Big data-based geological disaster prediction method, apparatus and device, and medium

Abstract

The invention provides a geological disaster prediction method and device based on big data, equipment and a medium, and relates to the technical field of geological disaster prediction. Comprising the steps of performing model training based on obtained current remote sensing image data, meteorological rainfall data and geographic information data of a to-be-predicted area and historical remote sensing image data, historical meteorological rainfall data and historical geographic information data corresponding to sample points in the to-be-predicted area to obtain a preset geological disaster prediction model; according to the method, the predicted geological disaster risk point in the to-be-predicted area is obtained, and the geological disaster risk early warning prompt information is generated and output, so that geological disaster early warning prediction calculation of the whole monitoring area is realized, and the accuracy and timeliness of geological disaster monitoring early warning are greatly improved.

Description

Technical Field

[0001] The present application relates to the technical field of geological disaster prediction, and specifically to a geological disaster prediction method, device, equipment and medium based on big data. Background Art

[0002] As an important part of monitoring information, geological disaster monitoring and early warning is an indispensable part of information construction, and its social value and economic benefits are immeasurable. With the continuous deepening of geological environment information construction and cloud platform construction, the deployment of geological disaster monitoring sensors and the amount of monitoring data in various places are also expanding rapidly.

[0003] The traditional monitoring method is to deploy disaster monitoring points in areas prone to geological disasters, and collect monitoring data through monitoring terminals, sensors, cameras and other equipment, using GPS, embedded software and hardware, sensors, multimedia and other technologies. The content includes surface displacement, deep displacement, crack displacement, rainfall, water level, stress, seepage pressure (groundwater) and video monitoring / monitoring data, as well as group monitoring and prevention data and inspection reports. Not only is the monitoring range very small and early warning judgment difficult, but a single type of sensor data is difficult to reflect changes in disaster-causing factors, seriously affecting the accuracy and timeliness of geological disaster monitoring and early warning.

[0004] Therefore, there is an urgent need to provide a geological disaster prediction method to calculate geological disaster early warnings and predictions for the entire monitoring area. Summary of the invention

[0005] The purpose of this application is to address the above problems and provide a geological disaster prediction method, device, equipment and medium based on big data to improve the accuracy and timeliness of geological disaster monitoring and early warning.

[0006] In the first aspect, the present application provides a geological disaster prediction method based on big data, comprising:

[0007] Obtain the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted;

[0008] Based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model, the predicted geological disaster risk points in the area to be predicted are obtained; the preset geological disaster prediction model is obtained by model training based on the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to the sample points in the area to be predicted; wherein the sample points are sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as the pouring points, and sample non-disaster points outside the basin that meet the preset conditions;

[0009] Generate and output geological disaster risk warning prompt information based on the predicted geological disaster risk points.

[0010] According to the technical solution provided by the present application, historical data of geological disasters in the area to be predicted are obtained; the historical data of geological disasters at least include historical disaster points and their historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data;

[0011] Based on the historical disaster points and their historical geographic information data, sample disaster points and sample non-disaster points in the area to be predicted are obtained, and according to the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data of the sample disaster points and the sample non-disaster points, as well as the disaster results corresponding to the sample disaster points and the sample non-disaster points, model training is performed based on a preset original prediction model to obtain the preset geological disaster prediction model; the preset original prediction model includes an SVM prediction model and / or a random forest prediction model, or a logistic regression prediction model.

[0012] According to the technical solution provided by the present application, the sample disaster points and sample non-disaster points in the area to be predicted are obtained based on the historical disaster points and their historical geographic information data, and the disaster results corresponding to the sample disaster points and the sample non-disaster points are trained based on the preset original prediction model to obtain the preset geological disaster prediction model, including:

[0013] Based on the historical geographic information data of the historical disaster point, the watershed where the historical disaster point is located is determined with the historical disaster point as the pouring point, and all grids in the watershed where the historical disaster point is located are used as sample disaster points, and sample non-disaster points are selected in an area outside the watershed where the historical disaster point is located at a first preset ratio, and the sample disaster points and the sample non-disaster points are used as the sample points;

[0014] The historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and corresponding disaster results of the sample points of the second preset ratio are used as a training set, and the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and corresponding disaster results of the remaining sample points are used as a test set;

[0015] The training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model.

[0016] According to the technical solution provided by the present application, the preset original prediction model is a logistic regression prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model, including:

[0017] Input the training set into the preset original prediction model:

[0018]

[0019] Perform model training; where y i , i=1,…n is the geological disaster response variable, which obeys the binomial distribution B(1,p i );p i is the geological disaster response variable y i The probability of being 1, β0,…β n is the coefficient, x i1 ,…x in is the geological disaster impact variable, L is the likelihood function, threshold is the probability threshold, is the maximum likelihood estimation of the geological hazard response variable, is the geological disaster response variable y i The maximum likelihood estimate of the probability that is 1;

[0020] Maximize the likelihood function and iteratively solve the coefficients β0,…β n The maximum likelihood estimate of obtaining a first intermediate prediction model;

[0021] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the first intermediate prediction model to calculate the predicted disaster results of the sample points of the test set, and the TPR value and the FPR value are calculated according to the predicted disaster results and the disaster results of the sample points of the test set. Based on the TPR value and the FPR value, feature form transformation, feature selection and probability threshold optimization are performed until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

[0022] According to the technical solution provided by the present application, the preset original prediction model is a random forest prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model, including:

[0023] A plurality of sample points are randomly extracted with replacement from the training set, a plurality of features are randomly extracted without replacement from the historical remote sensing image data, the historical meteorological rainfall data, and the historical geographic information data, and a plurality of decision trees are established with the ntree parameter as the optimization parameter to form a random forest;

[0024] The random forest partition is simulated multiple times to calculate the TPR value and the FPR value respectively, and the ntree parameter when the TPR value and the FPR value obtain the optimal value is used as the intermediate optimization parameter to obtain the second intermediate prediction model;

[0025] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the second intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the TPR value and the FPR value are calculated, and the intermediate optimization parameters are optimized until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

[0026] According to the technical solution provided by the present application, the preset original prediction model is an SVM prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model, including:

[0027] The SVM prediction model is established by taking the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the training set as features, taking the Gaussian kernel as the kernel function, and taking the misjudgment loss parameter and the Gaussian kernel parameter as optimization parameters;

[0028] The optimization parameters are simulated multiple times to calculate TPR values ​​and FPR values ​​respectively, and based on the TPR values ​​and FPR values, the misjudgment loss parameter and Gaussian kernel parameter corresponding to when the TPR value obtains a maximum value are used as the intermediate loss parameter and the intermediate Gaussian kernel parameter to obtain a third intermediate prediction model;

[0029] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the third intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the TPR value and the FPR value are calculated, and the intermediate loss parameters and intermediate Gaussian kernel parameters are optimized until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

[0030] According to the technical solution provided by the present application, the preset original prediction model includes a random forest prediction model and an SVM prediction model, and the preset geological disaster prediction model includes a first prediction sub-model established based on the random forest prediction model and a second prediction sub-model established based on the SVM prediction model; the acquisition of predicted geological disaster risk points in the area to be predicted based on the remote sensing image data, meteorological rainfall data, geographic information data and the preset geological disaster prediction model includes:

[0031] A first risk point obtained according to the first prediction sub-model based on the remote sensing image data, the meteorological rainfall data, and the geographic information data;

[0032] A second risk point obtained according to the second prediction sub-model based on the remote sensing image data, the meteorological rainfall data, and the geographic information data;

[0033] The union of the first risk point and the second risk point is used as the predicted geological disaster risk point.

[0034] In a second aspect, the present application provides a geological disaster prediction device based on big data, comprising:

[0035] The acquisition module is used to obtain the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted;

[0036] A prediction module, for obtaining predicted geological disaster risk points in the area to be predicted based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model; the preset geological disaster prediction model is obtained by model training based on historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to sample points in the area to be predicted; wherein the sample points are sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as pouring points, and sample non-disaster points outside the basin that meet preset conditions;

[0037] The early warning module is used to generate and output geological disaster risk early warning prompt information based on the predicted geological disaster risk points.

[0038] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described above when executing the computer program.

[0039] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and the computer program implements the method described above when executed by a processor.

[0040] Compared with the prior art, the beneficial effects of the present application are as follows: the geological disaster prediction method, device, equipment and medium based on big data provided by the present application obtain the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted; based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model, the predicted geological disaster risk points in the area to be predicted are obtained; the preset geological disaster prediction model is obtained by model training based on the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to the sample points in the area to be predicted; the target geological disaster risk points are obtained based on the predicted geological disaster risk points and preset conditions, and disaster risk warnings are issued for the target geological disaster risk points. By using relatively high-precision sensors to monitor historical data, and through meteorological rainfall big data, satellite remote sensing image big data and geographic information big data, based on the geological disaster prediction model established through machine learning algorithms such as logistic regression, random forest and SVM support vector machine, the geological disaster incidence rate of each point in the whole region is calculated at a certain level of accuracy, and geological disaster early warning prediction and calculation are realized for the entire monitoring area, and early warning is issued for locations in the whole region where geological disasters may occur in the future. The results can be extended to the entire geological disaster monitoring area, which not only realizes the geological disaster early warning prediction and calculation for the entire monitoring area, but also greatly improves the accuracy and timeliness of geological disaster monitoring and early warning.

[0041] It should be understood that the description of technical features, technical solutions, beneficial effects or similar language in this application does not imply that all features and advantages can be realized in any single embodiment. On the contrary, it is understood that the description of features or beneficial effects means that specific technical features, technical solutions or beneficial effects are included in at least one embodiment. Therefore, the description of technical features, technical solutions or beneficial effects in this specification does not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions and beneficial effects described in the present embodiment can also be combined in any appropriate manner. Those skilled in the art will understand that the embodiment can be realized without one or more specific technical features, technical solutions or beneficial effects of a specific embodiment. In other embodiments, additional technical features and beneficial effects can also be identified in a specific embodiment that does not embody all embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution in this embodiment, the drawings required for use in the description of the embodiment will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1A flowchart of a geological disaster prediction method based on big data provided in an embodiment of the present application;

[0044] Figure 2 It is a schematic diagram of the relationship between pouring points and watershed distribution;

[0045] Figure 3 A schematic diagram of the watershed where the debris flow disaster occurred and the sample disaster points;

[0046] Figure 4 A schematic diagram of the distribution of sample non-disaster points provided in the embodiment of the present application;

[0047] Figure 5 A schematic diagram of the ROC curve provided in the embodiment of the present application;

[0048] Figure 6 A histogram of coefficients corresponding to landform types provided in an embodiment of the present application;

[0049] Figure 7 A schematic diagram of the distribution of predicted geological disaster risk points obtained based on a preset geological disaster prediction model using a logistic regression prediction model as a preset original prediction model provided in an embodiment of the present application;

[0050] Figure 8 A schematic diagram of a probability grid of predicted geological disaster risk points obtained based on a preset geological disaster prediction model using a logistic regression prediction model as a preset original prediction model provided in an embodiment of the present application;

[0051] Fig. 9 A schematic diagram of the distribution of predicted geological disaster risk points obtained based on a preset geological disaster prediction model using a random forest prediction model as a preset original prediction model provided in an embodiment of the present application;

[0052] Fig.10 A line graph showing the trend of TPR and FPR values ​​provided in the embodiment of the present application;

[0053] Fig.11 A schematic diagram of the distribution of predicted geological disaster risk points obtained based on a preset geological disaster prediction model using an SVM prediction model as a preset original prediction model provided in an embodiment of the present application;

[0054] Fig.12 A schematic diagram of the actual distribution of disaster locations provided in the embodiments of the present application;

[0055] Fig.13 A schematic diagram of the distribution of the first risk points predicted by the first prediction sub-model provided in an embodiment of the present application;

[0056] Fig.14 A schematic diagram of the distribution of second risk points predicted by the second prediction sub-model provided in an embodiment of the present application;

[0057] Fig.15 A schematic diagram of a geological disaster prediction device based on big data provided in an embodiment of the present application;

[0058] Fig.16 A schematic diagram of a computer system of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to enable those skilled in the art to better understand the technical solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings. The description in this section is only exemplary and explanatory and should not have any limiting effect on the protection scope of the present application. Specifically, the described embodiments are only embodiments of a part of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work should fall within the scope of protection of the present application.

[0060] It should be noted that similar reference numerals and letters represent similar items in the following figures, so once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0061] In order to facilitate the understanding of this application, a brief introduction to the related technology is first given:

[0062] The geological and geographical environment in mountainous areas is complex, and landslides, debris flows, ground collapses, and ground subsidence are the main types of geological disasters. Among them, landslides and debris flows are widespread, active, and sudden, and are the main types of disasters that cause loss of life and property, restrict economic and social development, and are also the main targets of geological disaster prevention and control. The traditional method of relying on sensor monitoring faces the following difficulties:

[0063] (1) The monitoring range is very small

[0064] The monitoring range of a set of geological sensors is very small, usually less than 100 meters in radius. They can only focus on monitoring certain geological disaster-prone areas and cannot monitor all places in the entire area.

[0065] (2) Difficulty in early warning judgment

[0066] The traditional monitoring method that relies on sensors generally sets early warning thresholds for sensors. If the monitored value exceeds the threshold range, an alarm will be generated. However, the early warning threshold for a single sensor is generally set based on experience. Basically, a threshold is set for each sensor. However, the actual situation when a geological disaster occurs will vary depending on the terrain, surface cover and rainfall. It is difficult to reflect the changes in these disaster-causing factors through data from a single type of sensor.

[0067] In view of this, an embodiment of the present application provides a geological disaster prediction method based on big data. The method uses sensors with relatively high accuracy to monitor historical data, and uses meteorological rainfall big data, satellite remote sensing image big data and geographic information big data. Based on a geological disaster prediction model established through machine learning algorithms such as logistic regression, random forest and SVM support vector machine, the geological disaster incidence rate of each point in the entire region is calculated at a certain level of accuracy, and geological disaster early warning prediction calculations are realized for the entire monitored area. Warnings are issued for locations in the entire region where geological disasters may occur in the future. The results can be extended to the entire geological disaster monitoring area, which not only realizes the geological disaster early warning prediction calculations for the entire monitoring area, but also greatly improves the accuracy and timeliness of geological disaster monitoring and early warning.

[0068] In order to make the technical solution of the present application clearer and easier to understand, the geological disaster prediction method based on big data provided in the embodiment of the present application is introduced below.

[0069] like Figure 1 As shown in the figure, this figure is a flow chart of a geological disaster prediction method based on big data provided by this embodiment, and the method includes the following steps:

[0070] S101, obtaining current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted;

[0071] Specifically, in combination with the causes of landslides, mud-rock flows, etc. in disaster geology, remote sensing image data, meteorological rainfall data, and geographic information data are selected as analysis objects to comprehensively cover all locations in the area to be predicted, and obtain remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted at the current moment; wherein, the area to be predicted can be an area delineated according to demand, or it can be the entire geological disaster monitoring area, which can be set and adjusted according to actual conditions, and is not specifically limited here.

[0072] The remote sensing image data includes landform type, surface evapotranspiration (evap) and vegetation index. The vegetation index is Enhanced Vegetation Index (EVI) or Normalized Difference Vegetation Index (NDVI). The remote sensing image is a multispectral satellite remote sensing image or a hyperspectral satellite remote sensing image. In addition to providing visible light image data of multiple bands, it also provides image data of different bands such as near infrared and far infrared. It can effectively identify the type of surface cover that could not be identified by relying on a single full-color spectrum before. Combined with surface evapotranspiration evap and vegetation index EVI / NDVI, the surface cover of the area to be predicted can be monitored and analyzed. There are many calculation models for surface evapotranspiration. Taking the PM model with clear physical meaning and not affected by surface temperature as an example, soil evaporation and vegetation transpiration are calculated separately. The calculation formula is as follows:

[0073]

[0074] Where, λE is the latent heat flux (W / m2), λ is the latent heat coefficient of vaporization, s is the slope of the saturated water vapor pressure versus temperature curve (kPa / ℃), ρ is the air density (kg / m3), cP is the specific heat of air at constant pressure, which is 1 004J / (kg·K), and e s is the saturated water vapor pressure (kPa), e a is the actual water vapor pressure (kPa), γ is the psychrometric constant, r a is the aerodynamic impedance (s / m), r s is the canopy impedance (s / m).

[0075] The calculation formula of Enhanced Vegetation Index (EVI) is:

[0076]

[0077] Among them, EVI is the enhanced vegetation index, ρ NIR is the near-infrared reflectance value, ρ Red is the red light reflectance value, ρ Blue is the blue light reflection value.

[0078] The normalized difference vegetation index NDVI can partially eliminate the effects of solar altitude angle, satellite observation angle, terrain changes, clouds / shadows and atmospheric attenuation through ratio processing. The calculation formula is as follows:

[0079]

[0080] Among them, NDVI is the normalized difference vegetation index, ρ NIR is the near-infrared reflectance value, ρRed is the red light reflectance value.

[0081] There are many sources for obtaining meteorological rainfall data. The rainfall monitoring data from various meteorological stations of the Meteorological Bureau has the best quality, is timely, accurate, standard, and has a wide coverage. Spatial interpolation is performed based on the coordinate positions of each meteorological station, such as using the inverse distance weighting method, to obtain the rainfall at each location in the entire area. Among them, the inverse distance weighting method (IDW) is analogous to the law of gravity, and believes that the influence of the surrounding location points on themselves is inversely proportional to the square of the distance. The calculation formula is as follows:

[0082]

[0083] Among them, n is the number of surrounding location points participating in spatial interpolation, M i is the value of the i-th position, d i It should be noted that the meteorological rainfall data in this application include the daily rainfall RF_1 one day before the disaster, the daily rainfall RF_2 two days before the disaster, the daily rainfall RF_3 three days before the disaster, and the cumulative rainfall RF_for3Day three days before the disaster. Other meteorological rainfall data may also be included, which can be set and adjusted according to actual conditions and are not specifically limited here.

[0084] Geographic information data includes digital elevation model (DEM) data and slope data calculated based on the DEM data; DEM data is basic geographic information data, which records the height of the surface relative to the sea level at each place in the region; by performing slope analysis on DEM data, the steepness of the terrain at each place in the region, i.e., the slope, can be obtained; the slope is the angle between the normal and the plumb line, and the slope can be approximately solved using the slope formula of two adjacent points:

[0085]

[0086] Among them, θ is the slope, Δx, Δy, and Δz represent the differences between two adjacent points in x-coordinate, y-coordinate, and z-coordinate, respectively.

[0087] S102, based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model, obtain predicted geological disaster risk points in the area to be predicted; the preset geological disaster prediction model is obtained by model training based on the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to the sample points in the area to be predicted; wherein the sample points are sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as pouring points, and sample non-disaster points outside the basin that meet preset conditions;

[0088] Specifically, historical data of geological disasters in the area to be predicted are obtained, and the historical data of geological disasters at least include historical disaster points and their historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data; based on the historical disaster points and their historical geographic information data, sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as the pouring points and sample non-disaster points randomly selected at a first preset ratio outside the basin are obtained, and according to the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data of the sample disaster points and the sample non-disaster points and the disaster results corresponding to the sample disaster points and the sample non-disaster points, model training is performed based on the preset original prediction model to obtain the preset geological disaster prediction model; the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted are input into the preset geological disaster prediction model to obtain the predicted geological disaster risk points in the area to be predicted. Among them, the preset condition can be the first preset ratio of the sample non-disaster points to the total number of the sample points, or other conditions, which can be set and adjusted according to the actual situation, and are not specifically limited here. Wherein, the preset original prediction model includes an SVM prediction model and / or a random forest prediction model, or a logistic regression prediction model.

[0089] S103: Generate and output geological disaster risk warning prompt information based on the predicted geological disaster risk points.

[0090] Specifically, the geological disaster risk warning prompt information includes the geographical location information of the predicted geological disaster risk point, so as to prompt the staff to deploy risk avoidance measures for the predicted geological disaster risk point. It is understandable that in the actual geological disaster monitoring and early warning work, in addition to the prediction results of the present invention, it is also necessary to combine the reports and confirmations of monitoring personnel from various places to do a good job in disaster early warning management, so as to effectively protect the safety of people's lives and property, and prevent excessively frequent false alarms from affecting people's normal life and production.

[0091] On the basis of the above embodiment, further, the method further includes:

[0092] Acquire historical data of geological disasters in the area to be predicted; the historical data of geological disasters at least includes historical disaster points and their historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data;

[0093] Specifically, when obtaining the historical data of geological disasters in the area to be predicted, the parameter type of the historical data of geological disasters is consistent with the type and attribute of the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted collected during the prediction process.

[0094] Based on the historical disaster points and their historical geographic information data, sample disaster points and sample non-disaster points in the area to be predicted are obtained, and according to the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data of the sample disaster points and the sample non-disaster points, as well as the disaster results corresponding to the sample disaster points and the sample non-disaster points, model training is performed based on a preset original prediction model to obtain the preset geological disaster prediction model; the preset original prediction model includes an SVM prediction model and / or a random forest prediction model, or a logistic regression prediction model.

[0095] Specifically, based on the historical geographic information data of the historical disaster point, the watershed where the historical disaster point is located is determined with the historical disaster point as the pouring point, and all grids within the watershed where the historical disaster point is located are used as sample disaster points, and sample non-disaster points are selected in an area outside the watershed where the historical disaster point is located at a first preset ratio, and the sample disaster points and the sample non-disaster points are used as the sample points; the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and the corresponding disaster results of the sample points at a second preset ratio are used as a training set, and the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and the corresponding disaster results of the remaining sample points are used as a test set; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, so as to obtain the preset geological disaster prediction model.

[0096] On the basis of the above embodiment, further, the sample disaster points and sample non-disaster points in the area to be predicted are obtained based on the historical disaster points and their historical geographic information data, and the disaster results corresponding to the sample disaster points and the sample non-disaster points are respectively trained based on the preset original prediction model to obtain the preset geological disaster prediction model, including:

[0097] Based on the historical geographic information data of the historical disaster point, the watershed where the historical disaster point is located is determined with the historical disaster point as the pouring point, and all grids in the watershed where the historical disaster point is located are used as sample disaster points, and sample non-disaster points are selected in an area outside the watershed where the historical disaster point is located at a first preset ratio, and the sample disaster points and the sample non-disaster points are used as the sample points;

[0098] Specifically, based on the historical geographic information data (DEM data) of historical disaster points where disasters have occurred, the recorded historical disaster points are used as pouring points to determine the watershed where they are located, and all grids in the watershed are regarded as debris flow disaster points, that is, sample disaster points. Figure 2 As shown in the figure, the pour point is the point where water flows out of a certain area. By determining a pour point, we can determine an area based on the DEM data. The water in the area will be discharged from the pour point, and this area is the watershed determined by the pour point. For example, Figure 3 The river basin where the debris flow disaster occurred on May 10, 2014 is located. Figure 3 As shown, all grid points in the watershed are determined as disaster points. In the area outside the watershed where the historical disaster point is located, multiple location points are randomly selected at a first preset ratio as sample non-disaster points (such as Figure 4 The sample disaster points and the sample non-disaster points are taken as the sample points. In addition, the first preset ratio can be set and adjusted according to actual conditions. Preferably, the value of the first preset ratio is selected so that the sample non-disaster points account for 10% of the total number of sample points.

[0099] The historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and corresponding disaster results of the sample points of the second preset ratio are used as a training set, and the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and corresponding disaster results of the remaining sample points are used as a test set;

[0100] Specifically, for example, as shown in Table 1, 80% of the sample points are randomly selected from all the sample points, and the daily rainfall RF_1, daily rainfall RF_2, daily rainfall RF_3, cumulative rainfall RF_for3Day, slope, DEM data, landform type landform and their corresponding disaster results are used as training sets, and the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and their corresponding disaster results of the remaining 20% ​​of the sample points are used as test sets. Among them, the second preset ratio can be set and adjusted according to actual conditions; it should be noted that, preferably, both the training set and the test set include both sample disaster points and non-sample disaster points to improve the efficiency and accuracy of model training and testing.

[0101] The training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model.

[0102] Table 1

[0103]

[0104]

[0105] On the basis of the above embodiment, further, the preset original prediction model is a logistic regression prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, and the preset geological disaster prediction model is obtained, including:

[0106] Input the training set into the preset original prediction model:

[0107]

[0108] Perform model training; where y i , i=1,…n is the geological disaster response variable, which obeys the binomial distribution B(1,p i );p i is the geological disaster response variable y i The probability of being 1, β0,…β n is the coefficient, x i1 ,…x in is the geological disaster impact variable, L is the likelihood function, threshold is the probability threshold, is the maximum likelihood estimation of the geological hazard response variable, is the geological disaster response variable yi The maximum likelihood estimate of the probability that is 1;

[0109] Maximize the likelihood function and iteratively solve the coefficients β0,…β n The maximum likelihood estimate of obtaining a first intermediate prediction model;

[0110] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the first intermediate prediction model to calculate the predicted disaster results of the sample points of the test set, and the TPR value and the FPR value are calculated according to the predicted disaster results and the disaster results of the sample points of the test set. Based on the TPR value and the FPR value, feature form transformation, feature selection and probability threshold optimization are performed until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

[0111] Specifically, logistic regression is a regression model used to fit binary classification problems. In this application, it is assumed that the geological disaster response variable y i ,i=1,…n each follows a binomial distribution B(1,p i ),in:

[0112] p i =P(y i =1).

[0113] Logistic regression assumes that p i There is a linear relationship between the logit transformation and the independent variable, namely:

[0114]

[0115] Substituting into the likelihood function:

[0116]

[0117] Maximize the likelihood function and iteratively solve the maximum likelihood estimate of β Obtain the first intermediate prediction model, and then obtain p i Estimates

[0118] In order to predict the response variable as 0 or 1, it is necessary to set a suitable probability threshold, so that the predicted value of the response variable is:

[0119]

[0120] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the first intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the disaster results (true values) and the predicted disaster results of the sample points of the test set are subjected to frequency statistics, and a two-dimensional contingency table is listed as shown in Table 2:

[0121] Table 2

[0122]

[0123] The True Positive Rate (TPR) and False Positive Rate (FPR) are defined as the proportion of correctly predicting a value of 1 as 1 and the proportion of misjudging a value of 0 as 1, respectively:

[0124]

[0125]

[0126] Among them, a 00 is the frequency of correctly predicting 0 as 0, a 01 is the frequency of mispredicting 0 as 1, a 10 is the frequency of mispredicting 1 as 0, a 11 is the frequency of correctly predicting 1 as 1.

[0127] In order to improve the prediction accuracy as much as possible, it is necessary to reduce the FPR value on the basis of maximizing the TPR value. However, it should be noted that there is a correlation between the TPR value and the FPR value: the higher the probability threshold is set, the fewer sample points are predicted to be 1, and the TPR value and FPR value will be reduced accordingly; but when the probability threshold is set low, the more sample points are predicted to be 1, and the TPR value and FPR value will increase. Therefore, according to different probability thresholds, the corresponding TPR value and FPR value can be plotted, that is, the Receiver Operating Characteristic curve (ROC) is obtained, such as Figure 5As shown in the figure, the horizontal axis is the FPR value and the vertical axis is the TPR value. In order to make the model obtain the best prediction results, the probability value at the upper left vertex of the ROC curve is usually taken as the probability threshold, which is also the value that makes TPR-FPR reach the maximum. The area enclosed under the ROC curve is recorded as AUC (Area Under the Curve,). The maximum value of AUC is 1. The larger the value, the higher the accuracy of the model. Generally speaking, AUC reaching 0.7 to 0.9 can be considered as a prediction with high accuracy, and above 0.9 is considered as a prediction with very high accuracy.

[0128] Due to the asymmetric distribution of rainfall variables, it is considered to perform central standardization on rainfall variables, introduce different variables into the model, calculate the prediction accuracy of different models, and perform feature form transformation. The historical data of the area to be predicted is simulated 100 times, and the results are shown in Table 3 below:

[0129] Table 3

[0130]

[0131] In the above results, TPR fluctuates around 85%-86% and FPR fluctuates around 17%-18% for the original data and the center-normalized data. From the AUC criterion, the AUC for the original data and the center-normalized data does not change much, so in order to accurately interpret the regression coefficient, the center-normalized data is used for logistic regression.

[0132] In order to further screen out suitable variables, the whole model was selected according to the Akaike information criterion (AIC). The initial independent variables included: daily rainfall RF_1 one day before the disaster, daily rainfall RF_2 two days before the disaster, daily rainfall RF_3 three days before the disaster, slope, DEM data, landform type landform, surface evapotranspiration evap, and normalized vegetation index NDVI. The above independent variables were not removed after screening.

[0133] After feature form and feature selection, it was finally determined that the centrally standardized RF_1, RF_2, RF_3, dem, slope, evap, vni, and landforms were used as the features introduced into the model for model training, testing, and optimization. The above 8 variables were used to establish the logistic regression model coefficients as shown in Table 4:

[0134] Table 4

[0135]

[0136]

[0137] In the above table, except for the geomorphic variables, the coefficient β of the other variables represents: the odds ratio increases by one standard deviation of the variable. Increase (or decrease) exp(β), so the larger the coefficient β of the variable, the greater the influence of the variable. Obviously, the rainfall on the previous day has a greater impact on the occurrence of debris flow, and the rainfall on the previous two days has the second largest impact.

[0138] As for landform, in order to compare the influence of different landforms on the occurrence of debris flow, a bar chart is made for its coefficient, such as Figure 6 As shown, compared with landform 12, areas with landforms 32, 33, 42, 43, 52, 53, 62, and 63 are more prone to disasters, while areas with other landform characteristics are less likely to be affected by disasters.

[0139] For example, Figure 7 As shown in the figure, all the training data on May 11, 2014 are layered and colored according to the logistic regression predicted probability values. It can be clearly seen from the figure that the disaster rate along the three valleys with lower terrain is slightly higher than that in other areas. After IDW interpolation of the above points according to the logistic regression predicted probability values, the logistic regression probability grid can be obtained as follows Figure 8 As shown, the darker the color, the higher the predicted probability of a disaster. From the figure, we can clearly see three high-risk zones for debris flow.

[0140] On the basis of the above embodiment, further, the preset original prediction model is a random forest prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, and the preset geological disaster prediction model is obtained, including:

[0141] A plurality of sample points are randomly extracted with replacement from the training set, a plurality of features are randomly extracted without replacement from the historical remote sensing image data, the historical meteorological rainfall data, and the historical geographic information data, and a plurality of decision trees are established with the ntree parameter as the optimization parameter to form a random forest;

[0142] The random forest partition is simulated multiple times to calculate the TPR value and the FPR value respectively, and the ntree parameter when the TPR value and the FPR value obtain the optimal value is used as the intermediate optimization parameter to obtain the second intermediate prediction model;

[0143] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the second intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the TPR value and the FPR value are calculated, and the intermediate optimization parameters are optimized until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

[0144] Specifically, for example, if there are N sample points in the training set, and the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data include M features, then n samples are randomly selected with replacement from all N sample points, and m features are randomly selected without replacement from the M features as the classification basis of the decision tree, and a decision tree is formed by splitting under a certain criterion (such as information gain). This is repeated many times to establish a large number of decision trees to form a random forest. When a new sample comes in and needs to be predicted, it is predicted by each decision tree, and finally a simple vote is used to determine the predicted classification of the sample.

[0145] Since there are 6 characteristic values, namely daily rainfall RF_1 one day before the disaster, daily rainfall RF_2 two days before the disaster, daily rainfall RF_3 three days before the disaster, and cumulative rainfall RF_for3Day three days before the disaster, and DEM data and slope Slope are all continuous values, it is necessary to design a decision tree division method for continuous values.

[0146] Historical sample set D and continuous attribute a, assuming that a has n different values ​​on D, sort these values ​​from small to large and record them as {a1, a2,…, an}. Based on the partition point t, D can be divided into subsets Dt- and Dt+, where Dt- contains samples whose values ​​on attribute a are not greater than t, and Dt+ contains samples whose values ​​on attribute a are greater than t. Obviously, for adjacent attribute values ​​ai and ai+1, the partition results produced by t taking any value in the interval [ai, ai+1] are the same. The median point of the interval [ai, ai+1] is taken as the candidate partition point. The above process is simulated 100 times to obtain the TPR value and FPR value. According to the TPR value and FPR value, it is known that when the ntree parameter is 20, the prediction effect of the random forest model is the best, and the TPR value is 92.56% and the FPR value is 6.49%.

[0147] In terms of variable importance, RF_1 has an importance of 1075.711, which has the most significant impact on the occurrence of debris flow; followed by RF_2 and RF_3, and the influence of other variables is relatively weak. However, if there are too many values ​​of landforms, the random forest will tend to give this variable greater importance, so its importance value will be biased. For example, the geological disaster prediction results for May 11, 2014 were obtained by applying the preset geological disaster prediction model established with the random forest prediction model as the preset original prediction model. Fig. 9 shown.

[0148] On the basis of the above embodiment, further, the preset original prediction model is an SVM prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, and the preset geological disaster prediction model is obtained, including:

[0149] The SVM prediction model is established by taking the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the training set as features, taking the Gaussian kernel as the kernel function, and taking the misjudgment loss parameter and the Gaussian kernel parameter as optimization parameters;

[0150] The optimization parameters are simulated multiple times to calculate TPR values ​​and FPR values ​​respectively, and based on the TPR values ​​and FPR values, the misjudgment loss parameter and Gaussian kernel parameter corresponding to when the TPR value obtains a maximum value are used as the intermediate loss parameter and the intermediate Gaussian kernel parameter to obtain a third intermediate prediction model;

[0151] The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the third intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the TPR value and the FPR value are calculated, and the intermediate loss parameters and intermediate Gaussian kernel parameters are optimized until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

[0152] Specifically, in order to maximize the TPR value and reduce the FPR value, the misjudgment loss parameter (cost) and the Gaussian kernel parameter (gamma) are simulated multiple times to obtain the TPR and FPR values ​​under different models, and the parameters at the maximum point of TPR are found. As the gamma value increases, both the TPR and FPR values ​​tend to decrease. When gamma increases from 1 to 2, the TPR value decreases by an average of 1.3%, and the FPR value decreases by an average of 1.5%; when gamma increases from 2 to 3, the TPR value decreases by an average of 1.9%, and the FPR value decreases by an average of 1%; when gamma increases from 3 to 4, the TPR value decreases by an average of 2%, and the FPR value decreases by an average of 0.8%; when gamma increases from 4 to 5, the TPR value decreases by an average of 2%, and the FPR value decreases by an average of 0.4%. Plot a line graph of the decreasing trend of the TPR and FPR values ​​as shown below: Fig.10 As shown in the figure, from the trend of the line graph, as gamma increases, the TPR value decreases more and the FPR value decreases less. This shows that although the increase in gamma will lead to a decrease in the FPR value, the TPR value will also decrease, and the decrease will be greater, so the cost of reducing the FPR value is too high. When gamma increases from 1 to 2, the TPR value decreases by 1.3%, the FPR value decreases by 1.5%, and when gamma = 2, the TPR value remains at around 96%, so this increase is valuable. Finally, gamma = 2 and cost = 100 are selected as the optimal parameters for SVM. At this time, the TPR value of the model is 96.06%, and the FPR value is 4.7%, and the preset geological disaster prediction model is obtained.

[0153] The geological disaster prediction results for May 11, 2014 based on the preset geological disaster prediction model are as follows: Fig.11 shown.

[0154] On the basis of the above embodiment, further, the preset original prediction model includes a random forest prediction model and an SVM prediction model, the preset geological disaster prediction model includes a first prediction sub-model established based on the random forest prediction model and a second prediction sub-model established based on the SVM prediction model; the acquisition of predicted geological disaster risk points in the area to be predicted based on the remote sensing image data, meteorological rainfall data, geographic information data and the preset geological disaster prediction model includes:

[0155] A first risk point obtained according to the first prediction sub-model based on the remote sensing image data, the meteorological rainfall data, and the geographic information data;

[0156] A second risk point obtained according to the second prediction sub-model based on the remote sensing image data, the meteorological rainfall data, and the geographic information data;

[0157] The union of the first risk point and the second risk point is used as the predicted geological disaster risk point.

[0158] Specifically, the preset original prediction model includes a random forest prediction model and an SVM prediction model, the preset geological disaster prediction model includes a first prediction sub-model established based on the random forest prediction model and a second prediction sub-model established based on the SVM prediction model, the first risk point obtained according to the first prediction sub-model based on the remote sensing image data, meteorological rainfall data, and geographic information data is as shown in Table 5; the second risk point obtained according to the second prediction sub-model based on the remote sensing image data, meteorological rainfall data, and geographic information data is as shown in Table 6:

[0159] Table 5

[0160] True value / predicted value 0 1 0 53464 208 1 608 1152

[0161] Table 6

[0162] True value / predicted value 0 1 0 51689 182 1 2383 1178

[0163] In the two tables, the value of the first row and the first column represents the number of samples that were correctly predicted as non-disaster samples, the value of the first row and the second column represents the number of samples that were incorrectly predicted as non-disaster samples, the value of the second row and the first column represents the number of samples that were incorrectly predicted as disaster samples, and the value of the second row and the second column represents the number of samples that were correctly predicted as disaster samples. In comparison, the first prediction sub-model predicted more samples that were disaster samples, but correspondingly, more samples that were not disaster samples were incorrectly judged as disaster samples.

[0164] If the first prediction sub-model and the second prediction sub-model are combined, for any input sample, the first prediction sub-model and the second prediction sub-model are used to predict it respectively. If any one of them predicts that a disaster will occur, it will be determined as a predicted geological disaster risk point, that is, the union of the first risk point obtained by the first prediction sub-model and the second risk point obtained by the second prediction sub-model is used as the predicted geological disaster risk point; only when both the first prediction sub-model and the second prediction sub-model predict that no disaster will be occurred, it will be determined that it is not a predicted geological disaster risk point. Such prediction results are shown in Table 7:

[0165] Table 7

[0166] True value / predicted value 0 1 0 51256 60 1 2716 1300

[0167] It can be seen that in this case, only 60 disaster samples were not predicted, and the number of non-disaster samples increased by only 333 misjudged disasters compared to the results of random forest. Fig.12 As shown, the first risk point distribution predicted by the first prediction sub-model is as follows Fig.13 As shown, the second risk point distribution predicted by the second prediction sub-model is as follows Fig.14 shown.

[0168] The geological disaster prediction method based on big data provided in the present application obtains the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted; based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model, the predicted geological disaster risk points in the area to be predicted are obtained; the preset geological disaster prediction model is obtained by model training based on the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to the sample points generated by the historical data of geological disasters in the area to be predicted; the target geological disaster risk points are obtained based on the predicted geological disaster risk points and preset conditions, and disaster risk warnings are issued for the target geological disaster risk points. By using relatively high-precision sensors to monitor historical data, and through meteorological rainfall big data, satellite remote sensing image big data and geographic information big data, based on the geological disaster prediction model established through machine learning algorithms such as logistic regression, random forest and SVM support vector machine, the geological disaster incidence rate of each point in the whole region is calculated at a certain level of accuracy, and geological disaster early warning prediction and calculation are realized for the entire monitoring area, and early warning is issued for locations in the whole region where geological disasters may occur in the future. The results can be extended to the entire geological disaster monitoring area, which not only realizes the geological disaster early warning prediction and calculation for the entire monitoring area, but also greatly improves the accuracy and timeliness of geological disaster monitoring and early warning.

[0169] Combination of the above Figure 1-14 The geological disaster prediction method based on big data provided in the embodiment of the present application is introduced in detail. The geological disaster prediction device, electronic device and computer-readable storage medium based on big data provided in the embodiment of the present application will be introduced in conjunction with the accompanying drawings.

[0170] like Fig.15 As shown in the figure, this figure is a schematic diagram of a geological disaster prediction device based on big data provided by the present application, and the device includes:

[0171] The acquisition module 201 is used to acquire the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted;

[0172] The prediction module 202 is used to obtain the predicted geological disaster risk points in the area to be predicted based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model; the preset geological disaster prediction model is obtained by model training based on the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to the sample points in the area to be predicted; wherein the sample points are the sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as the pouring points and the sample non-disaster points outside the basin that meet the preset conditions;

[0173] The early warning module 203 is used to generate and output geological disaster risk early warning prompt information according to the predicted geological disaster risk points.

[0174] The geological disaster prediction device based on big data provided in the embodiment of the present application can correspond to executing the geological disaster prediction method based on big data described in the embodiment of the present application, and the above functions of each module of the device correspond to realizing Figure 1 For the sake of brevity, the corresponding process of the method shown will not be repeated here.

[0175] An embodiment of the present application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the method for predicting geological disasters based on big data as described in the above embodiment is implemented.

[0176] like Fig.16 As shown, the computer system 300 of the electronic device includes a CPU 301, which can perform various appropriate actions and processes according to the program stored in the ROM 302 or the program loaded from the storage part 308 to the RAM 303. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, the ROM 302 and the RAM 303 are connected to each other through a bus 304. The I / O interface 305 is also connected to the bus 304. Among them, CPU 301 represents a central processing unit, ROM 302 represents a read-only memory, RAM 303 represents a random access memory, and I / O represents input / output.

[0177] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube, a liquid crystal display, etc., and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed, so that a computer program read therefrom is installed into the storage section 308 as needed.

[0178] In particular, the process of the geological disaster prediction method based on big data described in the above embodiment can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains a program code for executing the geological disaster prediction method based on big data described in the above embodiment. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 309, and / or installed from the removable medium 311. When the computer program is executed by the CPU 301, the above functions defined in the present computer system 300 are executed.

[0179] An embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for predicting geological disasters based on big data as described in the above embodiment is implemented.

[0180] Specifically, the computer-readable storage medium may be included in the electronic device described in the above embodiment; or it may exist independently without being assembled into the electronic device. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the geological disaster prediction method based on big data as described in the above embodiment.

[0181] It should be noted that the computer-readable storage medium shown in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory, a read-only memory, an erasable programmable read-only memory, an optical fiber, a portable compact disk read-only memory, an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above.

[0182] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and its core ideas of this application. The above is only the preferred implementation method of this application. It should be pointed out that due to the limitations of textual expression and the objective existence of infinite specific structures, ordinary technicians in this technical field can make several improvements, modifications or changes without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes or combinations, or the direct application of the inventive concept and technical solution to other occasions without improvement, should be regarded as the scope of protection of this application.

Claims

1. A geological disaster prediction method based on big data, characterized in that: include: Obtain the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted; Based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model, the predicted geological disaster risk points in the area to be predicted are obtained; the preset geological disaster prediction model is obtained by model training based on the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to the sample points in the area to be predicted; wherein the sample points are sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as the pouring points, and sample non-disaster points outside the basin that meet the preset conditions; Generate and output geological disaster risk warning prompt information based on the predicted geological disaster risk points.

2. The method according to claim 1, characterized in that The method further comprises: Acquire historical data of geological disasters in the area to be predicted; the historical data of geological disasters at least includes historical disaster points and their historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data; Based on the historical disaster points and their historical geographic information data, sample disaster points and sample non-disaster points in the area to be predicted are obtained, and according to the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data of the sample disaster points and the sample non-disaster points, as well as the disaster results corresponding to the sample disaster points and the sample non-disaster points, model training is performed based on a preset original prediction model to obtain the preset geological disaster prediction model; the preset original prediction model includes an SVM prediction model and / or a random forest prediction model, or a logistic regression prediction model.

3. The method according to claim 2, characterized in that The method of obtaining sample disaster points and sample non-disaster points in the area to be predicted based on the historical disaster points and their historical geographic information data, and performing model training based on a preset original prediction model according to the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data of the sample disaster points and the sample non-disaster points, and the disaster results corresponding to the sample disaster points and the sample non-disaster points, to obtain the preset geological disaster prediction model includes: Based on the historical geographic information data of the historical disaster point, the watershed where the historical disaster point is located is determined with the historical disaster point as the pouring point, and all grids in the watershed where the historical disaster point is located are used as sample disaster points, and sample non-disaster points are selected in an area outside the watershed where the historical disaster point is located at a first preset ratio, and the sample disaster points and the sample non-disaster points are used as the sample points; The historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and corresponding disaster results of the sample points of the second preset ratio are used as a training set, and the historical remote sensing image data, historical meteorological rainfall data, historical geographic information data and corresponding disaster results of the remaining sample points are used as a test set; The training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model.

4. The method according to claim 3, characterized in that The preset original prediction model is a logistic regression prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model, including: Input the training set into the preset original prediction model: Perform model training; where y i , i=1,...n are geological disaster response variables, which obey the binomial distribution B(1, p i );p i is the geological disaster response variable y i The probability of being 1, β0, ... β n is the coefficient, x i1 ,...x in is the geological disaster impact variable, L is the likelihood function, threshold is the probability threshold, is the maximum likelihood estimation of the geological hazard response variable, is the geological disaster response variable y i The maximum likelihood estimate of the probability that is 1; Maximize the likelihood function and iteratively solve the coefficients β0, ...β n The maximum likelihood estimate of obtaining a first intermediate prediction model; The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the first intermediate prediction model to calculate the predicted disaster results of the sample points of the test set, and the TPR value and the FPR value are calculated according to the predicted disaster results and the disaster results of the sample points of the test set. Based on the TPR value and the FPR value, feature form transformation, feature selection and probability threshold optimization are performed until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

5. The method according to claim 3, characterized in that: The preset original prediction model is a random forest prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model, including: A plurality of sample points are randomly extracted with replacement from the training set, a plurality of features are randomly extracted without replacement from the historical remote sensing image data, the historical meteorological rainfall data, and the historical geographic information data, and a plurality of decision trees are established with the ntree parameter as the optimization parameter to form a random forest; The random forest partition is simulated multiple times to calculate the TPR value and the FPR value respectively, and the ntree parameter when the TPR value and the FPR value obtain the optimal value is used as the intermediate optimization parameter to obtain the second intermediate prediction model; The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the second intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the TPR value and the FPR value are calculated, and the intermediate optimization parameters are optimized until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

6. The method according to claim 3, characterized in that The preset original prediction model is an SVM prediction model; the training set is input into the preset original prediction model for model training, and the training structure is tested and optimized based on the test set until the prediction accuracy is greater than a preset threshold, thereby obtaining the preset geological disaster prediction model, including: The SVM prediction model is established by taking the historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the training set as features, taking the Gaussian kernel as the kernel function, and taking the misjudgment loss parameter and the Gaussian kernel parameter as optimization parameters; The optimization parameters are simulated multiple times to calculate TPR values ​​and FPR values ​​respectively, and based on the TPR values ​​and FPR values, the misjudgment loss parameter and Gaussian kernel parameter corresponding to when the TPR value obtains a maximum value are used as the intermediate loss parameter and the intermediate Gaussian kernel parameter to obtain a third intermediate prediction model; The historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data of the sample points of the test set are input into the third intermediate prediction model to calculate the predicted disaster results of the sample points of the test set. According to the predicted disaster results and the disaster results of the sample points of the test set, the TPR value and the FPR value are calculated, and the intermediate loss parameters and intermediate Gaussian kernel parameters are optimized until the prediction accuracy is greater than the preset threshold, so as to obtain the preset geological disaster prediction model.

7. The method according to claim 3, characterized in that The preset original prediction model includes a random forest prediction model and an SVM prediction model, and the preset geological disaster prediction model includes a first prediction sub-model established based on the random forest prediction model and a second prediction sub-model established based on the SVM prediction model; the acquisition of predicted geological disaster risk points in the area to be predicted based on the remote sensing image data, meteorological rainfall data, geographic information data and the preset geological disaster prediction model includes: A first risk point obtained according to the first prediction sub-model based on the remote sensing image data, the meteorological rainfall data, and the geographic information data; A second risk point obtained according to the second prediction sub-model based on the remote sensing image data, the meteorological rainfall data, and the geographic information data; The union of the first risk point and the second risk point is used as the predicted geological disaster risk point.

8. A geological disaster prediction device based on big data, characterized in that: include: The acquisition module is used to obtain the current remote sensing image data, meteorological rainfall data, and geographic information data of the area to be predicted; A prediction module, for obtaining predicted geological disaster risk points in the area to be predicted based on the remote sensing image data, meteorological rainfall data, geographic information data and a preset geological disaster prediction model; the preset geological disaster prediction model is obtained by model training based on historical remote sensing image data, historical meteorological rainfall data, and historical geographic information data corresponding to sample points in the area to be predicted; wherein the sample points are sample disaster points included in the basin determined by taking the historical disaster points in the area to be predicted as pouring points, and sample non-disaster points outside the basin that meet preset conditions; The early warning module is used to generate and output geological disaster risk early warning prompt information based on the predicted geological disaster risk points.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images