Intelligent identification of landslide disaster based on deep learning of UAV images

doi:10.16265/j.cnki.issn1003-3033.2024.07.2092

Abstract

Abstract:

An open-pit mine landslide identification method was proposed based on object-oriented annotation datasets and the Res-U-Net model to realize accurate identification and early warning of open-pit mile landslide disasters. Firstly, the mine landslide image data in the study area were obtained by UAV aerial survey. Secondly, the multi-scale-spectral segmentation method and threshold separation principle were applied to divide and classify the open-pit mine landslide data, and the landslide dataset was developed based on the object-oriented method. Then, the U-Net network was used as the infrastructure to propose a landslide identification semantic segmentation model based on Res-U-Net by integrating the residual module into each convolutional layer. Finally, the datasets constructed by different methods were used to identify landslides, and the Res-U-Net model was compared with the widely used semantic segmentation models, Fully Convolutional Networks (FCN), and U-net. The results indicated that the landslide data set based on object-oriented annotation had better landslide identification performance when compared to the traditional manual annotation dataset, resulting in improvements in identification accuracy, recall rate, F₁ score, and kappa coefficient of more than 12%. The landslide identification accuracy of the Res-U-Net model was more than 0.8, realizing the accurate landslide open-pit mine disaster identification.

Key words: unmanned aerial vehicle image, deep learning, landslide disaster, intelligent identification, object oriented, Res-U-Net

CLC Number:

X915.5

JIANG Song, LI Yanbo, HE Xuqian, HE Runfeng, ZHANG Chao, ZHANG Cunliang. Intelligent identification of landslide disaster based on deep learning of UAV images[J]. China Safety Science Journal, 2024, 34(7): 229-238.

Figures/Tables 20

Table 1

Checkpoint horizontal accuracy statisticsm

序号	测试点			像控点
序号	X	Y	$H 0$	x	y	H
1	545 937.520	3 753 813.645	1 322.815	545 937.524	3 753 813.639	1 322.793
2	545 661.386	3 753 701.391	1 284.536	545 661.377	3 753 701.379	1 284.505
3	545 754.316	3 753 480.454	1 225.059	545 754.308	3 753 480.444	1 225.031
4	546 031.346	3 753 592.878	1 351.895	546 031.338	3 753 592.870	1 351.869

Table 1

Table 2

Checkpoint elevation accuracy statisticsm

序号	较差
序号	$Δ x$	$Δ y$	$Δ S$	$Δ H$
1	-0.004	0.006	0.007	0.022
2	0.009	0.012	0.015	0.031
3	0.008	0.010	0.012	0.028
4	0.006	0.008	0.010	0.026

Table 2

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Table 3

Fig.6

Fig.7

Fig.8

Fig.9

Fig.10

Table 4

Table 5

Table 6

Fig.11

Fig.12

Fig.13

Table 7

References 16

[1]	曹永胜, 解治宇, 于庆磊, 等. 大孤山露天矿西北帮边坡稳定性分析及治理措施研究[J]. 金属矿山, 2020(1): 141-150.
	CAO Yongsheng, XIE Zhiyu, YU Qinglei, et al. Study on the governance measures and slope stability on the northwest slope of Dagushan open-pit mine[J]. Metal Mine, 2020(1): 141-150.
[2]	国家矿山安全监察局. 甘肃白银泓胜煤业有限责任公司(露天煤矿)“7·23”重大边坡坍塌事故案例[EB/OL]. (2023-09-28). https://www.chinamine-safety.gov.cn/zfxxgk/fdzdgknr/sgcc/sgalks/.
[3]	国家矿山安全监察局. 内蒙古阿拉善新井煤业有限公司露天煤矿“2·22”特别重大坍塌事故案例[EB/OL]. (2023-09-28). https://www.chinamine-safety.gov.cn/zfxxgk/fdzdgknr/sgcc/sgalks/.
[4]	BARLOW J, MARTIN Y, FRANKLIN S E. Detecting translational landslide scars using segmentation of landsat ETM+ and DEM data in the northern cascade mountains, British Columbia[J]. Canadian Journal of Remote Sensing, 2003, 29(4): 510-517.
[5]	JU Yuanzhen, XU Qiang, JIN Shichao, et al. Loess landslide detection using object detection algorithms in northwest China[J]. Remote Sensing, 2022, 14(5): DOI:10.3390/RS14051182.
[6]	JI Shunping, YU Dawen, SHEN Chaoyong, et al. Landslide detection from an open satellite imagery and digital elevation model dataset using attention boosted convolutional neural networks[J]. Landslides, 2020, 17(6): 1337-1352.
[7]	张蕴灵, 傅宇浩, 孙雨, 等. 结合深度神经网络的高分辨遥感影像滑坡检测[J]. 公路, 2021, 66(5): 188-194.
	ZHANG Yunling, FU Yuhao, SUN Yu, et al. Landslide detection from high-resolution remote sensing image using deep neural network[J]. Highway, 2021, 66(5): 188-194.
[8]	吴琪, 周创兵, 黄发明, 等. 基于双重注意力机制的滑坡识别方法优化[J]. 地质科技通报, 2022, 41(2): 246-253.
	WU Qi, ZHOU Chuangbing, HUANG Faming, et al. Optimiation of the landslide identification method based on a dual attention mechanism[J]. Bulletin of Geological Science and Technology, 2022, 41(2): 246-253.
[9]	巨袁臻, 许强, 金时超, 等. 使用深度学习方法实现黄土滑坡自动识别[J]. 武汉大学学报:信息科学版, 2020, 45(11): 1747-1755.
	JU Yuanzhen, XU Qiang, JIN Shichao, et al. Automatic object detection of loess landslide based on deep learning[J]. Geomatics and Information Science of Wuhan University, 2020, 45(11): 1747-1755.
[10]	席思远, 张西童, 王宁, 等. 倾斜摄影设备选型及像控点布设对高精度实景三维模型重建的影响[J]. 测绘通报, 2022(10): 86-92. doi: 10.13474/j.cnki.11-2246.2022.0299
	XI Siyuan, ZHANG Xitong, WANG Ning, et al. Influence of oblique photography equipment selection and image control point layout on high-precision 3D real scene model reconstruction[J]. Bulletin of Surveying and Mapping, 2022(10): 86-92. doi: 10.13474/j.cnki.11-2246.2022.0299
[11]	王朝辉, 吴昊, 孟将. 无人机倾斜摄影像控布点方案研究及精度分析[J]. 测绘通报, 2021(5):102-105,110. doi: 10.13474/j.cnki.11-2246.2021.0151
	WANG Zhaohui, WU Hao, MENG Jiang. Research and accuracy analysis of image control point arrangement schemes for UAV oblique photography[J]. Bulletin of Surveying and Mapping, 2021(5): 102-105,110. doi: 10.13474/j.cnki.11-2246.2021.0151
[12]	汤国安. 我国数字高程模型与数字地形分析研究进展[J]. 地理学报, 2014, 69(9): 1305-1325. doi: 10.11821/dlxb201409006
	TANG Guoan. Progress of DEM and digital terrain analysis in China[J]. Acta Geographica Sinica, 2014, 69(9): 1305-1325. doi: 10.11821/dlxb201409006
[13]	豆红强, 黄思懿, 简文彬, 等. 基于遥感数据的闽东南山区公路滑坡快速识别技术研究[J]. 自然灾害学报, 2023, 32(1): 217-227.
	DOU Hongqiang, HUANG Siyi, JIAN Wenbin, et al. Research on rapid identification technology of highway landslide in mountainous areas of southeast Fujian based on remote sensing data[J]. Journal of Natural Disasters, 2023, 32(1): 217-227.
[14]	卢冰, 李灿林, 冯薛龙, 等. 基于改进SRResNet深度学习网络的低照度图像超分辨率重建方法[J]. 科学技术与工程, 2022, 22(27): 12 045-12 052.
	LU Bing, LI Canlin, FENG Xuelong, et al. Super-resolution reconstruction method for low illumination images based on improved SRResNet deep learning network[J]. Science Technology and Engineering, 2022, 22(27): 12 045-12 052.
[15]	刘佳, 伍宇明, 高星, 等. 基于GEE和U-net模型的同震滑坡识别方法[J]. 地球信息科学学报, 2022, 24(7): 1275-1285. doi: 10.12082/dqxxkx.2022.210704
	LIU Jia, WU Yuming, GAO Xing, et al. Image recognition of co-seismic landslide based on GEE and U-net neural network[J]. Journal of Geo-information Science, 2022, 24(7): 1275-1285.
[16]	SHAO Zhengfeng, ZHOU Zifan, HUANG Xiao, et al. MRENet: Simultaneous extraction of road surface and road centerline in complex urban from very high-resolution images[J]. Remote Sensing, 2021, 13(2): DOI:10.3390/RS13020239.

特征类别	名称	代表含义	取值范围
光谱特征	L1-R	红色光谱	[32.22,215.45]
	L2-G	绿色光谱	[36.81,217.2]
	L3-B	蓝色光谱	[29.76,214.56]
	Brightness	亮度	[94.96,196.02]
	Max_diff	最大化差异特征	[0.227,2.247]
形状特征	r	长宽比	[1.003,22.41]
纹理特征	GLCM	灰度共生矩阵	[0.095,0.555]
地形特征	L4-Aspect	坡向	[54.89,240.24]
	L5-Curv	曲率	[97.79,113.41]
	L6-Sr	地形起伏度	[53.36,255]
	L7-Slope	坡度	[104.79,226.08]

模型结构	层数	操作	尺寸	步长	边缘填充	输出尺寸
输入层	Input_1	输入影像数据	3×3/1×1	1	—	1 024×1 024×3
输入层	Input_2	输入地形数据	3×3/1×1	1	—	1 024×1 024×4
编码层	Level_1	卷积	3×3/1×1	1	2/0	1 024×1 024×64
	Level_1	池化	2×2	2	0	512×512×64
	Level_2	卷积	3×3/1×1	1	2/0	512×512×128
	Level_2	池化	2×2	2	0	256×256×128
	Level_3	卷积	3×3/1×1	1	2/0	256×256×256
	Level_3	池化	2×2	2	0	128×128×256
	Level_4	卷积	3×3/1×1	1	2/0	128×128×512
	Level_4	池化	2×2	2	0	64×64×512
	Level_5	卷积	3×3/1×1	1	2/0	64×64×1024
解码层	Level_6	上采样	2×2	2	0	128×128×512
		跳跃连接	—	—	—	128×128×1024
		卷积	3×3/1×1	1	2/0	128×128×512
	Level_7	上采样	2×2	2	0	256×256×256
		跳跃连接	—	—	—	256×256×512
		卷积	3×3/1×1	1	2/0	256×256×256
	Level_8	上采样	2×2	2	0	512×512×128
		跳跃连接	—	—	—	512×512×256
		卷积	3×3/1×1	1	2/0	512×512×128
	Level_9	上采样	2×2	2	0	1 024×1 024×64
		跳跃连接	—	—	—	1 024×1 024×128
		卷积	3×3/1×1	1	2/0	1 024×1 024×64
输出层	Output_10	输出数据	1×1	1	0	1 024×1 024×1

混淆矩阵	参考结果
分类结果	—	滑坡	非滑坡	合计
	滑坡	TP	FP	TP+FP
	非滑坡	FN	TN	FN+TN
	合计	TP+FN	FP+TN	TP+FP+ FN+TN

数据输入	P	R	F₁	kappa
第1组	0.597	0.588	0.593	0.562
第2组	0.675	0.661	0.664	0.636
第3组	0.863	0.852	0.858	0.847

网络模型	P	R	F₁	kappa
U-Net	0.786	0.766	0.775	0.758
FCN	0.851	0.623	0.704	0.721
Res-U-Net	0.863	0.852	0.858	0.847