Abstract:

Field investigations of large earthquakes indicate that earthquakes with a magnitude greater than 6.5 often produce seismic surface rupture zones ranging from thousands of meters to tens of kilometers on the earth’s surface. The geometric structures of surface ruptures contain the kinematic characteristics of seismogenic structures, which can not only provide critical quantitative data for analyzing the spatial distribution law of co-seismic displacement of active faults and the width of active fault deformation zone, but also have an important significance for understanding the kinematic mechanism and deformation law of seismogenic faults.
At present, the conventional methods to obtain earthquake surface ruptures mainly include the field geological survey and visual interpretation of the remote sensing image. Although these two methods can get the rough geometry of coseismic surface ruptures, they both have certain limitations. The reliability of the field geological survey method is high. However, intra-continental earthquakes often occur in places with complicated topography, and lots of sites are difficult to reach, leading to incomplete data and failure to draw detailed features of the fracture zone. Meanwhile, the field geological survey is often time-consuming and laborious. Although the visual interpretation of remoting sensing images can be used to interpret surface fractures in areas that cannot be reached by the geological field survey, the result accuracy is vulnerable to the impacts of interpreters’ experience. The whole process of interpretation is still time-consuming and labor-intensive and the extraction results are mostly linear surface ruptures, so it is difficult to accurately obtain fine features such as the width of the surface rupture zone. Therefore, the automatic extraction of fine structures of seismic surface ruptures, especially micro-rupture surfaces, is an urgent problem in active tectonic studies.
The remote sensing images obtained through satellite platforms have low resolution and are susceptible to weather factors, and the extracted surface rupture fineness is not enough. The UAV platform, on the other hand, is low-cost to use, can fly at a low altitude, is not affected by clouds and fog, and can acquire images with a centimeter-level resolution, which provides conditions for extraction the fine structure of surface ruptures of large earthquakes. Thus, to solve the problem that it is difficult to obtain surface ruptures of large earthquakes quickly, this study proposes an object-oriented “Rough segmentation and Fine extraction” method based on object-oriented and color segmentation theories of color space, which realizes the semi-automatic extraction of features of seismic surface rupture zone. The processing workflow of the method is as follows: First, the original image is cropped by the custom irregular raster cropping method designed in this study to obtain ROI(the Region of Interest). Second, the color space of ROI is converted into HSV, and the HSV color space of ROI is segmented into surface rupture candidate area by using brightness and hue tone values. And then, the surface rupture candidate area is processed by expansion operation of binary mathematical morphology. Third, the surface rupture candidate area is transformed into a series of sub-area objects by the contour tracking method. Fourth, the surface rupture is refined using the spectral standard deviation, spectral mean and the length-width ratio of the smallest surrounding rectangle as characteristic parameters. Finally, the results are output as the vector surface of surface ruptures.
The effectiveness of the proposed method is analyzed by taking the high-resolution UAV image data of the M_S7.4 Maduo earthquake in Qinghai Province as an example. The results show that the proposed method can effectively remove the noises such as the river channel similar to the characteristics(i.e., the color and shape features)of the surface rupture and extract the delicate structures of the surface rupture zone quickly and accurately, except that several poor extraction results were caused by the limitation of image resolution and the destruction of surface rupture caused by river erosion. The extraction results are highly reliable and can be used to extract quantitative parameters of surface ruptures in large earthquakes. Thus, the semi-automatic extraction method of seismic surface ruptures established in this study can provide a feasible scheme for the rapid extraction of delicate structures from surface ruptures and analysis of surface deformation characteristics after a large earthquake.

Key words: object-oriented, color segmentation, high-resolution UAV images, Rough segmentation and Fine extraction, seismic surface rupture

摘要：

精细刻画大地震形成的地表破裂的几何学特征对于深入理解发震断裂的运动学机制和变形规律具有重要意义。野外地质调查和影像目视解译等传统获取地震地表破裂的方法往往费时费力, 且难以获得其精细特征。文中依据面向对象、 颜色空间色彩分割等理论, 基于高分辨率无人机影像, 提出了一种面向对象的“粗分割-精提取”方法, 以实现半自动提取地震地表破裂的矢量面。对2021年青海玛多 M_S7.4 地震的实验结果表明, 该方法能够有效去除与地表破裂光谱相似的河道等噪声, 快速准确地提取地表破裂带的精细结构。文中建立的地震地表破裂半自动化提取方法可为大地震发生后快速提取地表破裂精细结构和分析地表变形特征提供一个可行方案。

关键词: 面向对象, 色彩分割, 高分辨率无人机影像, 粗分割-精提取, 地震地表破裂