Method and system for identifying seismic liquefaction

Abstract

The invention discloses a method and system for identifying seismic liquefaction, and the method comprises the steps: collecting a remote sensing image of a to-be-evaluated region, and extracting the liquefaction related parameters of each pixel point; according to the liquefaction related parameters of each pixel point, determining a classification type representing whether the pixel point belongs to sand liquefaction by using a seismic liquefaction identification model; and forming a complete sand liquefaction pit area according to the classification type of the pixel points. According to the method, high-precision evaluation of the seismic liquefaction area and rapid analysis of sand liquefaction space distribution are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster identification, and in particular to a method and system for identifying earthquake liquefaction. Background Art

[0002] Coseismic sand liquefaction refers to the phenomenon that when an earthquake occurs, the sand on the site changes from solid to liquid due to the increase of pore water pressure and the decrease of effective stress. Sand liquefaction pits refer to volcanic-like coverings formed on the surface after the sand liquefies. Rapid acquisition of liquefaction spatial distribution in sand liquefaction pits is helpful for earthquake intensity assessment and earthquake emergency response.

[0003] At present, the identification of sand liquefaction pits usually relies on field investigations, which are often very time-consuming and labor-intensive. More importantly, in areas such as plateaus and foreign countries, it is often difficult to reach them within a period of time after an earthquake, and it is impossible to obtain the complete spatial distribution of liquefaction pits. Therefore, the development of automatic extraction methods based on satellite or aerial images can help overcome the shortcomings of field investigations.

[0004] Existing automatic identification of sand liquefaction pits can be summarized into two categories. The first category is an unsupervised classification strategy that does not require sand liquefaction pit samples. This strategy detects sand liquefaction pits by comparing pre-earthquake and post-earthquake images. These methods automatically obtain sand liquefaction without sand liquefaction pit samples. However, the performance of these methods usually depends on the imaging quality of the image. The imaging quality of the image is usually affected by factors such as imaging time, weather, and resolution. In addition, these methods usually fail when pre-earthquake images close to the time of the earthquake do not exist. The second category is a supervised classification strategy based on post-earthquake images. The supervised classification strategy requires a certain number of sand liquefaction pit samples to be collected in advance. This strategy can be further divided into pixel-based methods and object-based methods. Pixel-based methods focus on the spectral characteristics of sand liquefaction pits in each pixel without considering the morphological characteristics of sand liquefaction pits. This type of method avoids the use of pre-earthquake images and usually has more accurate and robust recognition results than unsupervised methods.

[0005] Due to the morphological characteristics of sand liquefaction pits (e.g., size, shape), object-based methods usually require very high-resolution images, such as unmanned aerial vehicle (UAV) images. Compared with UAV images, satellite-based images have lower resolution, making it difficult to identify the morphology of sand liquefaction pits. However, satellite-based images are easier to obtain and cover a larger area. These images are most used in earthquake emergency situations. Therefore, it is necessary to develop sand liquefaction pit identification methods based on satellite images.

[0006] Currently, pixel-based methods often require users to collect land cover maps to help select non-liquefied / liquefied pit samples. However, high-resolution land cover maps are difficult to obtain, and currently land cover maps are not used to identify sand liquefaction pits. At the same time, current supervised classification-based methods require users to draw liquefaction polygons very carefully. The boundaries of liquefaction are usually very fuzzy and difficult to draw, and the wrong range may lead to a decrease in recognition accuracy.

[0007] Therefore, the prior art needs to provide a seismic sand liquefaction identification solution that improves the accuracy and applicability of sand liquefaction type identification. Summary of the invention

[0008] The purpose of the present invention is to provide a seismic sand liquefaction identification scheme that improves the accuracy and applicability of sand liquefaction type identification.

[0009] In order to solve the above technical problems, an embodiment of the present invention provides a method for identifying earthquake liquefaction, including: collecting remote sensing images of the area to be evaluated, and extracting liquefaction-related parameters of each pixel point; according to the liquefaction-related parameters of each pixel point, using an earthquake liquefaction identification model, determining whether the pixel point belongs to the classification type of sand liquefaction; according to the classification type of the pixel point, forming a complete sand liquefaction pit area.

[0010] Preferably, the earthquake liquefaction identification model includes a first type of identification model and a second type of identification model, wherein, in the step of determining whether the pixel point belongs to the classification type of sand liquefaction according to the liquefaction-related parameters of each pixel point and using the earthquake liquefaction identification model, it includes: configuring the corresponding identification model according to the resolution level of the remote sensing image, including: when the spatial resolution of the remote sensing image is at the sub-meter or meter level, selecting the first type of identification model to determine the sand liquefaction type of the pixel point; when the spatial resolution of the remote sensing image is at the centimeter or decimeter level, selecting the second type of identification model to determine the sand liquefaction type of the pixel point.

[0011] Preferably, the earthquake liquefaction identification model includes a first type of identification model and a second type of identification model, wherein, in the step of determining the sand liquefaction type of the pixel point according to the liquefaction-related parameters of each pixel point and using the earthquake liquefaction identification model, it includes: judging the correlation between the data in the liquefaction-related parameters, and determining the required identification model according to at least one correlation coefficient, including: when at least one correlation coefficient is less than a preset coefficient threshold and there are a large number of artificial buildings in the area to be evaluated, selecting the first type of identification model to determine the sand liquefaction type of the pixel point; when at least one correlation coefficient exceeds the preset coefficient threshold and there are a large number of natural landscapes in the area to be evaluated, selecting the second type of identification model to determine the sand liquefaction type of the pixel point.

[0012] Preferably, the remote sensing image is a post-earthquake image based on different image types, and the image type is one or more of satellite images, remote sensing images, drone images and terrain images.

[0013] Preferably, the liquefaction-related parameters include terrain features, spectral band parameters and spectral index parameters, the terrain features include flat areas, swamps, grasslands, rivers, lakes, floodplains, roads and exposed rocks, the spectral band parameters include blue, green, red and near infrared, and the spectral index parameters include one or more of the normalized difference water index, the normalized difference vegetation index, the improved soil vegetation adjustment index, the salinity index and the particle size index.

[0014] Preferably, the method also includes: constructing the first type of recognition model, which includes: collecting remote sensing images of all image types collected in the target area within different post-earthquake time periods, and extracting liquefaction-related parameters of each pixel point in the remote sensing image to form liquefied sample points and non-liquefied sample points with different terrain characteristics, spectral band parameters and spectral index parameter combination information; establishing a first basic model for predicting liquefaction type, and using the liquefied sample points and non-liquefied sample points to train and verify the first basic model to form the first type of recognition model, wherein the first basic model includes a plurality of parallel first decision trees and a voting decision module located at the back end of the plurality of first decision trees.

[0015] Preferably, the method also includes: constructing the second type of recognition model, which includes: collecting remote sensing images of all image types collected in the target area within different post-earthquake time periods, and extracting liquefaction-related parameters of each pixel point in the remote sensing image to form liquefied sample points and non-liquefied sample points with different terrain characteristics, spectral band parameters and spectral index parameter combination information; establishing a second basic model for predicting liquefaction type, and using the liquefied sample points and non-liquefied sample points to train and verify the second basic model to form the second type of recognition model, wherein the second basic model includes multiple cascaded second decision trees and an output module for weighting the prediction results of all second decision trees, and the decision tree of the latter level is obtained according to the negative gradient of the sample points calculated by the decision tree of the previous level.

[0016] Preferably, the step of forming a complete sand liquefaction pit area according to the sand liquefaction type of the pixel point includes: using window templates with gradually increasing scales to perform grid pixel processing on the initial image containing the pixel point classification type distribution result information in turn, to obtain a first image after multi-level grid pixel processing; identifying the sand liquefaction area in the first image and calculating the fractal dimension of each sand liquefaction area; by comparing the fractal dimension of at least one sand liquefaction area with a preset fractal dimension threshold, deleting the sand liquefaction area whose fractal dimension exceeds the preset fractal dimension threshold, wherein the preset fractal dimension threshold is determined according to the minimum spatial capacity characteristics that allow the image to contain artificial buildings and / or roads; and forming a complete sand liquefaction pit area based on the remaining sand liquefaction area.

[0017] On the other hand, an embodiment of the present invention provides a system for identifying earthquake liquefaction, including: a feature extraction module, which is configured to collect remote sensing images of the area to be evaluated and extract liquefaction-related parameters of each pixel point; a sand liquefaction type identification module, which is configured to determine whether the characterizing pixel point belongs to the classification type of sand liquefaction based on the liquefaction-related parameters of each pixel point and using an earthquake liquefaction identification model; a liquefaction pit area formation module, which is configured to form a complete sand liquefaction pit area based on the classification type of the pixel point.

[0018] In addition, an embodiment of the present invention further provides a computer-readable storage medium storing computer-executable instructions, which can implement the steps of the method described above when executed by a processor.

[0019] Compared with the prior art, one or more embodiments of the above scheme may have the following advantages or beneficial effects:

[0020] The present invention proposes a method and system for identifying earthquake liquefaction. The method and system are based on a sand liquefaction identification method framework of binary classification. By adaptively selecting two types of typical representative machine learning automatic identification models (random forest and gradient boosting decision tree methods), high-precision evaluation of earthquake liquefaction pits in the area to be evaluated and rapid analysis of the spatial distribution of sand liquefaction are achieved. Various types of image information such as optical satellite images and limited samples can be effectively used to quickly and reliably identify sand liquefaction pits after an earthquake. It is widely applicable to various application scenarios and forms a general prediction method framework process.

[0021] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0023] Figure 1 A schematic diagram of the steps of a method for identifying earthquake liquefaction according to an embodiment of the present application.

[0024] Figure 2 A schematic diagram of a specific process of a method for identifying earthquake liquefaction according to an embodiment of the present application.

[0025] Figure 3 This is an example diagram of liquefaction pit area extraction results for identifying earthquake liquefaction in an embodiment of the present application.

[0026] Figure 4 A block diagram of a system module for identifying earthquake liquefaction according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will describe the implementation methods of the present invention in detail with reference to the accompanying drawings and embodiments, so that the implementation process of how the present invention applies technical means to solve technical problems and achieve technical effects can be fully understood and implemented accordingly. It should be noted that as long as there is no conflict, the various embodiments of the present invention and the various features in the embodiments can be combined with each other, and the technical solutions formed are all within the protection scope of the present invention.

[0028] In addition, the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions. Also, although a logical sequence is shown in the flowchart, in some cases, the steps shown or described can be performed in a sequence different from that here.

[0029] The terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments. Unless the context clearly indicates otherwise, the singular forms "one", "one" and "item" used herein are also intended to include plural numbers. It should also be understood that the terms "include" and / or "comprise" used herein specify the existence of stated features, integers, steps, operations, units and / or components, without excluding the existence or addition of one or more other features, integers, steps, operations, units, components and / or combinations thereof.

[0030] In order to solve the problems in the above-mentioned background technology, the present invention and the embodiment of the present application provide a method and system for identifying earthquake liquefaction. The method and system are a binary supervised classification framework, which does not require the help of a land cover map. Users can focus on selecting typical pixel samples of non-liquefied and liquefied pits as training samples. The constructed recognition model can quickly identify liquefaction based on satellite, telemetry, and drone images. In addition, the present invention also implements a binary classification framework by adaptively selecting a random forest (RF) model or a gradient boosted decision tree (GBDT) model suitable for the current prediction scenario.

[0031] Embodiment 1

[0032] Figure 1 A schematic diagram of the steps of a method for identifying earthquake liquefaction according to an embodiment of the present application. Figure 2 This is a schematic diagram of a specific process of the method for identifying earthquake liquefaction according to an embodiment of the present application. Figure 1 and Figure 2 , the specific steps of the method for identifying earthquake liquefaction (also called "earthquake liquefaction identification method") described in an embodiment of the present invention are described.

[0033] Step S110 , collecting remote sensing images of the area to be evaluated, and extracting liquefaction-related parameters of each pixel in the collected image.

[0034] In an embodiment of the present invention, the area to be evaluated represents an area whose earthquake liquefaction characteristics need to be understood. The collected remote sensing images are based on post-earthquake images of different image types. Moreover, the image type is one or more of satellite images, telemetry images, drone images, and terrain images.

[0035] Therefore, the remote sensing images collected by the embodiment of the present invention can be simultaneously applicable to satellite images, remote sensing images, drone images and terrain images.

[0036] Then, after acquiring the remote sensing image to be processed, step S110 will extract the liquefaction related parameters of each pixel in the current image.

[0037] In an embodiment of the present invention, liquefaction related parameters include: terrain features, spectral band parameters and spectral index parameters. Terrain features include but are not limited to: flat areas, swamps, grasslands, rivers, lakes, floodplains, roads and exposed rocks. Spectral band parameters include but are not limited to: blue, green, red and near infrared. Spectral index parameters include but are not limited to: one or more of normalized difference water index, normalized difference vegetation index, improved soil vegetation adjustment index, salinity index and particle size index.

[0038] Specifically, corresponding liquefaction related parameters are configured for each pixel point in the terrain image, that is, corresponding terrain features, spectral band parameters and spectral index parameters are generated for each pixel point, thereby entering step S120.

[0039] Step S120 , according to the liquefaction related parameters of each pixel point in the current terrain image obtained in step S110 , using the earthquake liquefaction identification model, determine the sand liquefaction type of each pixel point.

[0040] In the embodiment of the present invention, the earthquake liquefaction identification model includes: a first type of identification model and a second type of identification model. Before the implementation of step S120, the earthquake liquefaction identification method according to the embodiment of the present invention further includes constructing the first type of identification model and the second type of identification model respectively.

[0041] The process of building the first type of recognition model includes:

[0042] Step S1, collect remote sensing images of all image types collected in the target area in different post-earthquake time periods, and extract liquefaction-related parameters of each pixel point in the remote sensing image to form liquefied sample points and non-liquefied sample points with different terrain characteristics, spectral band parameters and spectral index parameter combination information.

[0043] In addition, in an embodiment of the present invention, the multiple remote sensing images collected in step S1 need to include non-liquefied areas and liquefied pit areas under various terrain features, and the collected terrain images also need to include all spectral band parameters, so that each pixel point extracted in step S1 can form various types of sample points including different terrain features, spectral band parameters and spectral index parameter combination information, so as to serve as model training data.

[0044] After completing the sample point extraction task, in step S2, a first basic model for predicting the liquefaction type is established, and all liquefied sample points and all non-liquefied sample points formed in step S1 are used to train and verify the first basic model to form a first type of recognition model.

[0045] In an embodiment of the present invention, the first basic model includes a plurality of first decision trees connected in parallel and a voting decision module located at the back end of the plurality of first decision trees. The input of each first decision tree is a liquefaction-related parameter, and the output is a voting result of the liquefaction type predicted by the decision tree. The input of the voting decision module is the voting results of the liquefaction type predicted by all the first decision trees, and the output is the final voting result of the liquefaction type predicted by the first basic model.

[0046] In step S2 described in the embodiment of the present invention, a first basic model based on a random forest machine learning framework is established, and then all liquefied sample points and all non-liquefied sample points are divided into a first data training set and a first data verification set according to a preset ratio. Then, the liquefaction-related parameters of each sample point in the first data training set and the corresponding sand liquefaction type are used as training input data and training output data, respectively, and a random forest learning algorithm is used to train the first basic model, and the first data verification set is used to verify the trained model. After the verification is qualified, a first type of recognition model is formed.

[0047] As a typical averaging method, random forest (RF) uses multiple decision trees to independently classify and judge the target, and then uses voting to predict the classification result of the target according to the number of votes. For the binary classification of the sand liquefaction type of the pixel point to be predicted by the present invention, each decision tree in RF will vote separately to determine whether a pixel is liquefied, and then combine these voting results to determine whether the pixel belongs to sand liquefaction according to the number of votes.

[0048] The process of building the second type of recognition model includes:

[0049] Step T1, collects remote sensing images of all image types collected in the target area in different post-earthquake time periods, and extracts liquefaction-related parameters of each pixel point in the remote sensing image to form liquefied sample points and non-liquefied sample points with different terrain characteristics, spectral band parameters and spectral index parameter combination information. Since step T1 of the present invention is similar to step S1, the implementation method of step T1 is not described in detail here.

[0050] After completing the sample point extraction task, in step T2, a second basic model for predicting the liquefaction type is established, and all liquefied sample points and all non-liquefied sample points formed in step T1 are used to train and verify the second basic model to form a second type of recognition model.

[0051] In an embodiment of the present invention, the second basic model includes a plurality of cascaded first decision trees and an output module for weighting the prediction results of all second decision trees. The decision tree of the next level is obtained according to the negative gradient of the sample points calculated by the decision tree of the previous level. The output of the output module is the final liquefaction type voting result predicted by the second basic model.

[0052] In step T2 described in the embodiment of the present invention, a second basic model based on a gradient boosting decision tree machine learning framework is established, and then all liquefied sample points and all non-liquefied sample points are divided into a second data training set and a second data verification set according to a preset ratio. Then, the liquefaction-related parameters of each sample point in the second data training set and the corresponding sand liquefaction type are used as training input data and training output data, respectively, and a gradient boosting decision tree learning algorithm is used to train the second basic model, and the second data verification set is used to verify the trained model. After the verification is qualified, a second type of recognition model is formed.

[0053] The basic idea of ​​the gradient boosted decision tree algorithm (GBDT) is to sort the decision trees in order. During the training process, each decision tree will give higher weights to samples that are misclassified by the previous decision tree. Finally, the prediction results of each decision tree are weighted to obtain the final prediction result.

[0054] RF and GBDT are powerful in processing high-dimensional data and selecting covariates, but the main differences between the two can be summarized into two aspects. The first is the sampling strategy: each tree in RF is constructed from a randomly selected sample subset with replacement from the complete sample set; the second is the classification result: RF determines the prediction result of the sand liquefaction type for each pixel by the majority vote of all decision trees in RF, while GBDT uses a weighted average method to integrate the prediction results of each decision tree to obtain the final prediction result of the sand liquefaction type for each pixel.

[0055] In this way, after completing the construction of the above-mentioned earthquake liquefaction identification model, the two types of identification models can be applied in step S120 to predict the sand liquefaction type (i.e., whether it is a sand liquefaction point) of each pixel point in the terrain image collected in step S110.

[0056] In order to improve the applicable scenarios of the earthquake liquefaction identification model of the present invention and at the same time ensure the reliability, accuracy and stability of the prediction results, in an embodiment of the present invention, the category of the identification model applicable to the current pixel to be evaluated can be adaptively selected according to the image type or the liquefaction-related parameters of the pixel.

[0057] In the first embodiment, the corresponding recognition model can be configured according to the resolution level of the acquired remote sensing image.

[0058] Specifically, in the first embodiment, when the spatial resolution of the remote sensing image is at the sub-meter or meter level, the first type of recognition model is selected to determine the sand liquefaction type of the pixel point. When the spatial resolution of the remote sensing image is a satellite image or other type of image at the sub-meter or meter level, the first type of recognition model is selected to determine the sand liquefaction classification type of all pixels in the remote sensing image.

[0059] In the second embodiment, when the spatial resolution of the remote sensing image is at the centimeter or decimeter level, the second type of recognition model is selected to determine the sand liquefaction type of the pixel point. When the spatial resolution of the remote sensing image is a satellite image or drone data or other type of image with high resolution at the centimeter or decimeter level, the second type of recognition model is selected to determine the sand liquefaction classification type of all pixels in the remote sensing image.

[0060] In the second embodiment, the correlation between the data in the pixel liquefaction related parameters in the current remote sensing image can also be determined, so as to determine the required recognition model according to at least one correlation coefficient.

[0061] Specifically, in the first embodiment, when at least one correlation coefficient is less than the preset coefficient threshold and there are a large number of artificial buildings in the area to be evaluated, the first type of recognition model is selected to determine the sand liquefaction type of the pixel point. When there are many liquefaction-related parameters with weak correlation parameters and there are a large number of artificial buildings in the study area, the first type of recognition model is selected to determine the sand liquefaction classification type of all pixels in the remote sensing image.

[0062] In the second embodiment, when at least one correlation coefficient exceeds the preset coefficient threshold and there are a large number of natural landscapes in the area to be evaluated, the second type of recognition model is selected to determine the sand liquefaction type of the pixel point. When there are many liquefaction-related parameters with strong correlation parameters and the study area is mainly natural landscapes, the second type of recognition model is selected to determine the sand liquefaction classification type of all pixels in the remote sensing image.

[0063] After determining the sand liquefaction type of each pixel point, the original remote sensing image is converted into a binary image operation result indicating whether each pixel point belongs to a sand liquefaction pit, thereby entering step S130.

[0064] Step S130, forming a complete sand liquefaction pit area in the terrain image of the area to be evaluated according to the sand liquefaction type of each pixel point in the terrain image predicted in step S120, thereby completing the final extraction of the liquefaction pit area.

[0065] In step S130, based on the sand liquefaction pit prediction results of all pixel points in the binary image obtained in step S120, an initial liquefaction pit area is first formed, and then morphological transformation is used to fill in the defective positions in the current initial liquefaction pit area (fill the gaps in the sand liquefaction pit area) and perform point deletion on discrete positions, thereby forming a morphologically adjusted sand liquefaction pit area.

[0066] In one embodiment, step S130 may utilize the following steps to complete the conversion from the binary image to the complete sand liquefaction pit area image.

[0067] Specifically, using window templates with gradually increasing scales, grid pixel processing is performed on the initial image containing the pixel classification type distribution result information in sequence (a single grid pixel processing includes sequentially performing morphological closing operations and resampling operations) to obtain a first image after multi-level grid pixel processing; the sand liquefaction area in the current first image is identified and the fractal dimension of each sand liquefaction area is calculated; by comparing the fractal dimension of at least one sand liquefaction area with a preset fractal dimension threshold, the sand liquefaction area whose fractal dimension exceeds the preset fractal dimension threshold is deleted; based on the remaining sand liquefaction areas, a complete sand liquefaction pit area is formed.

[0068] In the embodiment of the present invention, the preset fractal dimension threshold is determined according to the minimum spatial capacity feature that allows the image to contain artificial buildings and / or roads.

[0069] More specifically: First, a multi-scale morphological method based on a window template is used to merge and delete the binary image of the sand liquefaction classification at the pixel level. First, a window template based on the initial scale level of 3*3 is designed, and the binary image is processed with raster pixels based on the morphological closing operation; then, based on the image obtained by the raster processing of the window template at the initial scale level, the 3*3 operator is used to resample the result of the previous step, and the majority is used as the resampling result. Then, a window template of a 5*5 scale level slightly larger than the initial scale level is used, and the morphological closing operation and resampling operation are further carried out in sequence to process the raster pixels. By analogy, the size of the highest scale template is set according to the scale of sand liquefaction in the study area. Usually, the highest scale level can be set to 2-5 meters, and the specific needs need to be adjusted according to the quality of the classification results. Thus, the first image after the multi-level raster pixel processing is obtained.

[0070] Next, all the sand liquefaction areas in the current first image are identified, and the fractal dimension of each sand liquefaction pit polygon is calculated. After that, the sand liquefaction areas whose fractal dimension exceeds the preset fractal dimension threshold are deleted, so as to obtain the final required sand liquefaction pit spatial distribution area. In the embodiment of the present invention, the fractal dimension threshold is set according to the regional artificial buildings and / or road characteristics, so as to achieve the effect of deleting the artificial regular buildings, roads and other areas in the image according to the fractal dimension threshold.

[0071] Embodiment 2

[0072] The following takes a certain experimental area as an example to illustrate the specific implementation process of the earthquake liquefaction identification method described in the embodiment of the present invention.

[0073] The earthquake liquefaction identification method described in an embodiment of the present invention was tested based on the co-seismic sand liquefaction caused by an earthquake in a certain place on May 22, 2021.

[0074] The experimental area is about 45 square kilometers, located 30 kilometers west of the epicenter, between 97.9° and 98.02° east longitude and 34.64° and 34.74° north latitude. The elevation range of the experimental area is 4190m to 4512m, with a gentle slope and an average slope of less than 10°.

[0075] Gaofen-7 (GF-7) images were collected to identify coseismic sand liquefaction, and two other images were used for verification. First, the Gaofen-1D image (GF-1D) taken on April 29, 2021 was used as the pre-earthquake image; at the same time, the collected GF-1D image covered the entire experimental area. In addition, unmanned aerial vehicle (UAV) images taken on July 25, 2021 were also collected with a resolution of 0.1m.

[0076] Based on the processed GF-7 data, three types of related parameters for identifying sand liquefaction were produced, thus forming corresponding liquefaction related parameters for each pixel:

[0077] Terrain related parameters: elevation and slope. Since liquefaction usually occurs in flat and swampy areas, terrain parameters are mainly used to characterize terrain features.

[0078] Spectral band related parameters: blue (B), green (G), red (R), and near infrared (NIR) of GF-7.

[0079] Spectral index related parameters: Normalized Difference Water Index (NDWI), Normalized Difference Vegetation Index (NDVI), Modified Soil Vegetation Adjustment Index (MSAVI2), Salinity Index (Salinity), and Grain Size Index (GSI).

[0080] From GF-7 data, GF-1D data and drone images, a total of about 300 sand liquefaction pit sample points and 600 non-liquefaction pit sample points were collected. All liquefaction point data were located in obvious liquefaction pits and were not close to the edge as much as possible. Non-liquefaction sample points were distributed in different landscapes such as grasslands, rivers / lakes, floodplains, roads, and exposed rocks.

[0081] Figure 3 The liquefaction recognition results of the two recognition models, RF and GBDT, are shown. (a), (b) and (c) show the sand liquefaction pit areas predicted by RF using the pre-earthquake GF-1D image, the post-liquefaction GF-7 image and the post-liquefaction UAV image, respectively; (d), (e) and (f) show the sand liquefaction pit areas predicted by GBDT using the pre-earthquake GF-1D image, the post-liquefaction GF-7 image and the post-liquefaction UAV image, respectively.

[0082] Three types of images (i.e., GF-1D, GF-7, and UAV) are available in the experimental area. Compared with the pre-earthquake image (GF-1D), the surface changes in the post-earthquake image (GF-7) are obvious in the sand liquefaction area, and RF and GBDT can identify these places as sand liquefaction. In addition, the identified sand liquefaction area can be verified in the field UAV image. The earthquake liquefaction identification method proposed in the present invention can obtain sand liquefaction pits with higher accuracy. For example, in the Kappa coefficient, the recognition accuracy of RF is between 0.9-0.98, and the recognition accuracy of GBDT is between 0.87-0.97.

[0083] Embodiment 3

[0084] Based on the earthquake liquefaction identification method of the above embodiment 1 and embodiment 2, the embodiment of the present invention further provides a system for identifying earthquake liquefaction (also called "earthquake liquefaction identification system"). The earthquake liquefaction identification system is used to implement the above earthquake liquefaction identification method.

[0085] Figure 4 1 is a block diagram of a system module for identifying earthquake liquefaction according to an embodiment of the present application. Figure 4 As shown, the earthquake liquefaction identification system according to the embodiment of the present invention includes: a feature extraction module 41 , a sand liquefaction type identification module 42 and a liquefaction pit area formation module 43 .

[0086] Specifically, the feature extraction module 41 is implemented according to the method described in step S110 above, and is configured to collect remote sensing images of the area to be evaluated and extract liquefaction-related parameters of each pixel point; the sand liquefaction type identification module 42 is implemented according to the method described in step S120 above, and is configured to determine whether the pixel point belongs to the classification type of sand liquefaction based on the liquefaction-related parameters of each pixel point and the seismic liquefaction identification model; the liquefaction pit area formation module 43 is implemented according to the method described in step S130 above, and is configured to form a complete sand liquefaction pit area based on the classification type of the pixel point.

[0087] Embodiment 4

[0088] Based on the earthquake liquefaction identification method of the above-mentioned embodiment 1 and embodiment 2, the embodiment of the present invention provides a computer-readable storage medium. The storage medium stores a computer program, and the computer program is executed to run a method for identifying earthquake liquefaction. The computer program can run computer instructions, and the computer instructions include computer program code, which can be in source code form, object code form, executable file or some intermediate form.

[0089] Computer-readable storage media may include: any entity or device that can carry computer program code, recording media, USB flash drives, mobile hard disks, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0090] It should be noted that the content contained in computer-readable storage media can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, the content contained in computer-readable storage media does not include electrical carrier signals and telecommunication signals.

[0091] The present invention discloses a method and system for identifying earthquake liquefaction. The method and system are based on a sand liquefaction identification method framework of binary classification, and realize high-precision evaluation of earthquake liquefaction pits in the area to be evaluated and rapid analysis of the spatial distribution of sand liquefaction by adaptively selecting two types of typical representative machine learning automatic identification models (random forest and gradient boosting decision tree methods). It can effectively use various types of image information such as optical satellite images and limited samples to quickly and reliably identify sand liquefaction pits after an earthquake, and is widely applicable to various application scenarios and forms a general prediction method framework process.

[0092] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any person familiar with the technology within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

[0093] In the description of the present invention, unless otherwise specified, "plurality" means two or more than two; the orientations or positional relationships indicated by the terms "upper", "lower", "left", "right", "inner", "outer", "front end", "rear end", "head", "tail", etc. are based on the orientations or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0094] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "connected" and "connection" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0095] It should be understood that the embodiments disclosed in the present invention are not limited to the specific structures, processing steps or materials disclosed herein, but should be extended to equivalent substitutions of these features understood by ordinary technicians in the relevant field. It should also be understood that the terms used herein are only used for the purpose of describing specific embodiments and are not meant to be limiting.

[0096] The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0097] Although the embodiments disclosed in the present invention are as above, the contents described are only embodiments adopted for facilitating the understanding of the present invention and are not intended to limit the present invention. Any technician in the technical field to which the present invention belongs can make any modifications and changes in the form and details of the implementation without departing from the spirit and scope disclosed in the present invention, but the patent protection scope of the present invention shall still be subject to the scope defined in the attached claims.

Claims

1. A method for identifying earthquake liquefaction, characterized in that: include: Collect remote sensing images of the area to be evaluated and extract liquefaction-related parameters of each pixel; According to the liquefaction related parameters of each pixel point, using the earthquake liquefaction identification model, determining whether the pixel point belongs to the classification type of sand liquefaction; According to the classification type of pixel points, a complete sand liquefaction pit area is formed.

2. The method according to claim 1, characterized in that The earthquake liquefaction identification model includes a first type of identification model and a second type of identification model, wherein, according to the liquefaction related parameters of each pixel point, using the earthquake liquefaction identification model, the step of determining whether the pixel point belongs to the classification type of sand liquefaction includes: According to the resolution level of the remote sensing image, a corresponding recognition model is configured, including: When the spatial resolution of the remote sensing image is at the sub-meter or meter level, a first type of recognition model is selected to determine the sand liquefaction type of the pixel point; When the spatial resolution of the remote sensing image is at the centimeter or decimeter level, the second type of recognition model is selected to determine the sand liquefaction type of the pixel point.

3. The method according to claim 1, characterized in that The earthquake liquefaction identification model includes a first type of identification model and a second type of identification model, wherein the step of determining the sand liquefaction type of the pixel point by using the earthquake liquefaction identification model according to the liquefaction related parameters of each pixel point includes: Determine the correlation between the data in the liquefaction related parameters, and determine the required identification model according to at least one correlation coefficient, including: When at least one correlation coefficient is less than a preset coefficient threshold and there are a large number of artificial buildings in the area to be evaluated, the first type of recognition model is selected to determine the sand liquefaction type of the pixel point; When at least one correlation coefficient exceeds a preset coefficient threshold and there are a large number of natural landscapes in the area to be evaluated, the second type of recognition model is selected to determine the sand liquefaction type of the pixel point.

4. The method according to claim 2 or 3, characterized in that: The remote sensing image is a post-earthquake image based on different image types, and the image type is one or more of satellite images, remote sensing images, drone images and terrain images.

5. The method according to claim 4, characterized in that The liquefaction-related parameters include terrain features, spectral band parameters and spectral index parameters. The terrain features include flat areas, swamps, grasslands, rivers, lakes, floodplains, roads and exposed rocks. The spectral band parameters include blue, green, red and near infrared. The spectral index parameters include one or more of the normalized difference water index, the normalized difference vegetation index, the improved soil vegetation adjustment index, the salinity index and the particle size index.

6. The method according to claim 5, characterized in that The method further comprises: constructing the first type of recognition model, which comprises: Collect remote sensing images of all image types collected in the target area in different post-earthquake time periods, and extract liquefaction-related parameters of each pixel point in the remote sensing images to form liquefied sample points and non-liquefied sample points with different terrain characteristics, spectral band parameters and spectral index parameter combination information; A first basic model for predicting liquefaction type is established, and the liquefied sample points and non-liquefied sample points are used to train and verify the first basic model to form the first type recognition model, wherein: The first basic model includes a plurality of parallel first decision trees and a voting decision module located at the rear ends of the plurality of first decision trees.

7. The method according to claim 5 or 6, characterized in that: The method further comprises: constructing the second type of recognition model, which comprises: Collect remote sensing images of all image types collected in the target area in different post-earthquake time periods, and extract liquefaction-related parameters of each pixel point in the remote sensing images to form liquefied sample points and non-liquefied sample points with different terrain characteristics, spectral band parameters and spectral index parameter combination information; A second basic model for predicting liquefaction type is established, and the liquefied sample points and non-liquefied sample points are used to train and verify the second basic model to form the second type recognition model, wherein: The second basic model includes a plurality of cascaded second decision trees and an output module for weighting the prediction results of all the second decision trees, and the decision tree of the next stage is obtained according to the negative gradient of the sample points calculated by the decision tree of the previous stage.

8. The method according to any one of claims 1 to 7, characterized in that: The step of forming a complete sand liquefaction pit area according to the sand liquefaction type of the pixel point includes: Using window templates with gradually increasing scales, grid pixel processing is performed on the initial image containing pixel point classification type distribution result information in turn to obtain a first image after multi-level grid pixel processing; identifying sand liquefaction areas in the first image and calculating the fractal dimension of each sand liquefaction area; By comparing the fractal dimension of at least one sand liquefaction region with a preset fractal dimension threshold, the sand liquefaction region whose fractal dimension exceeds the preset fractal dimension threshold is deleted, wherein the preset fractal dimension threshold is determined according to the minimum spatial capacity feature that allows the image to contain artificial buildings and / or roads; Based on the remaining sand liquefaction area, a complete sand liquefaction pit area is formed.

9. A system for identifying earthquake liquefaction, characterized in that: include: A feature extraction module configured to collect remote sensing images of the area to be evaluated and extract liquefaction-related parameters of each pixel point; A sand liquefaction type identification module is configured to determine whether the pixel point belongs to the classification type of sand liquefaction according to the liquefaction related parameters of each pixel point and using the earthquake liquefaction identification model; The liquefaction pit area forming module is configured to form a complete sand liquefaction pit area according to the classification type of the pixel points.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the method according to any one of claims 1 to 8.

Images