 -235 23 2010-05-31 -96
10 2008-01-30 -67 24 2010-07-05 85
11 2008-08-04 15 25 2010-08-09 -370
12 2008-09-08 239 26 2010-09-13 80
13 2008-10-13 -200 27 2010-10-18 329
14 2009-04-06 462 - - -
Fig. 4 The spatial-temporal baselines of interferometric network in SBAS

图4 SBAS干涉网络时空基线

3.2 数据处理与分析

StaMPS-PS和StaMPS-SBAS参考方法的数据处理流程在StaMPS手册上有详细介绍,读者可以参考手册[13],在此不赘述。本方法的数据处理流程遵循图1,需要特别说明：由于本论文旨在证明改进方法在地表形变反演中优于参考方法,因此本实验在同质像元识别（即2.1节）过程中并未设置相干性阈值以提高数据处理效率,而是对所有像元都开展了同质像元识别,目的是为了尽可能多增加形变结果中可用的测量点（即使增加的数量非常有限）。
根据2.1节,同质像元识别采用的是K-S检验,涉及到显著性水平α及窗口大小的确定。由于实验区存在地形起伏的山区,因而窗口大小取值应适中。在兼顾SAR图像分辨率（4.5 m×7.8 m）以及数据处理效率情况下,将窗口大小确定为25×9,对应地面尺寸约为100 m × 100 m,该尺度与研究区地表覆盖基本单元尺度基本相当。而显著性水平α则测试了0.05,0.15,0.25,0.45,0.65,0.70,0.80,0.90等覆盖低中高区间的取值,α越大表明2个像元互为同质像元的条件越严,反之则越松。表2为不同α条件下,所有像元的同质像元数量的统计情况。通过分析表2,当α取值0.65和0.70及以上时像元的平均同质像元数量分别仅为36.00和7.52个,数量偏少,而α取值0.25及以下时,检测条件偏松,容易将很多非同类像元检测为同质像元。实验最终将α确定为0.45,保证了较严格的同质像元判别条件,此时像元的平均同质像元数量（74.31个）比较适中,同时地物间的区分度较好与地表有比较明显的对应,各像元的同质像元数量分布如图5所示。
Fig. 5 The quantitative distribution of statistically homogenous pixels identified by K-S in the radar coordinates when the value of α is 0.45

图5 雷达坐标系下,α取值为0.45时K-S检验识别出来的同质像元数量分布图

Tab. 2 The average value and the maximum value of the number of statistically homogenous pixels for all pixels under different confidence level α

表2 不同α条件下,所有像元的同质像元数量的平均值及最大值统计情况

α 平均值 最大值
0.05 151.58 225
0.15 132.26 225
0.25 107.00 225
0.45 74.31 224
0.65 36.00 208
0.70 7.52 166
0.80 7.52 166
0.90 7.52 166
图5是α取值0.45时,K-S检验识别出来的同质像元数量分布图。图5中,像元的灰度深浅表示该像元所拥有的同质像元数量,左下角颜色最深处为北京市中心城区,中心城区建筑物密集,是良好的点状散射体源,点状散射体同质像元数量一般较少,大部分像元的同质像元数量在50以下;而中间颜色较浅处为北京市郊区,郊区建筑物稀疏,包含大量未开发的土地、草地等,是良好的面状散射体源,面状散射体的同质像元数量一般较多,大部分像元的同质像元数量在150-200之间。
同质像元识别完成后,所有像元被赋予了一个新的“维度”信息-同质像元数量。按照流程图2所示,高相干像元选取之前需要确定一个同质像元数量阈值Dthr,根据这个阈值将所有像元分成2个点集：PS候选点及DS候选点。在确定该阈值之前,需要说明几点：
（1）由于DS点通常是成片分布,例如分布在未开垦过的裸地、草地、洪积扇等大面积的地表,因此DS像元的同质像元数量较多;而PS点由于其自身较强的散射特性,如单幢建筑可以认为是一个PS点,显然这类强散射的PS点所拥有的同质像元数量较少。
（2）根据幅度离差指数DA的定义（如式（3）所示）及Agram[25]的研究表明：由于PS点是永久散射体点,散射强烈而稳定,因此PS点的DA值相对于DS点而言,较小;当DA取0.25或以下的值时,所筛选的像元越可能是真正的PS点,含少量DS点;随着DA增大,越来越多的DS点会混入到所筛选的像元中。
DA=σA/mA （3）
式中：σA为像元的时间序列幅度标准差;mA为像元的时间序列幅度平均值[18]。
在说明以上情况后,同质像元数量阈值Dthr的确定过程为：将DA分别设置成0.25,0.30,0.35,0.40从所有像元中筛选满足条件的像元,并统计这些在不同DA值条件下筛选出来的像元的同质像元数量直方图。如图6所示,其中横坐标为同质像元数量,纵坐标为像元数的百分比。由于当DA较小时筛选出来的像元才越可能是真正的PS点,且其同质像元数量不可能很多,随着DA增大,DS会混入这些真正PS点中,导致直方图发生从图6（a）到图6（d）的变化过程。因此观察图6（a）及图6（b）（对应DA分别为0.25及0.30）,可知大多数PS点（由于此时DA值较小,认为筛选出来的像元大部分为PS点）所拥有的同质像元数量处于10以下,而图6（c）及图6（d）中随着DA值增大（对应DA分别为0.35及0.40）,越来越多的DS点也被选入,导致越来越多的像元拥有比10更多的同质像元数量,如20、50、100。基于此,为区分PS候选点和DS候选点,将Dthr取值为15,即当某像元的同质像元数量大于15时,其被划分为DS候选点,而当某像元的同质像元数量小于15时其被划分为PS候选点。
Fig. 6 The histogram of the number of statistically homogenous pixels for those pixels selected by different DA values

图6 不同DA条件下筛选出来的像元的同质像元数量分布直方图注：图6（a）-（d）,DA依次取值0.25, 0.30, 0.35, 0.40

在确定Dthr之后,根据流程图2,取DA为0.35从PS候选点中筛选高相干PS点,取DAD指数为0.52从DS候选点中筛选高相干DS点。筛选出来的高相干PS点及DS点利用式（2）进行加权相位滤波后被用于研究区地表形变估计。

3.3 结果分析与验证

除特别提到的参数外,其他参数都选用StaMPS中的默认处理参数或一致的处理参数。时序分析中参数的最优化问题不是本文讨论的主题。下文着重分析的内容主要包括：① 在不同相干性条件下3个形变结果中测量点数量、密度的对比分析;② 3个形变结果在相同测量点处的形变速率对比及分析;③ 3个结果与GPS数据的定量对比分析。
3.3.1 形变结果分析
基于图1所示数据处理流程,将残余轨道误差、大气延迟相位等去除后,2个参考方法及本方法的最终LOS（Line of Sight,LOS）向形变速率结果如 图7所示,参考区域统一为整个区域,参考区域内速率平均值为0。3个方法在初始的高相干像元数量、最终形变测量点数量、淘汰率及测量点密度上的统计情况如表3所示。在不考虑形变测量点精度条件下,本方法在形变测量点密度上占优,可达到39.60 km-1,超过了StaMPS-SBAS方法与StaMPS-PS方法的31.04 km-1及32.31 km-1,而图7则可以说明本方法在自然地表（如图7中红色椭圆所示山区）的形变测量点数量更多。综上,综合初选点数量、淘汰率、形变测量点密度及在诸如山区等自然地表的测量点数量可知,本方法优于2个参考方法。
Tab. 3 Comparison of the three methods in terms of the quantity of initial high coherent pixels, the quantity and the density of deformation measurement points

表3 3个方法在初始高相干像元数量、形变测量点数量及密度方面的对比

初始选取的高相干像元数量及对应相干性 形变测量点数量及对应相干性 淘汰率/% 形变测量点密度/(个/km2)
StaMPS-PS 354 116/0.31 93 646/0.39 73.6 32.31
StaMPS-SBAS 835 977/0.30 89 967/0.32 89.1 31.04
本方法 824 972/0.36 114 765/0.30 86.1 39.60
Fig. 7 The LOS directional deformation velocity in geographical coordinate system acquired by the three methods

图7 地理坐标系下3个方法的LOS向形变速率

由于相干性的高低与地表覆盖相关,为检测3个方法在自然地表等低相干情况下的形变反演能力,统计了图7结果中形变测量点在不同相干性条件下的数量分布,如表4所示。经统计发现相干性低于0.11时地表基本为植被,是典型的自然地表。表4中3类低相干条件下,本方法的形变测量点密度均最高,但优势有所不同：在时序平均相干性低于0.20及低于0.30的地表,本方法优势十分明显,形变测量点密度最大可达到StaMPS-SBAS方法密度的2倍,达到StaMPS-PS的7倍;而在相干性低于0.11时本方法虽有优势,但形变测量点密度都很小,最大值是本方法的3.61 km-1,这表明在特别低相干条件下（例如有植被覆盖）形变反演确实较难。回顾2.1节,为什么相干性低于一定阈值的像元没必要开展同质像元识别呢?以本实验区为例,时序平均相干性低于0.11的像元数占图像总像元的39.99%,但表4说明在特别低相干地区,如相干性低于0.11的地区,时序分析后保留下来的像元十分稀少,只占所有被保留像元的3.64%（不足4178个）,占图像总像元的0.013%。换言之,几乎所有时序平均相干性低于0.11的像元都在时序分析过程中都被剔除掉了,因此在同质像元识别过程中略过时序平均相干性低于0.11的像元能够提高约40%的处理效率,而所带来的最终形变测量点的减少的损失却是可忽略的。当然,在同质像元识别中,是否需要略过特别低相干的像元应该根据实验目的和应用需求而定。
Tab. 4 Comparison of the three methods in terms of the quantity and density of deformation measurement points in different coherence situation

表4 不同相干性情况下3个方法测量点数量及密度对比

形变测量
点总数		 时序平均相干性低于0.11的像元 时序平均相干性低于0.20的像元 时序平均相干性低于0.30的像元
占比/% 平均相干性 密度/
(个/km2)		 占比/% 平均相干性 密度/
(个/km2)		 占比/% 平均相干性 密度/
(个/km2)
StaMPS-PS 93 646 0.29 0.10 0.24 5.00 0.16 2.23 24.95 0.24 9.18
StaMPS-SBAS 89 967 2.95 0.10 2.29 18.03 0.15 7.73 45.02 0.21 16.58
本方法 114 765 3.64 0.10 3.61 27.10 0.15 14.81 53.71 0.20 25.22
在表3中,时序分析淘汰部分低质量像元后,两个参考方法的形变测量点的平均相干性都提高了,分别从0.31-0.39以及从0.30-0.32,而本方法的形变测量点的平均相干性却相较于初始高相干的相干性降低了,从0.36-0.30。这说明：① 本方法相较于参考方法更能够精准的选择高质量的像元点,初始选择的高相干像元相干性能达到0.36,质量最好,高于2个参考方法的0.31和0.30;② 本方法时序分析后保留了较多的存在于低相干地区的形变测量点,使得整体相干性降低,较适合低相干条件下的形变反演。本方法时序分析后形变测量点的平均相干性反而降低的原因可从表4中找到答案：表4中,本方法在相干性低于0.30时,形变测量点数量占比为53.71%,且平均相干性只有0.20,对应的数据在两个参考方法中分别为45.02%和0.21以及24.95%和0.24,这说明在2个参考方法中相干性大于0.30的像元数量占据多数（会拉高最后的相干性水平）,特别是在StaMPS-PS方法中,而在本方法中相干性大于0.30的像元占据少数,因此时序分析后形变测量点的平均相干性反而降低。综上,本方法时序分析过程中保留低相干像元点的能力大于StaMPS-SBAS方法,再大于StaMPS-PS方法。
3.3.2 形变结果对比与验证
从图7可看出,3个方法反演的形变速率趋势、特点及最大最小速率也基本一致,分别为-56.6~26.6 mm/yr,-66.9~24.5 mm/yr,-66.4~27.1 mm/yr （单位mm/yr,表示毫米每年,下同）。图8左边为3个方法在相同形变测量点处的速率差值图,右边为对应的速率差的统计直方图。从图8可知,图8（a）所示StaMPS-PS与StaMPS-SBAS之间的速率差异最大,速率差的标准差达到8.40 mm/yr,相比较之下,本方法与2个参考方法的速率差异（即图8（c）及8（e）所示）与之相比都小,当然从图8（b）、（d）,（f）可知,绝大部分点的速率差异在-10~10 mm/yr之间,速率差的平均值最大不超过0.53 mm/yr,速率差的标准差则不超过8.40 mm/yr,吻合很好。这反映本方法与2个成熟的参考方法对相同位置处的形变测量点的速率估计差异非常小,3个结果具有很好的一致性,本方法与2个参考方法享有同等的可靠性。
Fig.8 Comparisons of velocity differences between the three methods in the common measurement points

图8 3个方法相同测量点处的速率差值图及其直方图统计
注：（a）与（b）为StaMPS-PS与StaMPS-SBAS的差;（c）与（d）为StaMPS-PS与本方法的差;（e）与（f）为StaMPS-SBAS与本方法的差;（b）（d）（f）分别为（a）（c）（e）对应的直方图

最终形变结果的验证通过2种渠道：① 与已发表的文献进行对比;② 与连续GPS观测站的位移数据对比。从已发表的文献看,该研究区2007-2010年的形变特点为：中心城区相对稳定而沉降区域呈现出从东边到西北边环绕北京的特征[22,26-28],该趋势与图7结果一致。将GPS连续观测数据投影到LOS向后,取GPS位置邻近范围（约200 m）的形变测量点的相对位移的平均值,将该值与连续GPS站的相对位移进行对比。如图9所示,其中形变测量点与GPS观测站的时间参考都统一到了二者的相同时刻;同时,本方法及2个参考方法的测量点的相对位移与GPS站相对位移数据一致性较好,具体统计结果见表5。在4个连续GPS观测站处,本方法的形变测量点相对位移与GPS相对位移吻合较好,最大相对位移差的平均值不超过13.31 mm,标准差不超过9.21 mm,这与其他2个参考方法的结果互有优劣,整体差别不大,证明结果精度较高。
Tab. 5 The statistical results of the difference between the relative displacement of GPS and the relative displacements of the deformation measurement points at the corresponding positions of the three methods

表5 GPS相对位移与3个方法中对应位置处的形变测量点相对位移的差值的统计结果（mm）

BJSH CHAO CHPN DSQI
位移差平均值 位移差标准差 位移差平均值 位移差标准差 位移差平均值 位移差标准差 位移差平均值 位移差标准差
StaMPS-PS 2.43 3.34 8.36 6.69 2.38 1.42 15.76 10.33
StaMPS-SBAS 3.57 4.42 5.32 5.36 2.27 1.31 8.70 6.81
本方法 3.31 4.05 5.67 5.52 2.75 1.51 13.31 9.21
Fig. 9 Comparisons of relative displacement along the LOS direction between InSAR and the four GPS

图9 3个方法LOS向位移与GPS相对位移对比情况
注：图中CHAO、CHPN只覆盖部分时序InSAR时间段（GPS采样时间间隔为1 d）

4 结论

本文以保证地表形变监测精度为前提,以提高地表特别是自然地表的测量点数量和密度为主要研究目标,对StaMPS-SBAS方法中的高相干像元选择方法进行改进。利用2007-2010年覆盖北京中心城区、郊区及山区的27景ASAR数据,开展了本改进方法及2个参考方法StaMPS-PS及StaMPS-SB的形变监测实验,获取了该地区的LOS向形变速率结果,并将该结果与4个连续GPS站位移数据进行了对比分析。实验得到以下结论：
（1）相较于2个参考方法,本改进的SBAS方法更能有效选择高质量的像元点,并且能够提高最终形变结果上的测量点数量,测量点数量分别提高了22.6%及27.6%,同时在自然地表等低相干地区适用性最好,识别的目标点可达到参考方法的数倍,有利于山区滑坡等形变监测应用。
（2）随着相干性从低到高（从0.11-0.30）,本改进的SBAS方法相对于参考方法的优势不一,在低相干地区优势更明显,而对特别低相干地区（如本实验中相干性低于0.11的植被覆盖区）的形变反演则还存在一定难度。
（3）本改进的SBAS方法与两参考方法的形变结果趋势一致,与GPS位移数据吻合也较好,证明本方法结果可靠,有效。
（4）在同质像元识别过程中通过设定一相干性阈值,低于该阈值像元略过同质像元识别过程可以提高同质像元识别效率,而所带来测量点的损失是可忽略的。
（5）同质像元数量阈值Dthr的大小与具体的研究区、图像分辨率等相关,需要通过实验确定。

The authors have declared that no competing interests exist.

参考文献

原文顺序 | 文献年度倒序 | 文中引用次数倒序
[1]
Hanssen R F. Radar interferometry data interpretation and error analysis[J]. Journal of the Graduate School of the Chinese Academy of Sciences, 2001,2(1):V5-577-V5-580.This book is the product of five and a half years of research dedicated to the und- standing of radar interferometry, a relatively new space-geodetic technique for m- suring the earth's topography and its deformation. The main reason for undertaking this work, early 1995, was the fact that this technique proved to be extremely useful for wide-scale, fine-resolution deformation measurements. Especially the interf- ometric products from the ERS-1 satellite provided beautiful first results--several interferometric images appeared as highlights on the cover of journals such as Nature and Science. Accuracies of a few millimeters in the radar line of sight were claimed in semi-continuous image data acquired globally, irrespective of cloud cover or solar illumination. Unfortunately, because of the relative lack of supportive observations at these resolutions and accuracies, validation of the precision and reliability of the results remained an issue of concern. From a geodetic point of view, several survey techniques are commonly available to measure a specific geophysical phenomenon. To make an optimal choice between these techniques it is important to have a uniform and quantitative approach for describing the errors and how these errors propagate to the estimated parameters. In this context, the research described in this book was initiated. It describes issues involved with different types of errors, induced by the sensor, the data processing, satellite positioning accuracy, atmospheric propagation, and scattering character- tics. Nevertheless, as the first item in the subtitle "Data Interpretation and Error Analysis" suggests, data interpretation is not always straightforward.

DOI

[2]
Hooper A, Zebker H, Segall P, et al.A new method for measuring deformation on volcanoes and other natural terrains using InSAR persistent scatterers[J]. Geophysical Research Letters, 2004,31(23):1-5.A new InSAR Persistent Scatterers method is presented for analyzing episodic crustal deformation in non-urban environments, with application to volcanic settings.

DOI

[3]
高胜,曾琪明,焦健,等.永久散射体雷达干涉研究综述[J].遥感技术与应用,2016,31(1):86-94.lt;p>传统D\|InSAR技术虽然在地表形变测量应用中得到广泛应用,但其测量精度容易受失相干、大气效应、DEM误差等的影响,测量结果有时甚至不可靠。为了突破这些限制,在D\|InSAR技术中引入时间维,筛选出时序SAR图像上受失相干影响较小的像元,建立并解算相位模型,从而提取形变信息。PS\|InSAR技术自出现以来,经过不断改进,在受失相干影响较大的长期缓慢形变监测中发挥了重要作用,一直是近年来雷达干涉领域研究的重点问题。详细梳理了PS\|InSAR技术的发展历史,总结了关键技术环节的改进算法以及应用领域,指出现有PS\|InSAR技术仍然存在的问题以及可能的发展趋势。</p>

DOI

[ Gao S, Zeng Q, Jiao J, et al.A review on persistent scatterer interferometric synthetic aperture radar[J]. Remote Sensing Technology & Application, 2016,31(1):86-94. ]

[4]
Berardino P, Fornaro G, Lanari R, et al.A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms[J]. IEEE Transactions on Geoscience and Remote Sensing, 2002,40(11):2375-2383.

[5]
Schmidt D, Bürgmann R. Time-dependent land uplift and subsidence in the Santa Clara Valley, California[J]. Journal of Geophysical Research Solid Earth, 2003,108(B9): ETG 4-1.We invert 115 differential interferograms derived from 47 synthetic aperture radar (SAR) scenes for a time-dependent deformation signal in the Santa Clara Valley, California. The time-dependent deformation is calculated by performing a linear inversion that solves for the incremental range change between SAR scene acquisitions. A nonlinear range-change signal is extracted from the ERS InSAR data without imposing a model of the expected deformation. In the Santa Clara valley, cumulative land uplift is observed during the period from 1992 to 2000 with a maximum uplift of 41+/-18 mm centered north of Sunnyvale. Uplift is also observed east of San Jose. Seasonal uplift and subsidence dominates west of the Silver Creek fault near San Jose with a maximum peak-to-trough amplitude of 35 mm. The long-term uplift is attributed to the gradual recharge of the aquifer over the 8-year period. It is assumed that the pumpage and redistribution of groundwater is responsible for the seasonal subsidence, although a direct correlation is still under investigation. The pattern of seasonal versus long-term uplift provides a qualitative constraint on the spatial and temporal characteristics of water bearing units within the aquifer. The Silver Creek fault partitions the uplift behavior of the basin, suggesting that it acts as a hydrologic barrier to ground water flow. While no tectonic creep is observed along the fault, the development of a low permeability barrier that bisects the alluvium suggests that the fault has been active since the deposition of Quaternary units.

DOI

[6]
Gong W, Thiele A, Hinz S, et al.Comparison of small baseline interferometric SAR processors for estimating ground deformation[J]. Remote Sensing, 2016,8(4):330.The small Baseline Synthetic Aperture Radar (SAR) Interferometry (SBI) technique has been widely and successfully applied in various ground deformation monitoring applications. Over the last decade, a variety of SBI algorithms have been developed based on the same fundamental concepts. Recently developed SBI toolboxes provide an open environment for researchers to apply different SBI methods for various purposes. However, there has been no thorough discussion that compares the particular characteristics of different SBI methods and their corresponding performance in ground deformation reconstruction. Thus, two SBI toolboxes that implement a total of four SBI algorithms were selected for comparison. This study discusses and summarizes the main differences, pros and cons of these four SBI implementations, which could help users to choose a suitable SBI method for their specific application. The study focuses on exploring the suitability of each SBI module under various data set conditions, including small/large number of interferograms, the presence or absence of larger time gaps, urban/vegetation ground coverage, and temporally regular/irregular ground displacement with multiple spatial scales. Within this paper we discuss the corresponding theoretical background of each SBI method. We present a performance analysis of these SBI modules based on two real data sets characterized by different environmental and surface deformation conditions. The study shows that all four SBI processors are capable of generating similar ground deformation results when the data set has sufficient temporal sampling and a stable ground backscatter mechanism like urban area. Strengths and limitations of different SBI processors were analyzed based on data set configuration and environmental conditions and are summarized in this paper to guide future users of SBI techniques.

DOI

[7]
Hooper A.A multi-temporal InSAR method incorporating both persistent scatterer and small baseline approaches[J]. Geophysical Research Letters, 2008,35(16):96-106.Synthetic aperture radar (SAR) interferometry is a technique that provides high-resolution measurements of the ground displacement associated with many geophysical processes. Advanced techniques involving the simultaneous processing of multiple SAR acquisitions in time increase the number of locations where a deformation signal can be extracted and reduce associated error. Currently there are two broad categories of algorithms for processing multiple acquisitions, persistent scatterer and small baseline methods, which are optimized for different models of scattering. However, the scattering characteristics of real terrains usually lay between these two end-member models. I present here a new method that combines both approaches, to extract the deformation signal at more points and with higher overall signal-to-noise ratio than can either approach alone. I apply the combined method to data acquired over Eyjafjallaj kull volcano in Iceland, and detect time-varying ground displacements associated with two intrusion events.

DOI

[8]
Lanari R, Mora O, Manunta M, et al.A small-baseline approach for investigating deformations on full-resolution differential SAR interferograms[J]. Geoscience & Remote Sensing IEEE Transactions on, 2004,42(7):1377-1386.This paper presents a differential synthetic aperture radar (SAR) interferometry (DIFSAR) approach for investigating deformation phenomena on full-resolution DIFSAR interferograms. In particular, our algorithm extends the capability of the small-baseline subset (SBAS) technique that relies on small-baseline DIFSAR interferograms only and is mainly focused on investigating large-scale deformations with spatial resolutions of about 100 100 m. The proposed technique is implemented by using two different sets of data generated at low (multilook data) and full (single-look data) spatial resolution, respectively. The former is used to identify and estimate, via the conventional SBAS technique, large spatial scale deformation patterns, topographic errors in the available digital elevation model, and possible atmospheric phase artifacts; the latter allows us to detect, on the full-resolution residual phase components, structures highly coherent over time (buildings, rocks, lava, structures, etc.), as well as their height and displacements. In particular, the estimation of the temporal evolution of these local deformations is easily implemented by applying the singular value decomposition technique. The proposed algorithm has been tested with data acquired by the European Remote Sensing satellites relative to the Campania area (Italy) and validated by using geodetic measurements.

DOI

[9]
范锐彦,焦健,高胜,等. InSAR时序分析高相干目标选取方法比较研究[J].地球信息科学学报,2016,18(6):805-814.lt;p>合成孔径雷达干涉测量（Interferometric Synthetic Aperture Radar,InSAR）时序分析技术,利用时域上多幅SAR图像选择时间基线、空间基线满足一定条件的干涉对进行干涉,通过建立干涉相位形变模型获取地表形变信息。InSAR时序分析技术改善了差分干涉测量中时空失相干、大气延迟等问题,广泛应用于有关形变监测的多个领域,并逐渐成为获取长期地表形变趋势的重要手段。在InSAR时序分析中,针对不同应用选取合适的高相干目标选取方法,获得可靠的高相干目标,是获取精确可靠的地表形变信息的基础。本文通过分析永久散射体（Permanent Scatterers,PS）、分布式散射体（Distributed Scatterers,DS）选取方法的理论模型及其算法,研究其适用地物目标类型的异同,并分析、归纳总结了幅度相关法、相位分析法、信号杂波比法、相干性统计法等不同的高相干目标选取方法的优缺点。最后,以阿尔金断裂带西段部分区域为研究区,分别采用具有代表性的PS、DS选取方法开展该研究区域的选点实验,结果表明该研究区域DS选取方法比PS选取方法适用。本文方法为解决在不同地理区域进行应用研究时选取合适的选点方法提供参考。</p>

DOI

[ Fan R Y, Jiao J, Gao S, et al.2016. Comparison research of high coherent target selection based on InSAR time series analysis[J]. Journal of Geo-information Science, 2016,18(6):805-814. ]

[10]
Mora O, Mallorqui J J, Broquetas A.Linear and nonlinear terrain deformation maps from a reduced set of interferometric SAR images[J]. IEEE Transactions on Geoscience & Remote Sensing, 2003,41(10):2243-2253.

[11]
陈强,刘国祥,李永树,等.干涉雷达永久散射体自动探测--算法与实验结果[J].测绘学报,2006,35(2):112-117.

[Chen Q, Liu G X, Li Y S, et al.Automated detection of permanent scatterers in radar interferometry: Algorithm and testing results[J]. Acta Geodaetica Et Cartographica Sinica, 2015,35(2):112-117. ]

[12]
曲世勃,王彦平,洪文. PSDInSAR的永久散射体时序选择方法[J].电子与信息学报,2011,33(2):381-387.已有永久散射体(PS)识别技术均侧重于利用时序数据集的统计特性，而没有考虑到数据集的时序特性，这样的处理方式势必带来时序信息的浪费，造成部分PS点的漏选。该文重点关注一种永久散射体具有较好的相位稳定度但却在整个监测时间内不连续，称这种PS点为类永久散射体点。文中对类永久散射体的概念及特征进行了详细描述，利用仿真实验分析了类永久散射体应用的可行性，根据类永久散射体特性对其进行了有效选择，形成一种新的永久散射体选择方法&mdash;&mdash;时序选择法。同时对时序选择法以天津地区Envisat ASAR影像数据进行实验验证。通过类永久散射体的选择，PS点数量提高了、不均匀分布特性得到改善，同时保证了其较高的相干性。

DOI

[ Qu S B, Wang Y P, Hong W.The PS selection method using temporal information in PSDInSAR technique[J]. Dianzi Yu Xinxi Xuebao/Journal of Electronics & Information Technology, 2011,33(2):381-387. ]

[13]
Hooper A, Spaans K, Bekaert D, et al.StaMPS/MTI manual[J]. Delft institute of earth observation and space systems delft university of technology, Kluyverweg, 2010,1:2629.

[14]
Wegmǜller U.GAMMA IPTA Processing Example Luxemburg, GAMMA Technical Report[R]. Luxeburg: GAMMA, 2005.

[15]
丁伟. PSInSAR点目标提取及相位解缠技术研究[D].长沙:中南大学,2011.

[ Ding W.Study of PSInSAR on the technique of points selection and phase unwrapping[D]. Changsha: Central South University, 2011. ]

[16]
Ferretti A, Fumagalli A, Novali F, et al.A new algorithm for processing interferometric data-stacks: SqueeSAR[J]. IEEE Transactions on Geoscience and Remote Sensing, 2011,49(9):3460-3470.

[17 Kvam P H, Vidakovic B. Nonparametric statistics with applications to science and engineering[M]. Hoboken, New Jersey: John Wiley & Sons, Inc., 2007.

[18]
Ferretti A, Prati C, Rocca F.Permanent scatterers in SAR interferometry[J]. IEEE Transactions on geoscience and remote sensing, 2001,39(1):8-20.Temporal and geometrical decorrelation often prevents SAR interferometry from being an operational tool for surface deformation monitoring and topographic profile reconstruction. Moreover, atmospheric disturbances can strongly compromise the accuracy of the results. The authors present a complete procedure for the identification and exploitation of stable natural reflectors or permanent scatterers (PSs) starting from long temporal series of interferometric SAR images. When, as it often happens, the dimension of the PS is smaller than the resolution cell, the coherence is good even for interferograms with baselines larger than the decorrelation one, and all the available images of the ESA ERS data set can be successfully exploited. On these pixels, submeter DEM accuracy and millimetric terrain motion detection can be achieved, since atmospheric phase screen (APS) contributions can be estimated and removed. Examples are then shown of small motion measurements, DEM refinement, and APS estimation and removal in the case of a sliding area in Ancona, Italy. ERS data have been used.

DOI

[19]
罗小军,黄丁发,刘国祥.时序差分雷达干涉中永久散射体的自动探测[J].西南交通大学学报,2007,42(4):414-418.为了有效探测时序差分雷达干涉中的永久散射体,考虑永久散射体对雷达波的强反射特性和散射的稳定性,提出了探测永久散射体的振幅信息双阈值法.该方法根据振幅阈值选取具有强反射特性的点作为永久散射体候选点,然后根据振幅离差阈值,选取散射特性稳定的候选点作为永久散射体.利用上海地区的26幅ERS-1/2 C波段SAR(合成孔径雷达)图像进行了实验,证明该方法是有效和可靠的.

DOI

[ Luo X, Huang D, Liu G.Automated detection of permanent scatterers in time serial differential radar interferometry[J]. Journal of Southwest Jiaotong University, 2007,42(4):414-418. ]

[20]
Xia Y.Homogeneous pixel selection for distributed scatterers using multitemporal SAR data stacks[D]. Technical University of Munich (TUM), 2016.

[21]
何庆成,刘文波,李志明.华北平原地面沉降调查与监测[J].高校地质学报,2006,12(2):195-209.华北平原是世界上超采地下水最严重的地区,也是地面沉降面积最大的地区.大约有70,000 km2的地下水水位低于海平面.随着近20年来的经济快速发展、城市化进程的加快、地表水污染程度的增加、高层建筑的施工以及对地下水的需求与日俱增,华北平原地面沉降呈现加剧的趋势.对华北平原地面沉降调查与监测提出了一套探索性的工作思路,并在基础监测设施的建设以及综合分析研究方面取得一些积极成果.

DOI

[ Qing-Cheng H E, Liu W B, Zhi-Ming L I. Land subsidence survey and monitoring in the North China Plain[J]. Geological Journal of China Universities, 2006,12(2):195-209. ]

[22]
Ng A H-M, Ge L, Li X, et al. Monitoring ground deformation in Beijing, China with persistent scatterer SAR interferometry[J]. Journal of Geodesy, 2012,86(6):375-392.This paper investigated the long term ground deformation in Beijing, China, using persistent/permanent scatterer interferometry (PSI) techniques. GEOS-PSI (Geodesy and Earth Observing Systems-PSI), an in-house software developed at UNSW for PSI, has been applied to 41 ENVISAT ASAR images acquired over the metropolitan area of Beijing City between June 2003 and March 2009 and 24 ALOS PALSAR images (two Paths: 10 acquisitions from January 2007 to October 2008 and 14 acquisitions from February 2007 to September 2009). The results generated using these datasets from the two satellites were cross-validated. Correlations between the results of ENVISAT ASAR and ALOS PALSAR agreed very well. The horizontal and vertical displacement rate maps over Beijing City were obtained from the results generated with data acquired by both satellites over the period from 1st February 2007 to 1st November 2008. The results indicate that the displacements in Beijing City were mainly in the vertical direction. The majority of the easting displacement rates were in the range of 6110 mm/year to 1002mm/year, while the vertical rates were in the range of 61115 mm/year to 602mm/year. The possible cause for the ground deformation is groundwater extraction based on our research as well as earlier published studies.

DOI

[23]
Zebker H A, Villasenor J.Decorrelation in interferometric radar echoes[J]. IEEE Transactions on Geoscience and Remote Sensing, 1992,30(5):950-959.A radar interferometric technique for topographic mapping of surfaces promises a high resolution, globally consistent approach to generation of digital elevation models. One implementation approach, that of utilizing a single SAR system in a nearly repeating orbit, is attractive not only for cost and complexity reasons but also in that it permits inference of changes in the surface over the orbit repeat cycle from the correlation properties of the radar echoes. The various sources contributing to the echo correlation statistics are characterized, and the term which most closely describes surficial change is isolated. There is decorrelation increasing with time, but digital terrain model generation remains feasible.

DOI

[24]
Doin M P, Lodge F, Guillaso S, et al.Presentation of the small baseline NSBAS processing chain on a case example: The ETNA deformation monitoring from 2003 to 2010 using ENVISAT data[C]. FRINGE 2011 ESA Conference, 2011:954-966.

[25]
Agram P S.Persistent scatterer interferometry in natural terrain[D]. Stanford: Stanford university, 2010.

[26]
Zhang Y-Q, Wang S, Wang R.Research on monitoring land subsidence in Beijing plain area using PS-InSAR technology[J]. Spectroscopy and Spectral Analysis, 2014,34(7):1898-1902.In the present paper,the authors use permanent scatterers synthetic aperture radar interferometry(PS-InSAR)technique and 29 acquisitions by Envisat during 2003 to 2009 to monitor and analyze the spatial-temporal distribution and mechanism characterize of land subsidence in Beijing plain area.The results show that subsidence bowls have been bounded together in Beijing plain area,which covers Chaoyang,Changping,Shunyi and Tongzhou area,and the range of subsidence has an eastward trend.The most serious regional subsidence is mainly distributed by the quaternary depression in Beijing plain area.PS-Insar results also show a new subsidence bowl in Pinggu.What's more,the spatial and temporal distribution of deformation is controlled mainly by faults,such as Liangxiang-Shunyi fault,Huangzhuang-Gaoliying fault,and Nankou-Sunhe fault.The subsidence and level of groundwater in study area shows a good correlation,and the subsidence shows seasonal ups trend during November to March and seasonal downs trend during March to June along with changes in groundwater levels.The contribution of land subsidence is also influenced by stress-strain behavior of aquitards.The compaction of aquitards shows an elastic,plastic,viscoelastic pattern.

DOI PMID

[27]
Luo Y.Primary investigation of formation and genetic mechanism of land subsidence based on PS-InSAR technology in Beijing[J]. Spectroscopy and Spectral Analysis, 2014,34(8):2185-2189.Abstract The present paper adopts permanent scatterer interferometric synthetic aperture radar(PS-InSAR) technique to obtain land subsidence information in Beijing plain area. Then, combined with the time series of meteorological data, groundwater dynamic monitoring data, interferometric data and geological structure data, the formation and evolution mechanism of land subsidence were revealed. The results show that (1) Beijing regional land subsidence characteristics are obvious, more land subsidence funnel areas are interconnected, the settlement is influenced by rainfall recharge and exhibits seasonal fluctuation characteristics; (2) The land subsidence center and groundwater drawdown funnel centre are not fully consistent, unconfined aquifer and shallow confined aquifer are the major contribution factors and have greater impact on the land subsidence; (3) Land subsidence mainly occurred in the clay layer with a thickness of 50-70 m; (4) Land subsidence caused by tectonic controls is significant and the deformation gradient is great on both sides of the fault.

DOI PMID

[28]
Wang Y, Liu Y, Hu L.GPS-based research on changes of land subsidence in Beijing from 2007 to 2012[J]. Journal of Catastrophology, 2015,30:11-15.GPS data of Beijing from 2007 to 2012 are processed by Empirical Mode Decomposition(EMD)method and are compared with bury of groundwater to analyze the land subsidence in recent years.Firstly,according to the yearly land subsidence velocity during 2007-2012,settlement center in Beijing is discovered to be formed together as a Middle East settlement funnel.DSQI cumulative settlement reached 510mm,yearly land subsidence velocities of DSQI,CHAO,NLSH stations are respectively 85mm / a,41.7mm / a and 20mm / a.Secondly,trend terms in vertical trend sequence of DSQI and NLSH station are extracted according to EMD method to analyze the subsidence trend.Finally,the relation between bury of groundwater and subsidence trend of GPS stations is comparative analyzed.According to the comparison,it is found that subsidence trend is consistent well with the change of groundwater depth,and the groundwater level change is the main inducement of the occurrence and development of land subsidence.

Options
文章导航

/

〈 〉