A FRACTURE ZONE EXTRACTION METHOD FOR LIDAR POINT CLOUD BASED ON MULTI-SCALE NEURAL NETWORK WITH RS-CONV

doi:10.3969/j.issn.0253-4967.2024.03.013

Abstract

Abstract:

Fracture zones are geological formations resulting from the strong movement of the Earth’s crust, often manifesting as fragile and sensitive areas. These zones are closely linked to natural disasters such as earthquakes and landslides. Accurate extraction of fracture zones is crucial for quantitative studies of earthquake faults, providing a scientific basis for risk assessment and decision-making in earthquake prevention and mitigation. Thus, an in-depth study to determine their distribution patterns and surface geometry is essential for understanding earthquake dynamics and mechanisms.

This paper addresses the shortcomings of existing methods in extracting fracture zones from LiDAR point clouds, which often suffer from incomplete extraction, poor continuity, and high error rates. We propose a method based on a multi-scale neural network with RS-Conv to improve the automatic extraction of fault zones in complex terrain regions. Fracture zones exhibit complex morphologies and scale features; therefore, single-scale neighborhood point sets fail to reveal their intrinsic structural information fully. Our approach begins by constructing neighborhood point sets at different spatial scales to comprehensively examine geometric features at various levels within the point cloud. The RS-Conv operator effectively portrays the spatial relationship between the center point and neighboring points. We then build a multi-scale neural network model using the RS-Conv operator as the convolution module. This model captures the spatial relationships in the point cloud, efficiently extracting deep features at different scales. The extracted multi-scale features are concatenated to form a richer and more comprehensive feature representation, which is inputted into a fully connected layer to classify the centroid and solve the fracture zone extraction problem. We compared our method with the Tensor Decomposition and Deep Neural Networks(DNN)methods using the ISPRS point cloud dataset, the Sichuan-Yunnan point cloud dataset, and the Xianshuihe dataset. Results show that our method achieves the highest classification accuracy across all three datasets. Specifically, our method’s total classification error is only 0.3%, a reduction of 0.91% -2.79%compared to other methods. This significant error reduction demonstrates the accuracy, stability, and reliability of our proposed method in handling complex point cloud data. The main conclusions of this study are as follows:

(1)The construction of neighborhood point sets at different scales reveals that the combination of these scales significantly impacts the model’s classification performance. Selecting appropriate scale combinations is crucial for optimizing the model’s classification accuracy, facilitating better distinction between fracture zone points and non-fracture zone points.

(2)Compared to traditional and machine learning methods, the deep learning network model developed in this study shows significant advantages in extracting fracture zones from point clouds. The model can automatically learn deep features from point cloud data and process large-scale, high-dimensional point cloud datasets, thereby achieving more accurate fracture zone extraction in complex terrain conditions.

(3)Comparative experiments on different datasets further demonstrate the proposed method’s generalization ability. It is effective not only in extracting fracture zones under single terrain conditions but also in maintaining stable performance across multiple terrain conditions. This adaptability enhances the extraction of fracture zones in various terrain scenarios.

In conclusion, the method proposed in this paper offers a novel approach to fracture zone extraction. It achieves higher classification accuracy compared to existing traditional and machine learning methods, effectively addressing the challenge of fracture zone extraction in complex terrain areas.

Key words: LiDAR point cloud, multi-scale neighborhood point sets, deep learning, fracture zone extraction

摘要：

断裂带与地震、滑坡等自然灾害的发生有着密切关系, 精准提取断裂带不仅可为地震断层的定量化研究提供指导, 还可为地震灾害风险评估及防震减灾决策的制定提供科学依据。针对现有方法中LiDAR点云断裂带提取不完整、连续性差及错误率高等问题, 文中提出了一种基于 RS-Conv 的多尺度神经网络LiDAR点云断裂带提取方法, 以便更好地解决复杂地形区域的断裂带自动提取问题。该方法首先构建不同空间尺度的邻域点集, 从而更全面地考察点云的局部几何结构特征。考虑到RS-Conv算子能够很好地表征中心点与邻域点的空间关系, 文中以RS-Conv算子作为卷积模块构建了多尺度神经网络模型, 以提取出LiDAR点云不同尺度的深层次特征, 对其进行堆叠并输入到全连接层, 以完成对断裂带点的提取。最后, 在ISPRS点云数据集、川滇点云数据集和鲜水河数据集上对文中所述方法与张量分解方法和Deep Neural Networks(DNN)方法进行了对比实验, 结果表明, 文中方法的分类精度最高, 分类总误差最低仅为0.3%, 较其他方法降低了0.91%~2.79%, 证实了该方法在点云断裂带提取方面的优越性。

关键词: LiDAR点云, 多尺度邻域点集, 深度学习, 断裂带提取

SONG Dong-mei, WANG Hao, FENG Jia-xing, SHAN Xin-jian, WANG Bin. A FRACTURE ZONE EXTRACTION METHOD FOR LIDAR POINT CLOUD BASED ON MULTI-SCALE NEURAL NETWORK WITH RS-CONV[J]. SEISMOLOGY AND GEOLOGY, 2024, 46(3): 739-755.

宋冬梅, 王浩, 冯家兴, 单新建, 王斌. 基于RS-Conv的多尺度神经网络LiDAR点云断裂带提取方法[J]. 地震地质, 2024, 46(3): 739-755.

Figures/Tables 18

Fig. 1 Schematic diagram of RS-Conv operator framework.

Fig. 2 Flowchart of LiDAR point cloud fracture zone extraction method.

Fig. 3 Two-dimensional schematic diagram of the neighborhood search.

Fig. 4 Two-dimensional schematic diagram of multi-scale neighborhood search.

Fig. 5 Schematic diagram of MS RS-Conv point cloud fracture zone extraction framework.

Fig. 6 Sample display diagrams of ISPRS point cloud dataset.

Fig. 7 Sample display diagrams of Chuandian point cloud dataset.

Fig. 8 Sample display diagrams of Xianshuihe point cloud dataset.

Table1 The confusion matrix of classification results and error calculation method

分类结果			预测
		断裂带点	非断裂带点
真值	断裂带点	a	b
	非断裂带点	c	d
Ⅰ类误差		c/(c+d)
Ⅱ类误差		b/(a+b)
总误差		(b+c)/(a+b+c+d)

Table2 Results of fracture zone points extraction based on different search strategies and the number of neighborhood points

模型	邻域点搜索策略	邻域点个数	总误差/%	提取精度/%
MS RS-Conv	kNN	64、128、256	1.27	98.73
		64、128、512	0.52	99.48
		64、256、1024	1.24	98.76
	RPIB	64、128、256	1.46	98.54
		64、128、512	0.99	99.01
		64、256、1024	1.51	98.49

Table3 Performance comparison of three fracture zone extraction methods on ISPRS dataset

数据集	方法	Ⅰ类误差/%	Ⅱ类误差/%	总误差/%	提取精度/%
Samp51	张量分解	1.22	9.31	2.82	97.18
	DNN	0.46	4.95	1.35	98.65
	MS RS-Conv	0.41	2.56	0.86	99.14
Samp53	张量分解	2.43	17.37	3.83	96.17
	DNN	1.07	11.39	2.03	97.97
	MS RS-Conv	0.19	10.27	1.04	98.96

Fig. 9 The results of three fracture zone extraction methods on Samp51.

Fig. 10 The results of three fracture zone extraction methods on Samp53.

Table4 Performance comparison of three fracture zone extraction methods on Chuandian dataset

数据集	方法	Ⅰ类误差/%	Ⅱ类误差/%	总误差/%	提取精度/%
CD_1	张量分解	1.15	1.47	1.32	98.68
	DNN	0.77	0.97	0.88	99.12
	MS RS-Conv	0.29	0.50	0.41	99.59
CD_2	张量分解	1.39	1.31	1.36	98.64
	DNN	0.63	2.01	1.28	98.72
	MS RS-Conv	0.21	0.52	0.36	99.64

Fig. 11 The results of three fracture zone extraction methods on CD_1.

Fig. 12 The results of three fracture zone extraction methods on CD_2.

Table5 Performance comparison of three fracture zone extraction methods on Xianshuihe dataset

数据集	方法	Ⅰ类误差/%	Ⅱ类误差/%	总误差/%	提取精度/%
鲜水河点云断裂带	张量分解	7.74	1.07	2.13	97.87
	DNN	5.11	0.91	1.58	98.42
	MS RS-Conv	1.19	0.13	0.30	99.70

Fig. 13 The results of three fracture zone extraction methods on Xianshuihe fracture zone.

References 29

[1]	陈梦莹. 2019. 基于高分辨率无人机航测数据对郯庐断裂带沂沭段活动性的研究[D]. 北京: 中国地质大学.
	CHEN Meng-ying. 2019. Research on the activity of YiShu section in Tan Lu fault zone based on aerial survey data of high resolution UAV[D]. China University of Geosciences(Beijing), Beijing (in Chinese).
[2]	高帅坡. 2017. 基于无人机摄影测量技术的活动构造定量参数提取研究[D]. 北京: 中国地震局地质研究所.
	GAO Shuai-po. 2017. A quantitative parameters extraction study of active tectonics based on UAV photogrammetry technology[D]. Institute of Geology, China Earthquake Administration, Beijing (in Chinese).
[3]	郭文周, 邓宇, 何庐山, 等. 2017. 航摄与机载激光雷达技术在河道地形测量中的应用[J]. 水资源研究, 6(6): 7.
	GUO Wen-zhou, DENG Yu, HE Lu-shan, et al. 2017. Application of aerial photography and airborne LiDAR in river topography survey[J]. Journal of Water Resources Research, 6(6): 7 (in Chinese).
[4]	林洪彬, 刘彬, 张玉存. 2012. 基于多尺度张量分解的点云结构特征提取[J]. 中国机械工程, 23(15): 1833—1839.
	LIN Hong-bin, LIU Bin, ZHANG Yu-cun. 2012. Feature extraction of point cloud structure based on multi-scale tensor decomposition[J]. China Mechanical Engineering, 23(15): 1833—1839 (in Chinese).
[5]	彭检贵, 马洪超, 王宗跃, 等. 2010. 机载LiDAR点云的双阈值自动提取断裂线方法[J]. 测绘科学技术学报, 27(4): 5.
	PENG Jian-gui, MA Hong-chao, WANG Zong-yue, et al. 2010. Double threshold automatic extraction of fracture line from airborne LiDAR point cloud[J]. Journal of Geomatics Science and Technology, 27(4): 5 (in Chinese).
[6]	席英杰, 李克文, 徐延辉, 等. 2021. 一种用于地震断层图像识别的SPD-UNet模型[J]. 计算机工程, 47(12): 249—255.
	XI Ying-jie, LI Ke-wen, XU Yan-hui, et al. 2021. An SPD-UNet model for seismic fault image recognition[J]. Computer Engineering, 47(12): 249—255 (in Chinese).
[7]	谢小平, 白毛伟, 陈芝聪, 等. 2019. 龙门山断裂带北东段活动断裂的遥感影像解译及构造活动性分析[J]. 国土资源遥感, 31(1): 237—246.
	XIE Xiao-ping, BAI Mao-wei, CHEN Zhi-cong, et al. 2019. Remote sensing image interpretation and tectonic activity analysis of active fractures in the north-eastern section of the Longmenshan fault zone[J]. Remote Sensing for Land & Resources, 31(1): 237—246 (in Chinese).
[8]	徐景中, 万幼川, 张圣望. 2008. 基于机载激光雷达点云的断裂线自动提取方法[J]. 计算机应用, 28(5): 1214—1216.
	XU Jing-zhong, WAN You-chuan, ZHANG Sheng-wang. 2008. Automatic breakline extraction from LiDAR point clouds[J]. Journal of Computer Applications, 28(5): 1214—1216 (in Chinese).
[9]	徐彦怀. 2016. 基于卫星遥感影像和DEM融合的地质断裂带研究[J]. 测绘与空间地理信息, 39(11): 135—138.
	XU Yan-huai. 2016. The study on geological fault zone base on the merged Image between the satellite remote sensing Image and DEM[J]. Geomatics & Spatial Information Technology, 39(11): 135—138 (in Chinese).
[10]	张志, 张雪亭. 1999. 基于遥感影像信息的东昆仑活动断裂带研究[J]. 西安工程学院学报, 21(3): 5—8.
	ZHANG Zhi, ZHANG Xue-ting, 1999. Study on the East Kunlun active fault zone based on remote sensing image information[J]. Journal of Xi'an Polytechnic University, 21(3): 5—8 (in Chinese).
[11]	Chehrazi A, Rahimpour-Bonab H, Rezaee M R. 2013. Seismic data conditioning and neural network-based attribute selection for enhanced fault detection[J]. Petroleum Geoscience, 19(2): 169—183.
[12]	Chen C, Guo J, Wu H, et al. 2021. Performance comparison of filtering algorithms for high-density airborne LiDAR point clouds over complex Land Scapes[J]. Remote Sensing, 13(14): 2663.
[13]	Chen Z, Gao B, Devereux B. 2017. State-of-the-art: DTM generation using airborne LiDAR data[J]. Sensors, 17(1): 150.
[14]	Chu H J, Wang C K, Huang M L, et al. 2014. Effect of point density and interpolation of LiDAR-derived high-resolution DEMs on landscape scarp identification[J]. GIScience & Remote Sensing, 51(6): 731—747.
[15]	Liu Y, Fan B, Xiang S, et al. 2019. Relation-shape convolutional neural network for point cloud analysis [C]∥Proceedings of the IEEE/CVF conference on computer vision and pattern recognition: 8895—8904.
[16]	Martins V S, Kaleita A L, Gelder B K, et al. 2020. Deep neural network for complex open-water wetland mapping using high-resolution WorldView-3 and airborne LiDAR data[J]. International Journal of Applied Earth Observation and Geoinformation, 93:102215.
[17]	Maturana D, Scherer S. 2015. Voxnet: A 3D convolutional neural network for real-time object recognition[C]. In 2015 IEEE/RSJ international conference on intelligent robots and systems(IROS), Hamburg: 922—928.
[18]	Meng X, Currit N, Zhao K. 2010. Ground filtering algorithms for airborne LiDAR data: A review of critical issues[J]. Remote Sensing, 2(3): 833—860.
[19]	Qi C R, Su H, Mo K, et al. 2017a. Pointnet: Deep learning on point sets for 3D classification and segmentation[C]. In 2017 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Honolulu, HI: 652—660.
[20]	Qi C R, Su H, Nieβner M, et al. 2016. Volumetric and multi-view CNNs for object classification on 3d data[C]. In 2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR), Las Vegas, NV: 5648—5656.
[21]	Qi C R, Yi L, Su H, et al. 2017b. Pointnet++: Deep hierarchical feature learning on point sets in a metric space[C]. In 2017 Advances in Neural Information Processing Systems, California: 5105—5114.
[22]	Rottensteiner F, Sohn G, Gerke M, et al. 2014. Results of the ISPRS benchmark on urban object detection and 3D building reconstruction[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 93: 256—271.
[23]	Su H, Maji S, Kalogerakis E, et al. 2015. Multi-view convolutional neural networks for 3D shape recognition[C]. In 2015 IEEE International Conference on Computer Vision(ICCV). Santiago, Chile: 945—953.
[24]	Wang R. 2013. 3D building modeling using images and LiDAR: A review[J]. International Journal of Image and Data Fusion, 4(4): 273—292.
[25]	Wang R, Peethambaran J, Chen D. 2018. LiDAR point clouds to 3-D urban models: A review[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 11(2): 606—627.
[26]	Wu X, Liang L, Shi Y, et al. 2019. FaultSeg3D: Using synthetic data sets to train an end-to-end convolutional neural network for 3D seismic fault segmentation[J]. Geophysics, 84(3): IM35—IM45.
[27]	Yang B, Huang R, Dong Z, et al. 2016. Two-step adaptive extraction method for ground points and breaklines from LiDAR point clouds[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 119: 373—389.
[28]	Zhou Y, Parsons B, Elliott J R, et al. 2015. Assessing the ability of Pleiades stereo imagery to determine height changes in earthquakes: A case study for the El Mayor-Cucapah epicentral area[J]. Journal of Geophysical Research: Solid Earth, 120(12): 8793—8808.
[29]	Zhou Y, Tuzel O. 2018. Voxelnet: End-to-end learning for point cloud based3D object detection[C]. In 2018 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Salt Lake City, UT: 4490—4499.