基于深度学习的全极化SAR影像冰川边界识别
Identification of glaciers using fully polarimetric SAR data based on deep-learning
- 2023年27卷第9期 页码:2098-2113
纸质出版日期: 2023-09-07
DOI: 10.11834/jrs.20221541
引用
阅读全文PDF
扫 描 看 全 文
纸质出版日期: 2023-09-07
扫 描 看 全 文
引用
阅读全文PDF
冰川识别对于周边地区水资源与气候变化监测具有重要意义。全极化SAR影像包含地物表面散射、偶次散射、体散射、统计特性等丰富的特征,而深度学习能够充分挖掘影像信息,因此使用全极化SAR影像结合深度学习能够得到精确的冰川识别效果。本文基于喜马拉雅山脉西端ALOS2-PALSAR全极化影像,使用VGG16特征提取网络与全卷积神经网络模型U-net相结合的VGG16-unet对冰川进行识别。采用的特征包括极化相干矩阵对角线元素、Freeman-Durden、H/A/α、Pauli、VanZyl、Yamaguchi这5种极化分解参数共计19种特征。为了充分利用影像信息,对这些特征进行分析与组合,并比较它们之间的冰川识别精度,以选取最佳特征。由于冰川与非冰川的地形具有明显差异,因此将DEM、坡度、局部入射角等作为辅助特征与极化特征结合。通过对比不同极化特征分类精度得出,基于物理特性的Pauli、Freeman-Durden、VanZyl、Yamaguchi特征分类的精度较高,其中Pauli特征分类的精度最高,整体精度(OA)达到92.54%,平均用户交并比(mIoU)达到78.78%。加入地形数据后整体精度(OA)提升至94.34%,平均用户交并比(mIoU)提升至82.35%。为了进一步提高冰川的识别精度,提出了一种基于单波段特征整体精度(OA)及召回率(Recall)筛选出的SDV(表面散射、偶次散射、体散射)特征交叉组合方式,结果显示,该组合整体精度(OA)达到94.98%,用户交并比(mIoU)达到85.67%,比Pauli特征分类精度分别高出0.64%和 3.32%。上述结果表明,选择最佳的特征组合方式并结合深度学习在提升冰川识别精度中具有重要的作用。
Glacier identification is important for monitoring water resources and climate change in surrounding areas. Although optical images have achieved high accuracy in glacier boundary identification, optical images are affected by cloud cover, and reproducing information under the clouds is difficult. Fully polarized SAR images contain rich features, and deep learning can fully exploit image information. Therefore, using fully polarized SAR images combined with deep learning can compensate for the lack of optical images and obtain accurate glacier recognition results. In this paper, VGG16-unet (VGG16 combined with U-net) is used to identify glaciers based on ALOS2-PALSAR fully polarized images of the western part of the Himalayas. The features include the diagonal elements of the polarization coherence matrix, Freeman-Durden, H/A/α, Pauli, VanZyl, and Yamaguchi polarization decomposition parameters totaling 19 features. To make full use of the image information, these features are analyzed and combined, and the glacier recognition accuracies are compared to select the best features. Given evident differences between glacier and nonglacier topography, elevation, slope, and local incidence angle are combined with polarization features as auxiliary features.
Comparing the classification accuracy of different polarization features reveals the accuracy of Pauli, Freeman-Durden, VanZyl, and Yamaguchi features based on physical characteristics is higher, among which Pauli features have the highest accuracy with an Overall Accuracy (OA) of 92.54% and an average user intersection ratio (mIoU) of 78.78%. The OA is improved to 94.34%, and the mIoU is improved to 82.35% after adding the topographic data. In order to improve the recognition accuracy of glaciers further, a feature cross-combination approach is proposed, and results show the OA of the combination reaches 94.98%, and the mIoU reaches 85.67%, which are 0.64% and 3.32% higher than the classification accuracy of Pauli features, respectively.
Selecting the best feature combination method and combining with deep learning plays an important role in improving the accuracy of glacier recognition, and the use of neural networks combined with fully polarized SAR images can effectively compensate for the shortcomings of optical images in glacier boundary identification.
来自喜马拉雅山脉的冰川融水为周边地区居住的人口提供了至关重要的淡水资源(
冰川识别是冰川监测的重要组成部分,而使用遥感图像能够快速、准确的提取冰川边界以显示冰川变化与研究冰川物质平衡(
全极化SAR数据具有比传统SAR以及单极化SAR数据更丰富的信息,而极化分解可以量化表面和内部后向散射的贡献,提高对不同位置冰川的识别能力(
近年来,深度学习发展迅速,在AlexNet基础上发展出的VGG(
研究区位于喜马拉雅山脉西端克什米尔,区域内包含典型的中纬度、高海拔冰川,大部分地区高程处于3200—6500 m。该地区多数大型冰川地势起伏大,平均坡度处于17°左右,同一冰川内不同区域坡度具有明显差异(
图1 克什米尔冰川地理位置及其分布
Fig. 1 Location and distribution of Kashmir glaciers
采用ALOS2-PALSAR全极化数据。ALOS-2卫星搭载的PALSAR是一种SAR传感器,向地面物体发射L波段雷达波,并接收其后向散射信号。与ALOS-PALSAR相比,它在时间和空间分辨率、观测波段、极化方式以及数据传输的时间间隔方面有显著的提升。特别是全极化模式可以获取目标的所有极化特性,大大提高了成像雷达对目标信息的获取能力。本文使用了条带模式下的同一轨两景全极化SAR影像,空间分辨率为5 m。成像时间为2020年10月2日。
SRTM(Shuttle Radar Topography Mission)数据是由美国航空航天局(NASA)和国防部国家测绘局(NIMA)联合测量的,雷达影像数据覆盖全球陆地表面的80%以上,精度有1 arc-second 和3 arc-seconds两种,即30 m与90 m空间分辨率的数据。本文选取SRTM 30 m空间分辨率的数据作为原始数据,用来对SAR数据进行地理编码并计算雷达波在冰川表面的局部入射角以及地表的坡度信息。
GLIMS(Global Land Ice Measurements from Space)是一项使用光学卫星数据监测世界冰川的计划,而RGI(Randolph Glacier Inventory)是对GLIMS的补充,它将曾经的冰川地图和有用的属性相结合。RGI包含了全球的冰川轮廓,使用RGI 6.0(2017年7月28日)作为训练样本的标签数据。选取2020年10月21日的Landsat 8影像作为精度验证数据。
SAR影像,DEM与冰川边界数据按照
图2 基于深度学习的ALOS2-PALSAR影像冰川识别流程
Fig. 2 Deep learning based glacier recognition process for ALOS2-PALSAR images
3.1.1 数据预处理
SAR数据预处理分为4个部分,包括极化矩阵生成、极化滤波、极化分解和地理编码(红色虚线矩形,
极化雷达影像中的每一个像素的信息能够通过Sinclair散射矩阵表示。
S=[SHHSHVSVHSVV] | (1) |
式中,S是描述散射体对入射波散射现象的2×2散射矩阵,SHV表示发射的水平H极化波和接收的垂直V极化波的目标后向散射系数。为了从散射矩阵中提取物理信息,将矩阵转换为字典矩阵基表示的目标散射向量k:
k=[SHH SHV SVH SVV]T | (2) |
基于以上的等式,与极化相关的散射矩阵的二阶乘积可以定义为相干矩阵[T3](
T3=12[|SHH+SHV|2(SHH+SVV)(SHH-SVV)*2(SHH+SVV)S*HV(SHH-SVV)(SHH+SVV)*|SHH-SVV|22(SHH-SVV)S*HV2SHV(SHH+SVV)*2SHV(SHH-SVV)*4|SHV|2] | (3) |
相干斑是SAR成像中的一种散射现象,含有相干斑的后向散射系数影像不利于图像解译。为了消除相干斑的影响,使用Improved Lee sigma滤波器(
当雷达波与地面目标相互作用时极化状态会发生变化,这种变化受到目标的介电常数、结构、粗糙度和化学成份影响。通过对表面和体积分量的分解能够推断出冰面和冰内特性的独立信息,而极化分解可以量化表面和体积后向散射的贡献,从而提高区分不同冰川带的能力(Sharma等,2010)。
由于雷达的成像机制,地形因素会使影像产生几何畸变,并对雷达接收的后向散射产生一定影响(
3.1.2 极化特征组合
为了探究不同极化特征在冰川识别中的作用,选取了Pauli分解(
序号 | 特征 | 含义 |
---|---|---|
1 | SurfFD | Freeman-Durden极化分解中表面散射分量,PS=|SHH|2,其中PS表示表面散射功率 |
2 | DblFD | Freeman-Durden极化分解中偶次散射分量,PD=|SVV|2,其中PD表示偶次散射功率 |
3 | VolFD | Freeman-Durden极化分解中体散射分量,PV=2|SHV|2,其中PV表示体散射功率 |
4 | H | 熵值,衡量散射过程中的极化程度;H/A/α分解的参数 |
5 | A | 各向异性值,衡量第二与第三个散射机制的相对大小;H/A/α分解的参数 |
6 | α | 极化散射角,对散射机制的物理解释,取0°,45°,90°时分别代表表面散射,体散射,偶次散射;H/A/α分解的参数 |
7 | Surfpauli | Pauli极化分解中表面散射分量 |
8 | Dblpauli | Pauli极化分解中偶次散射分量 |
9 | Volpauli | Pauli极化分解中体散射分量 |
10 | T11 | 相干矩阵对角线分量,含有表面散射信息,T11 ∈ diag (T3) |
11 | T22 | 相干矩阵对角线分量,含有偶次散射信息,T22 ∈ diag (T3) |
12 | T33 | 相干矩阵对角线分量,含有体散射信息,T33 ∈ diag (T3) |
13 | Surfvan | VanZyl极化分解中表面散射分量 |
14 | Dblvan | VanZyl极化分解中偶次散射分量 |
15 | Volvan | VanZyl极化分解中体散射分量 |
16 | Surfyama | Yamaguchi极化分解中表面散射分量,T3=fsTsurface+ fdTdouble-bounce+ fvTvolume+fcThelix,式中Tsurface是表面散射拓展矩阵, fs为其拓展系数 |
17 | Dblyama | Yamaguchi极化分解中偶次散射分量,T3=fsTsurface+ fdTdouble-bounce+ fvTvolume+fcThelix,式中Tdouble-bounce是偶次散射拓展矩阵, fd为其拓展系数 |
18 | Volyama | Yamaguchi极化分解中体散射分量,T3=fsTsurface+ fdTdouble-bounce+ fvTvolume+fcThelix,式中Tvolume是体散射拓展矩阵, fv为其拓展系数 |
19 | Hlxyama | Yamaguchi极化分解中螺旋散射分量,T3=fsTsurface+ fdTdouble-bounce+ fvTvolume+fcThelix,式中Thelix是螺旋散射拓展矩阵, fc为其拓展系数 |
3.2.1 输入数据处理
输入数据处理如
图3 ALOS2-PALSAR影像及地形数据的合成与裁剪
Fig. 3 Synthesis and cropping of ALOS2-PALSAR images and topographic data
图4 神经网络数据增强
Fig. 4 Neural network data enhancement
图5 VGG16-unet训练、验证及测试数据分布
Fig. 5 Regional distribution of VGG16-unet train, validation and test data
3.2.2 网络结构设计
U-net网络呈U形结构,主要包括编码器和解码器部分,编码器通过卷积和池化对影像降维并提取特征,解码器采用上采样且与特征部分相同尺度的图像进行拼接,将浅层特征和深层特征结合起来,更有利于提取目标。为了使网络用于冰川识别,在U-net的基础上设计了VGG16-unet,并对其中的参数进行修改以确定最佳参数(
编号 | 池化层数量 | 卷积核大小 | 编码器卷积层 | 起始层卷积核数量 | 整体精度(OA) |
---|---|---|---|---|---|
1 | 4 | 5 | 2,2,2,2,2 | 64 | 0.8772 |
2 | 4 | 5 | 2,2,2,2,2 | 32 | 0.8901 |
3 | 4 | 5 | 2,2,3,3,3 | 64 | 0.8847 |
4 | 4 | 5 | 2,2,3,3,3 | 32 | 0.8984 |
5 | 4 | 3 | 2,2,2,2,2 | 64 | 0.8943 |
6 | 4 | 3 | 2,2,2,2,2 | 32 | 0.8854 |
7 | 4 | 3 | 2,2,3,3,3 | 64 | 0.8661 |
8 | 4 | 3 | 2,2,3,3,3 | 32 | 0.9246 |
9 | 5 | 3 | 2,2,2,2,2 | 64 | 0.9073 |
10 | 5 | 3 | 2,2,2,2,2 | 32 | 0.8866 |
11 | 5 | 3 | 2,2,3,3,3 | 64 | 0.9185 |
12 | 5 | 3 | 2,2,3,3,3 | 32 | 0.9163 |
13 | 6 | 3 | 2,2,3,3,3 | 32 | 0.8685 |
14 | 2 | 3 | 2,2,3,3,3 | 32 | 0.8792 |
图6 VGG16-unet池化与去池化过程
Fig. 6 VGG16-unet pooling and unpooling process
图7 VGG16-unet网络结构及其参数
Fig. 7 Structure and parameters of VGG16-unet network
用于精度评价的冰川边界数据是在Landsat 8 5,4,3波段合成的影像上参考RGI 6.0边界进行目视解译修改得到。精度评价的方法都是基于冰川识别结果与目视解译冰川轮廓来进行对比分析,包括精确度(P)、召回率(R)、F1分数(F1)、交并比(IoU)、平均交并比(mIoU)、整体精度(OA)。
P=TPTP+FP | (4) |
R=TPTP+FN | (5) |
F1=2×Precision×RecallPrecision+Recall | (6) |
IoU=TPTP+FN+FP | (7) |
mIoU=n∑iIoUn | (8) |
OA=TP+TNTP+TN+FP+FN | (9) |
式中,TP是预测结果为正类,实际是正类;FP是预测结果为正类,实际是负类;TN是预测结果为负类,实际是负类;FN是预测结果为负类,实际是正类;n为类别总数,i为类别序号。
将19个特征参数单独输入神经网络中进行识别对比整体精度(
序号 | 特征参数 | 整体精度(OA) | 召回率(Recall) |
---|---|---|---|
1 | SurfFD | 81.65 | 46.68 |
2 | DblFD | 78.1 | 45.76 |
3 | VolFD | 80.78 | 59.17 |
4 | H | 82.55 | 64.51 |
5 | A | 81.21 | 55.35 |
6 | α | 81.55 | 56.28 |
7 | Surfpauli | 80.14 | 56.97 |
8 | Dblpauli | 79.90 | 55.84 |
9 | Volpauli | 83.26 | 59.09 |
10 | T11 | 80.12 | 50.48 |
11 | T22 | 79.77 | 41.46 |
12 | T33 | 80.59 | 53.97 |
13 | Surfvan | 82.31 | 56.44 |
14 | Dblvan | 79.47 | 52.86 |
15 | Volvan | 80.61 | 51.58 |
16 | Surfyama | 82.33 | 65.21 |
17 | Dblyama | 77.65 | 50.48 |
18 | Volyama | 81.18 | 52.14 |
19 | Hlxyama | 75.54 | 52.97 |
利用SDV组合进行识别的整体精度最高,达到93.3%,平均用户交并比(mIoU)达到80.91%,召回率达到76.89%(
序号 | 极化分解 | 精确度 | 召回率 | F1分数 | IoU | mIoU | 整体精度 |
---|---|---|---|---|---|---|---|
1 | H/A/α | 80.85 | 70.02 | 75.05 | 60.06 | 74.59 | 90.65 |
2 | Yamaguchi | 90.92 | 61.41 | 73.31 | 57.86 | 73.81 | 91.02 |
3 | VanZyl | 75.94 | 84.13 | 79.83 | 66.42 | 78.07 | 91.46 |
4 | Freeman-Durden | 89.42 | 69.1 | 77.96 | 63.88 | 77.38 | 92.15 |
5 | T_matrix | 85.72 | 74.96 | 79.98 | 66.64 | 78.89 | 92.46 |
6 | Pauli | 87.8 | 73.03 | 79.73 | 66.3 | 78.78 | 92.54 |
7 | SDV | 88.22 | 76.89 | 82.17 | 69.73 | 80.91 | 93.3 |
为了进一步分析冰川识别精度与极化特征的关系,选取包含岩石、表碛、冰、雪多种地表类型的G1冰川区域作为样本,对该区域内多种极化特征进行分析(
图8 18种极化特征实例及样本点分布
Fig. 8 Examples of 18 polarization characteristics and sample point distribution
图9 不同冰川地表类型中岩石、表碛、冰、雪的18个极化特征分布
Fig. 9 Distribution of 18 polarization features of bare rock, debris, ice and snow in different glacial surface types
序号 | 极化分解 | 精确度 | 召回率 | F1分数 | IoU | mIoU | 整体精度 | |||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
含∂ | 不含∂ | 含∂ | 不含∂ | 含∂ | 不含∂ | 含∂ | 不含∂ | 含∂ | 不含∂ | 含∂ | 不含∂ | |||||||
1 | T_matrix | 91.36 | 85.72 | 74.77 | 74.96 | 82.24 | 79.98 | 69.83 | 66.64 | 81.1 | 78.89 | 93.51 | 92.46 | |||||
2 | H/A/α | 90.07 | 80.85 | 77.36 | 70.02 | 83.23 | 75.05 | 71.28 | 60.06 | 81.93 | 74.59 | 93.74 | 90.65 | |||||
3 | Yamaguchi | 89.07 | 90.92 | 79.02 | 61.41 | 83.74 | 73.31 | 72.03 | 57.86 | 82.35 | 73.81 | 93.84 | 91.02 | |||||
4 | Freeman-Durden | 89.64 | 89.42 | 80.13 | 69.1 | 84.62 | 77.96 | 73.34 | 63.88 | 83.18 | 77.38 | 94.15 | 92.15 | |||||
5 | VanZyl | 88.81 | 75.94 | 81.08 | 84.13 | 84.77 | 79.83 | 73.56 | 66.42 | 83.29 | 78.07 | 94.15 | 91.46 | |||||
6 | Pauli | 88.8 | 87.8 | 82.18 | 73.03 | 85.36 | 79.73 | 72.03 | 66.3 | 82.35 | 78.78 | 94.34 | 92.54 | |||||
7 | SDV | 89.05 | 88.22 | 85.54 | 76.89 | 87.26 | 82.17 | 77.4 | 69.73 | 85.67 | 80.91 | 94.98 | 93.3 |
加入局部入射角、坡度、DEM数据后7组特征的冰川识别精度均有显著提升,选取一张512×512像素影像观察不同极化特征加入地形因子后的分类情况(
图10 不同极化特征组合冰川识别结果对比
Fig. 10 Comparison of glacier identification results for different combinations of polarization features
通过
图11 G1冰川剖面线分布(背景数据采用Landsat 8 5,4,3波段)
Fig. 11 Distribution of profile lines in G1 glacier (The background data were using the 5,4,3 bands of Landsat 8)
图12 G1冰川地形因子剖面分析曲线
Fig. 12 G1 Glacier topographic factor profile analysis curve
图13 基于深度学习的克什米尔ALOS2-PALSAR影像冰川识别分类结果
Fig. 13 Deep learning based classification results for glacier identification in ALOS2-PALSAR images of Kashmir
图14 VGG16-unet与机器学习冰川识别结果对比
Fig. 14 Comparison of glacier recognition results between VGG16-unet and machine learning
为了验证VGG16-unet方法在冰川识别上的可迁移性,选取位于藏东南地区左贡县旺达镇(
图15 基于ALOS1-PALSAR的VGG16-unet训练、验证及测试数据分布
Fig. 15 VGG16-unet train, validation and test data distribution based on ALOS1-PALSAR
图16 基于深度学习的旺达镇ALOS1-PALSAR影像冰川识别分类结果
Fig. 16 Deep learning based classification results for glacier identification in ALOS1-PALSAR images of Wangda town
冰川识别精度如
本文基于ALOS2-PALSAR全极化影像数据,利用VGG16与全卷积神经网络模型U-net结合生成的VGG16-unet,对19个极化特征进行对比、分析与组合,并结合地形数据进行冰川边界的识别,得到的结果显示(1)VGG16-unet神经网络对冰川区地物物理特性的差异更加敏感,结合物理特性相关的Pauli、Freeman-Durden、VanZyl、Yamaguchi、T矩阵对角元素进行识别能够得到更高的冰川识别精度,通过筛选得到的SDV组合特征精度最高,加入地形因子后冰川的识别精度能够达到94.98%;将19种极化分解特征输入神经网络时,冰川散射机制中占主导的表面散射与体散射的识别精度高于偶次散射的识别精度,虽然与统计特性相关的H、A、α都能得到较高的精度,但进行组合输入时由于特征冗余,使得其冰川识别精度最低;将地形数据作为辅助特征与极化特征相结合输入神经网络能够大幅提升冰川的识别精度,但是会影响极化特征分离度在冰川识别中的作用;特征是冰川识别的关键,特征显著的表碛、冰、雪等大型冰川能够得到良好的识别,而与岩石相似的小型冰川识别效果较差。(2)综上所述,利用全极化SAR影像结合神经网络进行冰川边界的识别能够有效弥补光学影像受云层影响导致的云及云影下冰/雪信息的缺失。与
Badrinarayanan V, Kendall A and Cipolla R. 2017. SegNet: a deep convolutional encoder-decoder architecture for image segmentation. IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(12): 2481-2495 [DOI: 10.1109/TPAMI.2016.2644615] [百度学术]
Brun F, Berthier E, Wagnon P, Kääb A and Treichler D. 2017. A spatially resolved estimate of high mountain Asia glacier mass balances from 2000 to 2016. Nature Geoscience, 10(9): 668-673 [DOI: 10.1038/ngeo2999] [百度学术]
Callegari M, Carturan L, Marin C, Notarnicola C, Rastner P, Seppi R and Zucca F. 2016. A Pol-SAR analysis for alpine glacier classification and snowline altitude retrieval. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 9(7): 3106-3121 [DOI: 10.1109/JSTARS.2016.2587819] [百度学术]
Cloude S R. 1985. Target decomposition theorems in radar scattering. Electronics Letters, 21(1): 22-24 [DOI: 10.1049/el:19850018] [百度学术]
Cloude S R and Pottier E. 1996. A review of target decomposition theorems in radar polarimetry. IEEE Transactions on Geoscience and Remote Sensing, 34(2): 498-518 [DOI: 10.1109/36.485127] [百度学术]
Freeman A and Durden S L. 1998. A three-component scattering model for polarimetric SAR data. IEEE Transactions on Geoscience and Remote Sensing, 36(3): 963-973 [百度学术]
Harris C, Arenson L U, Christiansen H H, Etzelmüller B, Frauenfelder R, Gruber S, Haeberli W, Hauck C, Hölzle M, Humlum O, Isaksen K, Kääb A, Kern-Lütschg M A, Lehning M, Matsuoka N, Murton J B, Nözli J, Phillips M, Ross N, Seppälä M, Springman S M and Vonder Mühll D. 2009. Permafrost and climate in Europe: monitoring and modelling thermal, geomorphological and geotechnical responses. Earth-Science Reviews, 92(3/4): 117-171 [DOI: 10.1016/j.earscirev.2008.12.002] [百度学术]
Huang L, Li Z, Tian B S, Zhou J M and Chen Q. 2014. Recognition of supraglacial debris in the Tianshan mountains on polarimetric SAR images. Remote Sensing of Environment, 145: 47-54 [DOI: 10.1016/j.rse.2014.01.020] [百度学术]
Immerzeel W W, van Beek L P H and Bierkens M F P. 2010. Climate change will affect the Asian water towers. Science, 328(5984): 1382-1385 [DOI: 10.1126/science.1183188] [百度学术]
Lee J S, Wen J H, Ainsworth T L, Chen K S and Chen A J. 2009. Improved sigma filter for speckle filtering of SAR imagery. IEEE Transactions on Geoscience and Remote Sensing, 47(1): 202-213 [DOI: 10.1109/TGRS.2008.2002881] [百度学术]
Nie Y, Zhang Y L, Liu L S and Zhang J P. 2010. Monitoring glacier change based on remote sensing in the Mt. Qomolangma national nature preserve, 1976-2006. Acta Geographica Sinica, 65(1): 13-28 [百度学术]
聂勇, 张镱锂, 刘林山, 张继平. 2010. 近30年珠穆朗玛峰国家自然保护区冰川变化的遥感监测. 地理学报, 65(1): 13-28 [DOI: 10.11821/xb201001003] [百度学术]
Parrella G, Hajnsek I and Papathanassiou K P. 2016. Polarimetric decomposition of L-band PolSAR backscattering over the austfonna ice cap. IEEE Transactions on Geoscience and Remote Sensing, 54(3): 1267-1281 [DOI: 10.1109/TGRS.2015.2477168] [百度学术]
Paul F, Bolch T, Briggs K, Kääb A, McMillan M, McNabb R, Nagler T, Nuth C, Rastner P, Strozzi T and Wuite J. 2017. Error sources and guidelines for quality assessment of glacier area, elevation change, and velocity products derived from satellite data in the Glaciers_cci project. Remote Sensing of Environment, 203: 256-275 [DOI: 10.1016/j.rse.2017.08.038] [百度学术]
Paul F, Bolch T, Kääb A, Nagler T, Nuth C, Scharrer K, Shepherd A, Strozzi T, Ticconi F, Bhambri R, Berthier E, Bevan S, Gourmelen N, Heid T, Jeong S, Kunz M, Lauknes T R, Luckman A, Boncori J P M, Moholdt G, Muir A, Neelmeijer J, Rankl M, Van Looy J and Van Niel T. 2015. The glaciers climate change initiative: methods for creating glacier area, elevation change and velocity products. Remote Sensing of Environment, 162: 408-426 [DOI: 10.1016/j.rse.2013.07.043] [百度学术]
Ronneberger O, Fischer P and Brox T. 2015. U-net: convolutional networks for biomedical image segmentation//18th International Conference on Medical Image Computing and Computer-Assisted Intervention. Munich: Springer: 234-241 [DOI: 10.1007/978-3-319-24574-4_28] [百度学术]
Scherler D, Bookhagen B and Strecker M R. 2011. Spatially variable response of Himalayan glaciers to climate change affected by debris cover. Nature Geoscience, 4(3): 156-159 [DOI: 10.1038/ngeo1068] [百度学术]
Sharma J J, Hajnsek I, Papathanassiou K P and Moreira A. 2011. Polarimetric decomposition over glacier ice using long-wavelength airborne polSAR. IEEE Transactions on Geoscience and Remote Sensing, 49(1): 519-535 [DOI: 10.1109/TGRS.2010.2056692] [百度学术]
Shi J C and Dozier J. 1993. Measurements of snow-and glacier-covered areas with single-polarization SAR. Annals of Glaciology, 17: 72-76 [百度学术]
Simonyan K and Zisserman A. 2015. Very deep convolutional networks for large-scale image recognition. arXiv:1409.1556 [DOI: 10.48550/arXiv.1409.1556] [百度学术]
Singh G, Venkataraman G, Yamaguchi Y and Park S E. 2014. Capability assessment of fully polarimetric ALOS-PALSAR data for discriminating wet snow from other scattering types in mountainous regions. IEEE Transactions on Geoscience and Remote Sensing, 52(2): 1177-1196 [DOI: 10.1109/TGRS.2013.2248369] [百度学术]
Takeuchi S and Oguro Y. 2003. A comparative study of coherence patterns in C-band and L-band interferometric SAR from tropical rain forest areas. Advances in Space Research, 32(11): 2305-2310 [DOI: 10.1016/S0273-1177(03)90558-8] [百度学术]
Thakur P K, Aggarwal S P, Arun G, Sood S, Kumar A S, Mani S and Dobhal D P. 2017. Estimation of snow cover area, snow physical properties and glacier classification in parts of western Himalayas using C-band SAR data. Journal of the Indian Society of Remote Sensing, 45(3): 525-539 [DOI: 10.1007/s12524-016-0609-y] [百度学术]
Usami N, Muhuri A, Bhattacharya A and Hirose A. 2016. PolSAR wet snow mapping with incidence angle information. IEEE Geoscience and Remote Sensing Letters, 13(12): 2029-2033 [DOI: 10.1109/LGRS.2016.2621891] [百度学术]
van Zyl J J, Arii M and Kim Y. 2011. Model-based decomposition of polarimetric SAR covariance matrices constrained for nonnegative eigenvalues. IEEE Transactions on Geoscience and Remote Sensing, 49(9): 3452-3459 [DOI: 10.1109/TGRS.2011.2128325] [百度学术]
Vijay S and Braun M. 2018. Early 21st century spatially detailed elevation changes of Jammu and Kashmir glaciers (Karakoram-Himalaya). Global and Planetary Change, 165: 137-146 [DOI: 10.1016/j.gloplacha.2018.03.014] [百度学术]
Yamaguchi Y, Moriyama T, Ishido M and Yamada H. 2005. Four-component scattering model for polarimetric SAR image decomposition. IEEE Transactions on Geoscience and Remote Sensing, 43(8): 1699-1706 [DOI: 10.1109/TGRS.2005.852084] [百度学术]
Yao G H, Ke C Q, Zhou X B, Lee H, Shen X Y and Cai Y. 2020. Identification of alpine glaciers in the central Himalayas using fully polarimetric L-band SAR data. IEEE Transactions on Geoscience and Remote Sensing, 58(1): 691-703 [DOI: 10.1109/TGRS.2019.2939430] [百度学术]
Zhao Q H, Guo S B, Li X L and Li Y. 2018. Polarimetric SAR sea ice classification based on target decompositional features. Acta Geodaetica et Cartographica Sinica, 47(12): 1609-1620 [百度学术]
赵泉华, 郭世波, 李晓丽, 李玉. 2018. 利用目标分解特征的全极化SAR海冰分类. 测绘学报, 47(12): 1609-1620 [DOI: 10.11947/j.AGCS.2018.20170551] [百度学术]
相关作者
相关机构