A landslide identification model training method, landslide identification method and device

Abstract

The present application relates to the technical field of disaster prevention and mitigation, and discloses a landslide identification model training method, a landslide identification method and device. The present application determines each region in a preset range as an overlapping region and a non-overlapping region through the spatial superposition of first landslide information and second landslide information. The overlapping region is a region where the remote sensing image is accurately judged, and the non-overlapping region is a region where the remote sensing image is incorrectly judged. The target remote sensing image feature is determined through the remote sensing image feature values corresponding to the remote sensing image features in the overlapping region and the non-overlapping region. The pre-constructed landslide identification model is trained through the target remote sensing image feature and the second landslide information of the overlapping region, so that the model learns the correlation between the remote sensing image feature and the landslide information, so that the trained model has the ability to determine the landslide information according to the remote sensing image feature. The present application uses the trained landslide identification model to make the landslide identification process more efficient and accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of disaster prevention and mitigation, and particularly relates to a landslide identification model training method, a landslide identification method and device. BACKGROUND

[0002] Landslide identification is very important for geological disaster risk assessment and risk assessment analysis, and the purpose is to obtain the spatial distribution information of landslide regions in a specific area, which usually includes detailed attribute information such as spatial distribution position, landslide area, and landslide boundary of the landslide.

[0003] At present, landslide identification methods mainly include field investigation methods, remote sensing-based visual interpretation methods, and remote sensing-based automatic identification methods. However, the field investigation method is high in cost and difficult to complete the landslide identification task in a large area; the remote sensing visual interpretation method is time-consuming and depends on expert experience and field verification; and the remote sensing-based automatic identification method is high in efficiency, but its accuracy is greatly affected by the quality of sample data.

[0004] In summary, the existing landslide identification methods usually have different advantages and disadvantages. Therefore, a landslide identification method with low cost, high efficiency and high accuracy is an urgent problem to be solved. SUMMARY

[0005] Therefore, the present application provides a landslide identification model training method, a landslide identification method and device to solve the problems of low efficiency and low accuracy of landslide identification.

[0006] In a first aspect, the present application provides a landslide identification model training method, which comprises:

[0007] obtaining remote sensing images in a preset range and first landslide information corresponding to each region in the preset range, the first landslide information being used to represent whether the corresponding region has a landslide; interpreting the remote sensing images to obtain second landslide information corresponding to each region, the second landslide information being used to represent whether the corresponding region has a landslide; performing spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions; determining target remote sensing image features based on feature values of the remote sensing images corresponding to the overlapping regions and the non-overlapping regions; and training a pre-constructed landslide identification model based on the feature values of the target remote sensing images corresponding to each region and the second landslide information corresponding to the overlapping regions to obtain a landslide identification model.

[0008] The present application determines each region in a preset range as an overlapping region and a non-overlapping region by spatial superposition of the first landslide information and the second landslide information, the overlapping region being a region where the remote sensing image is accurately judged, and the non-overlapping region being a region where the remote sensing image is inaccurately judged. The target remote sensing image feature is determined through the remote sensing image feature values corresponding to the remote sensing image features in the overlapping region and the non-overlapping region, and the pre-constructed landslide recognition model is trained through the target remote sensing image feature and the second landslide information of the overlapping region, so that the model learns the association between the remote sensing image feature and the landslide information, so that the trained model has the ability to determine the landslide information according to the remote sensing image feature. The trained landslide recognition model can solve the problems of low efficiency, small range, high cost, and low accuracy in the existing landslide recognition method, and make the landslide recognition process more efficient and accurate.

[0009] In an optional embodiment, the first landslide information and the second landslide information corresponding to each region are analyzed by spatial superposition to obtain overlapping regions and non-overlapping regions, including:

[0010] It is judged whether the first landslide information and the second landslide information corresponding to each region are consistent; when the first landslide information and the second landslide information are consistent, the corresponding region is determined as an overlapping region; when the first landslide information and the second landslide information are inconsistent, the corresponding region is determined as a non-overlapping region.

[0011] In this embodiment, the consistency of the first landslide information and the second landslide information is used to judge whether the determination of the landslide condition of the same region by the two different sources of data is consistent, thereby ensuring the accuracy of the label in the subsequent training process, and indirectly making the output result of the model more accurate.

[0012] In an optional embodiment, the target remote sensing image feature is determined based on the remote sensing image feature values corresponding to the overlapping region and the non-overlapping region, including:

[0013] The first remote sensing image feature value corresponding to the overlapping region and the second remote sensing image feature value corresponding to the non-overlapping region are obtained; the separation degree of the corresponding remote sensing image feature is determined based on the first remote sensing image feature value and the second remote sensing image feature value; the remote sensing image feature with a separation degree greater than or equal to a preset separation degree threshold is determined as the target remote sensing image feature.

[0014] The embodiment determines the separation degrees of the remote sensing image features based on the first remote sensing image feature values and the second remote sensing image feature values. The greater the separation degree of the remote sensing image feature is, the easier it is to distinguish the overlapping area and the non-overlapping area based on the remote sensing image feature. Therefore, the embodiment determines the features with the separation degree greater than or equal to the preset separation degree threshold as the target remote sensing image features. The purpose is to facilitate the training of the model based on the target remote sensing image features and the second landslide information of the overlapping area, so that the model can better learn the relevance between the target remote sensing image features and the second landslide information, and the result predicted by the final model is more accurate.

[0015] In an optional implementation, the first landslide information corresponding to each area in the preset range is obtained, including:

[0016] The target influence factor and the third landslide information corresponding to each area in the preset range are obtained, the third landslide information being information actually collected about whether the corresponding area has a landslide; the pre-constructed landslide prediction model is trained based on the target influence factor and the third landslide information corresponding to each area, to obtain a trained landslide prediction model; and the landslide distribution of the preset range is predicted by using the landslide prediction model, to obtain the first landslide information corresponding to each area.

[0017] In an optional implementation, the target influence factor is obtained, including:

[0018] The geographic information and the environmental information of each area in the preset range are obtained; all the geographic information, the environmental information, and the third landslide information are statistically analyzed to determine a plurality of initial influence factors causing the landslide; and the target influence factor is selected from the plurality of initial influence factors according to a preset rule.

[0019] In an optional implementation, the target influence factor is selected from the plurality of initial influence factors according to a preset rule, including:

[0020] The correlation between each initial influence factor in each area and the third landslide information is analyzed to obtain a first correlation coefficient between each initial influence factor in each area and the landslide; a random forest method is used to determine a second correlation coefficient between each initial influence factor and the landslide under the comprehensive action of all the initial influence factors; an initial weight corresponding to each initial influence factor in each area is determined based on the first correlation coefficient and the second correlation coefficient; and the target influence factor is selected from the plurality of initial influence factors based on the initial weight, a preset weight threshold, and the third landslide information.

[0021] In an optional implementation, the target influence factor is selected from the plurality of initial influence factors based on the initial weight, the preset weight threshold, and the third landslide information, including:

[0022] The initial weights corresponding to the plurality of initial influence factors in each region are compared with the preset weight threshold respectively, and the region with the initial weight greater than or equal to the preset weight threshold is determined as the first target region; the region determined to have a landslide according to the second landslide information is determined as the second target region; the first target region and the second target region are subjected to spatial superposition analysis to determine an overlap rate; when the overlap rate is greater than or equal to a preset overlap threshold, the second correlation coefficient corresponding to the initial influence factor is adjusted; when the adjusted second correlation coefficient is greater than or equal to a preset correlation coefficient threshold, the corresponding initial influence factor is determined as the target influence factor.

[0023] In a second aspect, the present application provides a landslide identification method, comprising:

[0024] A remote sensing observation image of a region to be identified is obtained; the remote sensing observation image is input into a landslide identification model trained by the training method of the landslide identification model provided in the first aspect or any of the embodiments thereof, to obtain an identification result of whether the region to be identified has a landslide.

[0025] In a third aspect, the present application provides a training device of a landslide identification model, comprising:

[0026] The acquisition module is configured to acquire remote sensing images in a preset range and first landslide information corresponding to each region in the preset range, the first landslide information being used to represent whether the corresponding region has a landslide; the interpretation module is configured to interpret the remote sensing images to obtain second landslide information corresponding to each region, the second landslide information being used to represent whether the corresponding region has a landslide; the analysis module is configured to perform spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions; the determination module is configured to determine target remote sensing image features based on feature values of the remote sensing images corresponding to the overlapping regions and the non-overlapping regions; and the training module is configured to train a pre-constructed landslide identification model based on the feature values of the target remote sensing images corresponding to each region and the second landslide information corresponding to the overlapping regions, to obtain the landslide identification model.

[0027] In an optional embodiment, the analysis module comprises:

[0028] The judgment submodule is configured to judge whether the first landslide information and the second landslide information corresponding to each region are consistent; the first determination submodule is configured to determine the corresponding region as an overlapping region when the first landslide information and the second landslide information are consistent; and the second determination submodule is configured to determine the corresponding region as a non-overlapping region when the first landslide information and the second landslide information are inconsistent.

[0029] In an optional embodiment, the determination module comprises:

[0030] The first obtaining sub-module is configured to obtain a first remote sensing image feature value corresponding to the overlapping area and a second remote sensing image feature value corresponding to the non-overlapping area.

[0031] In an optional implementation, the obtaining module comprises:

[0032] The second obtaining sub-module is configured to obtain a target influence factor and third landslide information corresponding to each area in the preset range, the third landslide information being information about whether the corresponding area actually collects landslide information; the training sub-module is configured to train the pre-constructed landslide prediction model based on the target influence factor and the third landslide information corresponding to each area, to obtain the trained landslide prediction model; and the prediction sub-module is configured to use the landslide prediction model to predict the landslide distribution of the preset range, to obtain the first landslide information corresponding to each area.

[0033] In a fourth aspect, the present application provides a landslide identification device, comprising:

[0034] The obtaining module is configured to obtain a remote sensing observation image of a to-be-identified area; the training module is configured to input the remote sensing observation image into the landslide identification model trained by the training method of the landslide identification model provided in the first aspect or any of the corresponding embodiments, to obtain an identification result about whether the to-be-identified area has landslide.

[0035] In a fifth aspect, the present application provides a computer device, comprising a memory and a processor, which are communicatively connected with each other, the memory stores computer instructions, and the processor executes the computer instructions to perform the training method of the landslide identification model of the first aspect or any of the corresponding embodiments, or to perform the landslide identification method of the second aspect or any of the corresponding embodiments.

[0036] In a sixth aspect, the present application provides a computer readable storage medium, which stores computer instructions, and the computer instructions are used to make a computer execute the training method of the landslide identification model of the first aspect or any of the corresponding embodiments, or to execute the landslide identification method of the second aspect or any of the corresponding embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0038] Figure 1 is a flowchart of a training method of a landslide identification model according to an embodiment of the present application;

[0039] Figure 2 is a flowchart of another training method of a landslide identification model according to an embodiment of the present application;

[0040] Figure 3 is a flowchart of still another training method of a landslide identification model according to an embodiment of the present application;

[0041] Figure 4 is a flowchart of yet another training method of a landslide identification model according to an embodiment of the present application;

[0042] Figure 5 is a flowchart of a landslide identification method according to an embodiment of the present application;

[0043] Figure 6 is a structural block diagram of a training device of a landslide identification model according to an embodiment of the present application;

[0044] Figure 7 is a structural block diagram of a landslide identification device according to an embodiment of the present application;

[0045] Figure 8 is a hardware structure schematic diagram of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application.

[0047] Since the existing landslide identification methods have different advantages and disadvantages, the embodiments of the present application provide a training method of a landslide identification model, which improves the accuracy of the model by training the landslide identification model, so as to achieve the effect of efficiently and accurately identifying landslides by applying the landslide identification model.

[0048] According to the embodiment of the present application, a landslide identification model training method is provided. It should be noted that 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, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0049] In the present embodiment, a landslide identification model training method is provided, which can be used in a computer device, Figure 1 is a flowchart of the landslide identification model training method according to the embodiment of the present application, as Figure 1 shown, the flow includes the following steps:

[0050] Step S101, acquiring remote sensing images in a preset range and first landslide information corresponding to each region in the preset range.

[0051] Specifically, the first landslide information is used to represent whether the corresponding region has a landslide. For example, the first landslide information can be 1 or 0, 1 indicating a landslide and 0 indicating no landslide. The first landslide information can be measured data collected in the preset range, or landslide information predicted based on a pre-trained landslide prediction model.

[0052] Specifically, the preset range can be a region with rich geological features, which can be selected at will by those skilled in the art, and the present embodiment does not make specific limitations. Each region in the preset range is a unit obtained by dividing the preset range according to a certain division rule. The division rule is determined by those skilled in the art according to the actual situation, which is not specifically limited here.

[0053] Step S102, interpreting the remote sensing images to obtain second landslide information corresponding to each region.

[0054] Specifically, based on the landslide interpretation mark, the remote sensing images obtained in step S101 are visually interpreted by those skilled in the art, thereby obtaining the landslide distribution of the preset range, i.e., the second landslide information corresponding to each region.

[0055] Specifically, the second landslide information is used to represent whether the corresponding region has a landslide. The difference between the second landslide information and the first landslide information is that their information sources are different.

[0056] Step S103, performing spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions.

[0057] Specifically, each region of the preset range corresponds to the first landslide information and the second landslide information. The embodiment determines that each region is an overlapping region or a non-overlapping region by spatially superimposing and analyzing the first landslide information and the second landslide information of the same region.

[0058] In step S104, the target remote sensing image feature is determined based on the remote sensing image feature values corresponding to the overlapping region and the non-overlapping region, respectively.

[0059] Specifically, the remote sensing image features include spectral band values, brightness, texture features, etc., and the remote sensing image feature values corresponding to the same remote sensing image feature in different regions are not completely the same.

[0060] Specifically, the target remote sensing image feature is selected from all remote sensing image features based on the remote sensing image feature values corresponding to the overlapping region and the non-overlapping region, respectively.

[0061] In step S105, the pre-constructed landslide recognition model is trained based on the target remote sensing image feature values corresponding to each region and the second landslide information corresponding to the overlapping region, to obtain a landslide recognition model.

[0062] Specifically, the feature values of the target remote sensing image corresponding to the overlapping region are input into the pre-constructed landslide recognition model, the feature values of the input target remote sensing image are predicted by the landslide recognition model, and the predicted landslide information is obtained. The second landslide information is taken as the true value, and the second landslide information and the predicted landslide information are substituted into the loss function to obtain the corresponding loss value, until the loss value no longer changes with the input data, the model is determined to be converged, at this time, the training operation of the model is stopped, and a trained landslide recognition model is obtained. The pre-constructed landslide recognition model is a random forest machine learning model, and the loss function uses the least squares method to fit the error sum of squares of the observed value and the predicted value.

[0063] The training method of the landslide recognition model provided in the embodiment determines each region in the preset range as an overlapping region and a non-overlapping region by spatially superimposing the first landslide information and the second landslide information, the overlapping region is an accurate region of the remote sensing image, and the non-overlapping region is an inaccurate region of the remote sensing image. The target remote sensing image feature is determined by the remote sensing image feature values corresponding to the overlapping region and the non-overlapping region, respectively, the pre-constructed landslide recognition model is trained by the target remote sensing image feature and the second landslide information of the overlapping region, so that the model learns the correlation between the remote sensing image feature and the landslide information, so that the trained model has the ability to determine the landslide information according to the remote sensing image feature. The trained landslide recognition model can solve the problems of low efficiency, small range, high cost, and low accuracy in the existing landslide recognition method, and make the landslide recognition process more efficient and accurate.

[0064] A landslide identification model training method is provided in the embodiment, which can be used for a computer device, Figure 2 is a flowchart of the landslide identification model training method according to the embodiment of the present application, as Figure 2 shown, the flow includes the following steps:

[0065] Step S201, acquiring remote sensing images in a preset range and first landslide information corresponding to each region in the preset range respectively.

[0066] Specifically, in the step S201, the first landslide information corresponding to each region in the preset range is acquired, including:

[0067] Step S2011, acquiring target influence factors and third landslide information corresponding to each region in the preset range.

[0068] Specifically, when the first landslide information is not the measured data collected in the preset range, the third landslide information is information about whether the corresponding region actually collected has landslide.

[0069] In some optional embodiments, in the step S2011, the target influence factors are acquired, including:

[0070] Step a1, acquiring geographic information and environmental information of each region in the preset range.

[0071] Specifically, the geographic information includes topographic information and geological information, wherein the topographic information is information such as elevation, slope, slope direction, etc., and the geological information is information such as fault distribution, stratum lithology, etc. The environmental information includes information such as vegetation type distribution, land use type, human engineering (roads, water conservancy and hydraulic facilities), etc.

[0072] Step a2, statistically analyzing all the geographic information, environmental information and third landslide information to determine a plurality of initial influence factors causing landslide.

[0073] Specifically, the initial influence factors are information related to landslide determined from the geographic information and environmental information after statistical analysis.

[0074] Illustratively, it is found through statistical analysis that landslides occur more frequently on slopes with a topographic slope of 20-35 degrees, and if the lithology of the slope is mainly soil rock and other easy sliding lithology types, the probability of landslide occurrence will increase significantly, and the change of water level in the reservoir area will also affect the stability of the slope to a certain extent. Therefore, the slope, stratum lithology and water level are information related to landslide, and the above three are determined as initial influence factors.

[0075] Step a3, screening target influence factors from the plurality of initial influence factors according to a preset rule.

[0076] Specifically, the step a3 comprises:

[0077] The step a31 comprises: performing correlation analysis on each initial influence factor in each region and the third landslide information respectively to obtain a first correlation coefficient between each initial influence factor in each region and the landslide.

[0078] Specifically, since the geographical information, the environmental information and the third landslide information corresponding to each region in the preset range are not completely the same, the influence of the same initial influence factor on the landslide in different regions is also different.

[0079] After determining the initial influence factors in the preset range, correlation analysis needs to be performed on each initial influence factor and the third landslide information in each region to obtain a first correlation coefficient corresponding to each influence factor in different regions. The first correlation coefficient is used to represent the influence degree of the initial influence factor on the landslide.

[0080] The step a32 comprises: determining a second correlation coefficient between each initial influence factor and the landslide under the comprehensive action of all initial influence factors by using a random forest method.

[0081] Specifically, the second correlation coefficient is the influence degree of a certain initial influence factor on the landslide relative to other initial influence factors under the action of all initial influence factors. The difference between the first correlation coefficient and the second correlation coefficient is that the first correlation coefficient is an absolute result of the single action of the initial influence factor, and the second correlation coefficient is a relative result under the joint action of multiple initial influence factors. The first correlation coefficients corresponding to the same initial influence factor in different regions are not completely the same, but the second correlation coefficients corresponding to the same initial influence factor in different regions are completely the same.

[0082] The step a33 comprises: determining an initial weight corresponding to each initial influence factor in each region based on the first correlation coefficient and the second correlation coefficient.

[0083] Specifically, the initial weight is the influence degree of a certain influence factor in all initial influence factors causing the landslide, and the value range thereof is 0 to 1.

[0084] Specifically, the initial weight is determined by using Bayesian probability, and the specific determination manner is as follows:

[0085]

[0086] Wherein, f(L i ) is the initial weight corresponding to the i th initial influence factor, f(E i ) is the first correlation coefficient corresponding to the i th initial influence factor, and f(L|E i) is a second correlation coefficient corresponding to the ith initial influence factor, and n is the number of initial influence factors.

[0087] Step a34, based on the initial weight, the preset weight threshold and the third landslide information, the target influence factor is selected from the plurality of initial influence factors.

[0088] Specifically, the above step a34 includes:

[0089] Step a341, the initial weight corresponding to each region in the plurality of initial influence factors is compared with the preset weight threshold, and the region whose initial weight is greater than or equal to the preset weight threshold is determined as the first target region.

[0090] Specifically, the preset weight threshold is an index for measuring the influence degree of the initial influence factor in a certain region on the landslide. If the initial weight of a certain initial influence factor is greater than or equal to the preset weight threshold, it means that the initial influence factor has a greater influence on the landslide, otherwise, it means that the influence is smaller. The preset weight threshold can be set according to the actual situation, which is not limited here.

[0091] Illustratively, the preset weight threshold is 0.5, the initial weight corresponding to the initial influence factor a in each region in the preset range is compared with the preset weight threshold 0.5, and all regions whose initial weight is greater than or equal to 0.5 are determined as the first target region. It should be noted that each initial influence factor has a corresponding first target region.

[0092] Step a342, the region where the landslide occurs determined according to the second landslide information is determined as the second target region.

[0093] Illustratively, in the preset range, all regions whose second landslide information is equal to 1 are determined as the second target region. The second landslide information equal to 1 means that the region has a landslide.

[0094] Step a343, spatial overlay analysis is performed on the first target region and the second target region to determine the overlap rate.

[0095] Illustratively, the overlap rate is the ratio of the number of regions contained in the first target region to the number of regions contained in the second target region. Since the first target region is the region whose initial weight of the initial influence factor a is greater than or equal to 0.5, and the second target region is the region where the landslide occurs, the higher the overlap rate, the higher the relevance of the initial influence factor a to the landslide, and the lower the overlap rate, the lower the relevance of the initial influence factor a to the landslide.

[0096] Step a344, when the overlap rate is greater than or equal to the preset overlap threshold, the second correlation coefficient corresponding to the initial influence factor is adjusted.

[0097] Specifically, the preset overlap threshold is a minimum value representing that the initial influence factor is strongly related to the landslide. When the overlap rate of the first target region and the second target region is greater than or equal to the preset overlap threshold, it is indicated that the initial landslide influence factor a is strongly related to the landslide. At this time, the second correlation coefficient of the initial influence factor a can be adjusted, and the adjustment manner can be to increase 10% on the basis of the original second correlation coefficient. It should be emphasized that steps a341 to a344 are performed for each initial influence factor to determine the correlation of each initial influence factor with the landslide, respectively.

[0098] Step a345, when the adjusted second correlation coefficient is greater than or equal to the preset correlation coefficient threshold, the corresponding initial influence factor is determined as the target influence factor.

[0099] Specifically, the preset correlation coefficient threshold is a standard for determining whether the initial influence factor can be determined as the target influence factor. The preset correlation coefficient threshold can be determined according to the size of the second correlation coefficient corresponding to all initial influence factors. The specific determination manner is not limited here.

[0100] Step S2012, training the pre-constructed landslide prediction model based on the target influence factor and the third landslide information corresponding to each region, to obtain a trained landslide prediction model.

[0101] Specifically, the target influence factor and the third landslide information are input into the landslide prediction model, and the landslide prediction model is trained with the third landslide information as a label. The training process of the model is stopped when the loss value between the predicted landslide information based on the target influence factor and the third landslide information does not change with the input of the target influence factor, and a trained landslide prediction model is obtained.

[0102] The above-mentioned manner of the embodiment enables the landslide prediction model to directly predict the more accurate landslide distribution of the to-be-predicted region based on the data corresponding to the target influence factor of the to-be-predicted region.

[0103] Step S2013, predicting the landslide distribution in the preset range by using the landslide prediction model, to obtain the first landslide information corresponding to each region, respectively.

[0104] Step S202, interpreting the remote sensing image to obtain the second landslide information corresponding to each region, respectively. For details, please refer to Figure 1 The step S102 of the embodiment shown in the figure will not be repeated here.

[0105] Step S203, performing spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region, respectively, to obtain overlapping regions and non-overlapping regions. For details, please refer to Figure 1Step S103 of the illustrated embodiment will not be described here again.

[0106] Step S204, based on the remote sensing image feature values corresponding to the overlapping area and the non-overlapping area respectively, determine the target remote sensing image feature. For details, please refer to Figure 1 Step S104 of the illustrated embodiment will not be described here again.

[0107] Step S205, based on the feature values of the target remote sensing image corresponding to each area and the second landslide information corresponding to the overlapping area, train the pre-constructed landslide identification model to obtain the landslide identification model. For details, please refer to Figure 1 Step S105 of the illustrated embodiment will not be described here again.

[0108] Through the statistics and analysis of data from different sources, the final recognition accuracy and recognition efficiency are higher under the joint action of the two models, which meets the needs of reservoir area geological disaster assessment and prevention.

[0109] In this embodiment, a landslide identification model training method is provided, which can be used for a computer device, Figure 3 is a flowchart of the landslide identification model training method according to the embodiment of the present application, as Figure 3 The flowchart includes the following steps:

[0110] Step S301, obtaining remote sensing images of a preset range and first landslide information corresponding to each area in the preset range. For details, please refer to Figure 2 Step S201 of the illustrated embodiment will not be described here again.

[0111] Step S302, interpreting the remote sensing image to obtain second landslide information corresponding to each area. For details, please refer to Figure 1 Step S102 of the illustrated embodiment will not be described here again.

[0112] Step S303, spatially superimposing and analyzing the first landslide information and the second landslide information corresponding to each area to obtain overlapping areas and non-overlapping areas.

[0113] Specifically, the above step S303 includes:

[0114] Step S3031, determining whether the first landslide information and the second landslide information corresponding to each area are consistent.

[0115] Step S3032, when the first landslide information and the second landslide information are consistent, determining the corresponding area as an overlapping area.

[0116] Step S3033, when the first landslide information and the second landslide information are inconsistent, determining the corresponding area as a non-overlapping area.

[0117] Step S304, based on the remote sensing image feature values corresponding to the overlapping region and the non-overlapping region respectively, determine the target remote sensing image feature. For details, please refer to Figure 1 Step S104 of the embodiment shown will not be repeated here.

[0118] Step S305, based on the feature values of the target remote sensing image corresponding to each region and the second landslide information corresponding to the overlapping region, train the pre-constructed landslide identification model to obtain a landslide identification model. For details, please refer to Figure 1 Step S105 of the embodiment shown will not be repeated here.

[0119] In this embodiment, by the consistency of the first landslide information and the second landslide information, it is determined whether the determination of the landslide situation of the same region by the data of two different sources is consistent, thereby ensuring the accuracy of the label in the subsequent training process, and indirectly making the output result of the model more accurate.

[0120] In this embodiment, a landslide identification model training method is provided, which can be used in a computer device, Figure 4 is a flowchart of the landslide identification model training method according to an embodiment of the present application, as Figure 4 shown, the flowchart includes the following steps:

[0121] Step S401, obtaining remote sensing images of a preset range and first landslide information corresponding to each region in the preset range. For details, please refer to Figure 2 Step S201 of the embodiment shown will not be repeated here.

[0122] Step S402, interpreting the remote sensing images to obtain second landslide information corresponding to each region. For details, please refer to Figure 1 Step S102 of the embodiment shown will not be repeated here.

[0123] Step S403, spatially superimposing and analyzing the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions. For details, please refer to Figure 3 Step S303 of the embodiment shown will not be repeated here.

[0124] Step S404, based on the remote sensing image feature values corresponding to the overlapping region and the non-overlapping region respectively, determine the target remote sensing image feature.

[0125] Specifically, the above step S404 includes:

[0126] Step S4041, obtaining the first remote sensing image feature value corresponding to the overlapping region and the second remote sensing image feature value corresponding to the non-overlapping region.

[0127] In step S4042, the separation degree of the corresponding remote sensing image feature is determined based on the first remote sensing image feature value and the second remote sensing image feature value.

[0128] For example, the first remote sensing image feature value corresponding to the remote sensing image feature A in each overlapping area is obtained, and the first average value and the first variance are determined based on all the first remote sensing image feature values. The second remote sensing image feature value corresponding to the remote sensing image feature A in each non-overlapping area is obtained, and the second average value and the second variance are determined based on all the second remote sensing image feature values. The separation degree of the remote sensing image feature A is determined based on the first average value, the first variance, the second average value and the second variance. The greater the separation degree, the more sensitive the feature is.

[0129] The specific determination method of the separation degree is as follows:

[0130]

[0131] Wherein, B j The separation degree of the jth remote sensing image feature is represented, m1 represents the first average value corresponding to the jth remote sensing image feature, m2 represents the second average value corresponding to the jth remote sensing image feature, σ1 represents the first variance corresponding to the jth remote sensing image feature, and σ2 represents the second variance corresponding to the jth remote sensing image feature.

[0132] In step S4043, the remote sensing image feature with a separation degree greater than or equal to a preset separation degree threshold is determined as the target remote sensing image feature.

[0133] Specifically, the preset separation degree threshold is a threshold for measuring the sensitivity of the remote sensing image feature to the overlapping area and the non-overlapping area. The preset separation degree threshold can be determined by a person skilled in the art according to the actual situation, and the present embodiment is not limited specifically.

[0134] In step S405, the pre-constructed landslide recognition model is trained based on the feature value of the target remote sensing image corresponding to each area and the second landslide information corresponding to the overlapping area, and the landslide recognition model is obtained.

[0135] In the present embodiment, the separation degree corresponding to each remote sensing image feature is determined based on the first remote sensing image feature value and the second remote sensing image feature value. The greater the separation degree of the remote sensing image feature, the easier it is to distinguish the overlapping area and the non-overlapping area based on the remote sensing image feature. Therefore, in the present embodiment, the feature with a separation degree greater than or equal to a preset separation degree threshold is determined as the target remote sensing image feature. The purpose is to facilitate the training of the model based on the target remote sensing image feature and the second landslide information of the overlapping area in the subsequent process, so that the model can better learn the correlation between the target remote sensing image feature and the second landslide information, so that the final model prediction result is more accurate.

[0136] The application also provides a landslide identification method, which can be used for a computer device, Figure 5 is a flowchart of the landslide identification method according to an embodiment of the application, as shown in the figure, the flowchart comprises the following steps: Figure 5

[0137] In step S501, a remote sensing observation image of a region to be identified is obtained.

[0138] In step S502, the remote sensing observation image is input into a pre-trained landslide identification model to obtain an identification result of whether a landslide occurs in the region to be identified.

[0139] As one of the most optimal embodiments in the application, the present application will be described in detail in combination with an actual application scenario.

[0140] A local region in the warehouse area is selected as a sample area (a preset range), the sample area is divided into m sample units (regions) through grid division, field investigation is performed on the sample area, and geographical information, environmental information and landslide information (third landslide information) corresponding to each sample unit are collected. All the geographical information, environmental information and landslide information in the sample area are integrated, counted and analyzed to obtain n initial influence factors related to landslides, which are a1, a2, …, an respectively. n .

[0141] In each sample unit, the correlation between each initial influence factor and landslides is analyzed according to the data corresponding to each initial influence factor and the landslide information of the sample unit, and a first correlation coefficient corresponding to each initial influence factor in each region is obtained, such as α m 1, α m 2, …, α m , n . Wherein, α m , n represents the first correlation coefficient of the nth initial influence factor in the mth sample unit.

[0142] Based on all the initial influence factors in each sample unit and the third landslide information of each sample unit, a random forest method is used to determine the second correlation coefficient between each initial influence factor and landslides under the comprehensive action of all the initial influence factors, such as β1, β2, …, β n .

[0143] According to the first correlation coefficient corresponding to each initial influence factor in each region and the second correlation coefficient corresponding to each initial influence factor, an initial weight corresponding to each initial influence factor in each region is determined, and the initial weights are γ m 1, γ m 2, …, γ m , n ​wherein, γ m , n indicating the initial weight of the nth initial influence factor in the mth sample unit.

[0144] The initial weight of each of the n initial influence factors in each sample unit is compared with a preset weight threshold, and a region with an initial weight greater than or equal to the preset weight threshold is determined as a first target region. Each initial influence factor corresponds to a respective first target region.

[0145] Remote sensing images of the sample area are collected, and the remote sensing images are visually interpreted to obtain second landslide information corresponding to each sample unit. The region determined to have a landslide according to the second landslide information is determined as a second target region.

[0146] The first target region corresponding to each initial influence factor is respectively spatially superimposed with the second target region to obtain an overlap rate corresponding to each initial influence factor. When the overlap rate of a certain initial influence factor is greater than or equal to a preset overlap threshold, the second correlation coefficient corresponding to the initial influence factor is increased by 10%, and the increased second correlation coefficient is compared with a preset correlation coefficient threshold. When the second correlation coefficient is greater than or equal to the preset correlation coefficient threshold, the initial influence factor is determined as a target influence factor.

[0147] The target influence factor determined by the above method and the third landslide information are input into a pre-constructed landslide prediction model, the model is trained to obtain a landslide prediction model, and the remote sensing images of the sample area are input into the landslide prediction model to obtain predicted landslide information (first landslide information).

[0148] The first landslide information and the second landslide information corresponding to each sample unit are spatially superimposed, and the sample units with consistent first landslide information and second landslide information are determined as overlapping regions, and the sample units with inconsistent first landslide information and second landslide information are determined as non-overlapping regions.

[0149] The first remote sensing image feature values corresponding to each remote sensing image feature in the overlapping region and the second remote sensing image feature values corresponding to each remote sensing image feature in the non-overlapping region are collected, and the separation degree corresponding to each remote sensing image feature is determined based on the first remote sensing image feature values and the second remote sensing image feature values. The remote sensing image feature with a separation degree greater than or equal to a separation degree threshold is determined as a target remote sensing image.

[0150] The second landslide information corresponding to the overlapping region and the target remote sensing image are input into a pre-constructed landslide identification model, the model is trained to obtain a trained landslide identification model.

[0151] The remote sensing image of the to-be-identified region is acquired, the remote sensing image is input into the trained landslide identification model, and an identification result of whether the to-be-identified region has a landslide is obtained, so that the identification of the landslide distribution is completed.

[0152] In the embodiment, a training device of a landslide identification model is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be described herein again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, implementation of hardware, or a combination of software and hardware, is also possible and contemplated.

[0153] The embodiment provides a training device of a landslide identification model, as shown in Figure 6 The training device comprises:

[0154] The acquisition module 601 is configured to acquire remote sensing images of a preset range and first landslide information corresponding to each region in the preset range, the first landslide information being used to represent whether the corresponding region has a landslide.

[0155] The interpretation module 602 is configured to interpret the remote sensing images to obtain second landslide information corresponding to each region, the second landslide information being used to represent whether the corresponding region has a landslide.

[0156] The analysis module 603 is configured to perform spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions.

[0157] The determination module 604 is configured to determine target remote sensing image features based on feature values of the remote sensing images corresponding to the overlapping regions and the non-overlapping regions.

[0158] The training module 605 is configured to train a pre-constructed landslide identification model based on the feature values of the target remote sensing images corresponding to each region and the second landslide information corresponding to the overlapping regions, to obtain the landslide identification model.

[0159] In an optional embodiment, the analysis module 603 comprises:

[0160] The judgment sub-module is configured to judge whether the first landslide information and the second landslide information corresponding to each region are consistent.

[0161] The first determination sub-module is configured to determine the corresponding region as an overlapping region when the first landslide information and the second landslide information are consistent.

[0162] The second determination sub-module is configured to determine the corresponding region as a non-overlapping region when the first landslide information and the second landslide information are inconsistent.

[0163] In an optional implementation, the determining module 604 comprises:

[0164] The first obtaining sub-module is configured to obtain a first remote sensing image feature value corresponding to the overlapping area and a second remote sensing image feature value corresponding to the non-overlapping area.

[0165] The third determining sub-module is configured to determine the separation degree of the corresponding remote sensing image feature based on the first remote sensing image feature value and the second remote sensing image feature value.

[0166] The fourth determining sub-module is configured to determine the remote sensing image feature with the separation degree greater than or equal to the preset separation degree threshold as the target remote sensing image feature.

[0167] In an optional implementation, the obtaining module 601 comprises:

[0168] The second obtaining sub-module is configured to obtain the target influence factor and the third landslide information corresponding to each area in the preset range, and the third landslide information is information actually collected about whether the corresponding area has a landslide.

[0169] The training sub-module is configured to train the pre-constructed landslide prediction model based on the target influence factor and the third landslide information corresponding to each area, to obtain the trained landslide prediction model.

[0170] The prediction sub-module is configured to use the landslide prediction model to predict the landslide distribution of the preset range, to obtain the first landslide information corresponding to each area.

[0171] In an optional implementation, the second obtaining sub-module comprises:

[0172] The obtaining unit is configured to obtain geographical information and environmental information of each area in the preset range.

[0173] The statistical analysis unit is configured to statistically analyze all the geographical information, environmental information, and third landslide information, to determine a plurality of initial influence factors causing the landslide.

[0174] The screening unit is configured to screen the target influence factor from the plurality of initial influence factors according to a preset rule.

[0175] In an optional implementation, the screening unit comprises:

[0176] The analysis sub-unit is configured to perform correlation analysis on each initial influence factor in each area and the third landslide information, to obtain a first correlation coefficient between each initial influence factor in each area and the landslide.

[0177] The first determining sub-unit is configured to determine, by using a random forest method, a second correlation coefficient between each initial influence factor and the landslide under the comprehensive action of all the initial influence factors.

[0178] The first determining sub-unit is configured to determine, based on the first correlation coefficient and the second correlation coefficient, an initial weight corresponding to each initial influence factor in each region.

[0179] The screening sub-unit is configured to screen, based on the initial weight, a preset weight threshold, and the third landslide information, a target influence factor from the plurality of initial influence factors.

[0180] In this embodiment, a landslide identification device is also provided, which is used to implement the above-mentioned embodiments and preferred embodiments, and will not be described again. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware, or a combination of software and hardware is also possible and is contemplated.

[0181] This embodiment provides a landslide identification device, as shown in Figure 7 , comprising:

[0182] The acquisition module 701 acquires a remote sensing observation image of a region to be identified.

[0183] The determination module 702 inputs the remote sensing observation image into a landslide identification model trained by the training method of the landslide identification model provided in the first aspect or any of the corresponding embodiments, to obtain an identification result of whether the region to be identified has a landslide.

[0184] The further function descriptions of the above-mentioned various modules and units are the same as those of the above-mentioned corresponding embodiments, and will not be described again.

[0185] The training device of the landslide identification model in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory executing one or more software or fixed programs, and / or other devices that can provide the above-mentioned functions.

[0186] The embodiment of the present application also provides a computer device with the above-mentioned Figure 6 training device of the landslide identification model or Figure 7 landslide identification device.

[0187] Please refer to Figure 8 , Figure 8 is a structural schematic diagram of a computer device provided by an optional embodiment of the present application, as shown in Figure 8As shown, the computer device includes one or more processors 10, memory 20, and interfaces 30 for the various components to communicate with one another. The various components communicate through one or more buses, and can be mounted on a common motherboard or in other manners as appropriate. The processor 10 can execute instructions, for example, stored in the memory 20 to display graphical information for a GUI on an external input / output device, such as a display device coupled to the interface. In some optional implementations, multiple processors and / or multiple buses can be employed as appropriate, such as about the memory 20. Also, various components can be distributed, such as over a network to provide greater functionality and / or redundancy. For example, components can be located on either or both of the same device or distributed among multiple devices. Figure 8 The processor 10 is taken as an example.

[0188] The processor 10 can be a central processing unit, a network processor, or a combination thereof. The processor 10 can further include a hardware chip. The hardware chip can be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device can be a complex programmable logic device, a field programmable logic device, a general array logic, or any combination thereof.

[0189] The memory 20 stores instructions that are executable by the at least one processor 10 to cause the at least one processor 10 to perform the methods illustrated in the above embodiments.

[0190] The memory 20 can include a program region and a data region. The program region can store an operating system and applications required by at least one function. The data region can store data created according to the use of the computer device, and the like. In addition, the memory 20 can include a high-speed random access memory, and can further include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some optional implementations, the memory 20 can optionally include a memory that is remotely located with respect to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0191] The memory 20 can include a volatile memory, such as a random access memory, and can also include a non-volatile memory, such as a flash memory, a hard disk, or a solid state disk. The memory 20 can further include a combination of the above-mentioned types of memories.

[0192] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or communication networks.

[0193] The embodiments of the present application further provide a computer readable storage medium, and the method according to the embodiments of the present application can be implemented in hardware, firmware, or recorded in a storage medium, or be implemented as computer codes stored in a remote storage medium or a non-transitory machine readable storage medium and downloaded through a network and stored in a local storage medium, so that the method described herein can be processed by such software on a storage medium using a general purpose computer, a special purpose processor, or programmable or special hardware. The storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid state disk, etc. Further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that the computer, the processor, the microprocessor controller, or the programmable hardware includes a storage component that can store or receive software or computer codes, when the software or computer codes are accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiments is implemented.

[0194] Although the embodiments of the present application are described in conjunction with the drawings, various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the present application, and such modifications and changes fall within the scope defined by the appended claims.

Claims

1. A method for training a landslide identification model, characterized in that, The method comprises: acquiring remote sensing images in a preset range and first landslide information corresponding to each region in the preset range, the first landslide information being used to represent whether landslide occurs in the corresponding region; interpreting the remote sensing images to obtain second landslide information corresponding to each region, the second landslide information being used to represent whether landslide occurs in the corresponding region, and the second landslide information being different from the first landslide information in information source; performing spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions; determining a target remote sensing image feature based on remote sensing image feature values corresponding to the overlapping regions and the non-overlapping regions: acquiring first remote sensing image feature values corresponding to the overlapping regions and second remote sensing image feature values corresponding to the non-overlapping regions; determining a separation degree of the corresponding remote sensing image feature based on the first remote sensing image feature values and the second remote sensing image feature values; and determining a remote sensing image feature with a separation degree greater than or equal to a preset separation degree threshold as the target remote sensing image feature; training a pre-constructed landslide identification model based on feature values of target remote sensing images corresponding to each region and second landslide information corresponding to the overlapping regions to obtain the landslide identification model: inputting feature values of each target remote sensing image corresponding to the overlapping regions into the pre-constructed landslide identification model, predicting the input target remote sensing image feature values by the landslide identification model to obtain predicted landslide information; taking the second landslide information as a true value, substituting the second landslide information and the predicted landslide information into a loss function to obtain a corresponding loss value, until the loss value no longer changes with the input data, determining that the model converges, at which time the training operation of the model is stopped, and a trained landslide identification model is obtained.

2. The training method of claim 1, wherein, The spatial superposition analysis on the first landslide information and the second landslide information corresponding to each region to obtain overlapping regions and non-overlapping regions comprises: determining whether the first landslide information and the second landslide information corresponding to each region are consistent; when the first landslide information and the second landslide information are consistent, determining the corresponding region as an overlapping region; when the first landslide information and the second landslide information are inconsistent, determining the corresponding region as a non-overlapping region.

3. The training method of claim 1, wherein, Acquiring first landslide information corresponding to each region in the preset range comprises: acquiring target influence factors and third landslide information corresponding to each region in the preset range, the third landslide information being information actually collected on whether landslide occurs in the corresponding region; training a pre-constructed landslide prediction model based on the target influence factors and the third landslide information corresponding to each region to obtain a trained landslide prediction model; using the landslide prediction model to predict landslide distribution in the preset range to obtain first landslide information corresponding to each region.

4. The training method of claim 3, wherein, Acquiring target influence factors comprises: acquiring geographic information and environmental information of each region in the preset range; Statistically and analytically processing all of the geographic information, the environmental information and the third landslide information to determine a plurality of initial influence factors causing the landslide to occur; Screening the target influence factor from the plurality of initial influence factors according to a preset rule.

5. The training method of claim 4, wherein, The screening of the target influence factor from the plurality of initial influence factors according to a preset rule comprises: performing correlation analysis on each of the initial influence factors in each region and the third landslide information respectively to obtain a first correlation coefficient between each of the initial influence factors in each region and the landslide; determining a second correlation coefficient between each of the initial influence factors and the landslide under the comprehensive action of all of the initial influence factors by using a random forest method; determining an initial weight corresponding to each of the initial influence factors in each region based on the first correlation coefficient and the second correlation coefficient; screening the target influence factor from the plurality of initial influence factors based on the initial weight, a preset weight threshold and the third landslide information.

6. The training method of claim 5, wherein, The screening of the target influence factor from the plurality of initial influence factors based on the initial weight, a preset weight threshold and the third landslide information comprises: comparing the initial weight corresponding to each of the initial influence factors in each region with the preset weight threshold respectively, and determining a first target region in which the initial weight is greater than or equal to the preset weight threshold; determining a second target region in which the landslide occurs according to the second landslide information; performing spatial overlay analysis on the first target region and the second target region to determine an overlap rate; adjusting the second correlation coefficient corresponding to the initial influence factor when the overlap rate is greater than or equal to a preset overlap threshold; determining the initial influence factor corresponding to the second correlation coefficient greater than or equal to a preset correlation coefficient threshold as the target influence factor after the adjustment.

7. A method of landslide identification, characterized by, comprises: obtaining a remote sensing observation image of a region to be identified; inputting the remote sensing observation image into a landslide identification model trained by the training method of the landslide identification model of any one of claims 1-6 to obtain an identification result of whether the region to be identified has a landslide. 8.A landslide identification model training device, characterized by comprising: The device comprises: an acquisition module configured to acquire remote sensing images of a preset range and first landslide information corresponding to each region in the preset range, the first landslide information being used to represent whether the corresponding region has a landslide; an interpretation module configured to interpret the remote sensing images to obtain second landslide information corresponding to each region, the second landslide information being used to represent whether the corresponding region has a landslide, and the second landslide information being different from the first landslide information in information source; an analysis module configured to perform spatial overlay analysis on the first landslide information and the second landslide information corresponding to each region respectively to obtain overlapping regions and non-overlapping regions; and an output module configured to output the overlapping regions and the non-overlapping regions. The determination module is configured to determine target remote sensing image features based on feature values of remote sensing images corresponding to the overlapping area and the non-overlapping area respectively. The training module is configured to train a pre-constructed landslide identification model based on feature values of target remote sensing images corresponding to each area and second landslide information corresponding to the overlapping area, to obtain the landslide identification model.

9. The training device of claim 8, wherein, The analysis module includes: The judgment submodule is configured to determine whether the first landslide information and the second landslide information corresponding to each area are consistent. The first determination submodule is configured to determine a corresponding area as the overlapping area when the first landslide information and the second landslide information are consistent. The second determination submodule is configured to determine a corresponding area as the non-overlapping area when the first landslide information and the second landslide information are inconsistent.

10. The exercise device of claim 8, wherein, The acquisition module includes: The second acquisition submodule is configured to acquire target influence factors and third landslide information corresponding to each area in a preset range, the third landslide information being information about whether a landslide occurs in a corresponding area actually collected. The training submodule is configured to train a pre-constructed landslide prediction model based on the target influence factors and the third landslide information corresponding to each area, to obtain a trained landslide prediction model. The prediction submodule is configured to predict a landslide distribution of the preset range by using the landslide prediction model, to obtain the first landslide information corresponding to each area.

11. Landslide identification apparatus, characterized in that The acquisition module is configured to acquire a remote sensing observation image of a region to be identified. The training module is configured to input the remote sensing observation image into a landslide identification model trained by the training method of the landslide identification model according to any one of claims 1-6, to obtain an identification result about whether a landslide occurs in the region to be identified. The memory and the processor are communicatively connected, and the memory stores computer instructions.

12. A computer device, comprising: The processor executes the computer instructions to perform the training method of the landslide identification model according to any one of claims 1-6 or the landslide identification method according to claim 7. ​ 13. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions for causing a computer to execute the landslide identification model training method in any one of claims 1 to 6, or to execute the landslide identification method in claim 7.

Images