Emergency monitoring of remote sensing for flood inundation region based on SAR texture and LightGBM

doi:10.3969/j.issn.1000-5641.2023.03.009

Abstract

Abstract:

In response to the need for high timeliness and accuracy monitoring for inundation region during flood disaster, a new extraction method of water areas based on SAR texture and LightGBM was proposed. Compared with other methods, such as the SDWI water index, SVM, RF and GBDT methods, it shows that the accuracy of water extraction of river, lake and flooded area is beyond 98% and higher than other methods. Meanwhile, the operating efficiency of the proposed method is 20 ~ 100 times higher than other methods, which greatly improves the timeliness of inundation emergency monitoring during flood disaster.

Key words: flood disaster, SAR remote sensing, image texture, LightGBM, water extraction

CLC Number:

TP79

Cheng SUN, Fang SHEN, Rugang TANG. Emergency monitoring of remote sensing for flood inundation region based on SAR texture and LightGBM[J]. Journal of East China Normal University(Natural Science), 2023, 2023(3): 82-92.

Figures/Tables 12

Fig.1

Fig.2

Table 1

Expressions of texture feature and their description"

纹理特征	表达式	描述
均值	${{\rm{ME}} = }\sum\nolimits_{i,j = 0}^{N - 1} {i{P_{i,j} } }$	均值反映图像灰度局部变化情况, 从河流两侧到河流中央, 灰度值快速降低到低值.
方差	${{\rm{VA}} = }\sum\nolimits_{i,j = 0}^{N - 1} {i{P_{i,j} } } {(i - {{\rm{ME}}})^2}$	方差是像元素与均值偏差的度量, 衡量图像分布的分散和变化情况, 反映纹理的强度信息.
均质性	${{\rm{HOM}} = }\sum\nolimits_{i = 0}^{N - 1} {\sum\nolimits_{j = 0}^{N - 1} { {P_{i,j} } } /(1 + { {(i - j)}^2})}$	均质性反映图像纹理的同质性, 用来度量图像纹理局部的变化量. 同质性最高值时, 纹理最稀疏; 同质性取最低值时, 纹理最密集. 水体区块为影像中的高同质区, 纹理稀疏.
反差	${ {\rm{CON} } = }\sum\nolimits_{n = 0}^{N - 1} {n_{}^2} \left\{ {\sum\nolimits_{i = 0}^{N - 1} {\sum\nolimits_{ j = 0,\atop \\ \left\| {i - j} \right\| = n }^{\,N - 1} { {P_{i,j} } } } } \right\}$	反差反映图像的清晰度和纹理沟纹的深浅程度, 值越大, 纹理沟纹越深, 说明在图像中亮度值变化越快, 越能突出水体边缘区域快速的亮度变化.
相异度	${{\rm{DI}} = }\sum\nolimits_{i,j = 0}^{N{- }1} {i{P_{i,j} }\left\| {i - j} \right\|}$	相异性与反差呈线性正相关, 在河流边远地区局部反差大, 河流中央地区局部反差小.
熵	${{\rm{ENT}} = }\sum\nolimits_{i = 0}^{N - 1} {\sum\nolimits_{j = 0}^{N - 1} { {P_{i,j} }\lg {P_{i,j} } } }$	熵表示图像中纹理的非均匀程度或复杂程度, 重点反映图像灰度的变化速度, 水体中央的熵较小, 在复杂的岸边熵值较大.
角二阶矩	${ {\rm{ASM} } = }\sum\nolimits_{i,j = 0}^{N - 1} {i{P_{i,j}^2 }}$	角二阶矩是GLCM元素值的平方和, 反映了图像灰度分布均匀程度和纹理粗细度, 重点突出大型面状水体内部纹理均匀程度较低的特点.
相关性	${ {\rm{COR} } = }\left\| {\sum\nolimits_{i = 0}^{N - 1} {\sum\nolimits_{j = 0}^{N - 1} {ij{P_{i,j} } - {u_1}{u_2} } } } \right\| /({\sigma _1}{\sigma _2})$	相关性反映图像纹理的一致性, 度量GLCM在行或列方向熵的相似程度, 值越大, 代表矩阵元素越均匀, 突出水体内部无稳定变化区域.

Table 1

Fig.3

Table 2

Fig.4

Table 3

Fig.5

Fig.6

Fig.7

Table 4

Fig.8

References 26

1	DANIEL A, GERALD F, VALENTIJN V. Modeling flood hazard zones at the sub-district level with the rational model integrated with GIS and remote sensing approaches. Water, 2015, 7 (7): 3531- 3564.
2	李香颜, 陈怀亮, 李友. 洪水灾害卫星遥感监测与评估研究综述. 中国农业气象, 2009, 30 (1): 102- 108.
3	孙亚勇, 李小涛, 杨锋杰, 等. 基于星载SAR数据的山区水体提取方法研究. 中国水利水电科学研究院学报, 2014, 12 (3): 258- 263.
4	贾诗超, 薛东剑, 李成绕, 等. 基于Sentinel-1数据的水体信息提取方法研究 [J]. 人民长江, 2019, 50(2): 213-217.
5	杨存建, 周成虎, 魏一鸣, 等. 星载SAR图像中居民地专题信息增强的方法探讨. 遥感技术与应用, 1998, (4): 5- 8.
6	LÓPEZ-CALOCA A A, TAPIA-SILVA F O, LÓPEZ F, et al. Analyzing short term spatial and temporal dynamics of water presence at a basin-scale in Mexico using SAR data. GIScience & Remote Sensing, 2020, 57 (7): 985- 1004.
7	SANDRA B, DOUG S, BRIAN B, et al. Operational surface water detection and monitoring using radarsat 2. Remote Sensing, 2016, 8 (4): 285.
8	LYU W T, YU Q Z, YU W X. Water extraction in SAR images using GLCM and support vector machine [C]// IEEE International Conference on Signal Processing. IEEE, 2010.
9	胡德勇, 李京, 陈云浩, 等. 单波段单极化SAR图像水体和居民地信息提取方法研究. 中国图象图形学报, 2008, (2): 257- 263.
10	吴永辉, 计科峰, 郁文贤. 利用SVM的全极化、双极化与单极化SAR图像分类性能的比较. 遥感学报, 2008, (1): 46- 53.
11	童莹萍, 冯伟, 全英汇, 等. 基于GF-3与Sentinel-2A图像融合的高精度水体提取算法. 无线电工程, 2021, 51 (8): 750- 753.
12	孟然, 沈蔚, 纪茜, 等. GBDT模型在遥感水深反演中的应用. 环境生态学, 2021, 3 (5): 1- 5.
13	张长水. 机器学习面临的挑战. 中国科学:信息科学, 2013, 43 (12): 1612- 1623.
14	欧东璟. 基于多分类器融合的高光谱遥感图像分类 [D]. 济南: 山东大学, 2019.
15	刘新伟, 黄武斌, 蒋盈沙, 等. 基于LightGBM算法的强对流天气分类识别研究. 高原气象, 2021, (4): 909- 918.
16	张国庆, 昌宁. 基于LightGBM的银行信用卡违约研究. 科技资讯, 2019, 17 (12): 8- 9.
17	王恒, 姜亚楠, 张欣, 等. 基于梯度提升算法的岩性识别方法. 吉林大学学报(地球科学版), 2021, 51 (3): 940- 950.
18	IAIN H W. 微波遥感导论 [M]. 北京: 科学出版社, 2014: 3.
19	HARALICK R M, SHANMUGAM K, DINSTEIN I. Textural features for image classification. Studies in Media and Communication, 1973, SMC-3 (6): 610- 621.
20	GONG M, LI Y, JIAO L, et al. Sar change detection based on intensity and texture changes. ISPRS Journal of Photogrammetry and Remote Sensing, 2014, 93, 123- 135.
21	GONG L X, WANG C, WU F, et al. Earthquake-induced building damage detection with post-event sub-meter VHR terraSAR-X staring spotlight Imagery. Remote Sensing, 2016, 8 (11): 887- 888.
22	胡召玲, 李海权, 杜培军. SAR图像纹理特征提取与分类研究. 中国矿业大学学报, 2009, 38 (3): 422- 427.
23	李强, 张景发, 龚丽霞, 等. SAR图像纹理特征相关变化检测的震害建筑物提取. 遥感学报, 2018, 22 (S1): 128- 138.
24	马晓君, 沙靖岚, 牛雪琪. 基于LightGBM算法的P2P项目信用评级模型的设计及应用. 数量经济技术经济研究, 2018, 35 (5): 144- 160.
25	张梓童, 王恒, 高世臣, 等. 基于特征选择和梯度提升算法的高光谱遥感地物识别. 数学的实践与认识, 2019, 49 (22): 187- 196.
26	杨魁, 杨建兵, 江冰茹. Sentinel-1卫星综述. 城市勘测, 2015, (2): 24- 27.

参数	取参范围	步长	最优参数
树的最大深度	1 ~ 8	1	8
迭代次数	1 ~ 300	2	200
叶节点数	5 ~ 100	5	95
叶节点的最小样本数	10 ~ 100	5	20
最大直方图数目	10 ~ 300	5	255
学习率	0.01 ~ 0.10	0.01	0.03

极化方式	SVM		RF		GBDT		LightGBM
极化方式	精度/%	训练耗时	精度/%	训练耗时	精度/%	训练耗时	精度/%	训练耗时
VH	94.73	3:07:04	96.18	1:08:17	96.60	1:02:37	98.28	0:02:42
VV	94.21	3:20:04	95.31	1:09:36	95.86	1:02:06	97.43	0:02:41
VH + VV	95.19	4:03:49	96.23	1:38:51	96.91	1:03:43	98.40	0:02:59

区域	提取方法	总体精度/%	IoU/%	F₁指标	Kappa 系数
河道区	SDWI	97.79	70.90	0.83	0.83
	SVM	98.80	81.38	0.89	0.89
	RF	98.89	79.88	0.89	0.88
	GBDT	98.90	83.70	0.91	0.91
	WEGL	99.15	84.82	0.92	0.92
湖泊区	SDWI	98.16	79.94	0.89	0.88
	SVM	99.04	88.14	0.92	0.92
	RF	98.95	85.76	0.92	0.92
	GBDT	99.02	86.83	0.93	0.92
	WEGL	99.13	88.85	0.94	0.94
洪水淹没区	SDWI	95.01	53.96	0.70	0.68
	SVM	97.32	69.89	0.82	0.81
	RF	98.11	71.53	0.83	0.82
	GBDT	98.34	75.05	0.86	0.85
	WEGL	98.69	81.80	0.90	0.89

[1]	WANG Wen-Jing, YAN Jun-Ping, LIU Yong-Lin, CAO Yong-Wang. Climate change and droughtflood disasters responses in South China [J]. Journal of East China Normal University(Natural Sc, 2016, 2016(2): 81-89.
[2]	WANG Xiao-li;CHEN Xi-qing;ZHANG Er-feng. On Fundamental Problems about Flood Disasters and Flood Insurance in the Taihu Basin(Chinese) [J]. Journal of East China Normal University(Natural Sc, 2005, 2005(4): 43-50.