Multi-time scale landslide deformation prediction method and related device

Abstract

The invention provides a multi-time-scale landslide deformation prediction method and a related device, and relates to the technical field of deep learning, landslide deformation and geological disaster prediction. Comprising the steps of performing downsampling processing on original landslide deformation time sequence data observed from a target area to obtain a time sequence data set; decomposing the time sequence data set to obtain a trend information set and a period information set; supplementing part of trend information with a smaller time scale in the trend information set to obtain a supplemented trend information set; supplementing partial period information with a relatively large time scale in the period information set to obtain a supplemented period information set; fusing the trend information and the period information under the same time scale in the supplemented trend information set and the supplemented period information set to obtain a fused time sequence data set; and processing the fused time sequence data set by using a preset time sequence prediction model to obtain a landslide deformation prediction result of the target area. The scheme can improve the accuracy of the prediction result.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, specifically to the field of deep learning, landslide deformation, and geological disaster prediction technology, and in particular to a multi-time scale landslide deformation prediction method, device, electronic device, computer-readable storage medium, and computer program product. Background Art

[0002] Landslide geological disasters are extremely destructive, not only causing damage to village and town infrastructure, but also threatening the lives and property of residents. For example, when a reservoir is filled with water, hundreds of landslides are reactivated, putting them in an intermittent creep state. Under the influence of periodic changes in reservoir water levels and heavy rainfall, many landslides in the reservoir area exhibit a stepped deformation pattern, alternating between accelerated and decelerated deformation states.

[0003] Specifically, reservoir water level and its fluctuations may periodically change the hydromechanical boundaries of landslides, while surface runoff and groundwater recharge caused by precipitation promote the hydrodynamic changes of landslides. The combined effects of these two make the evolution mechanism of landslides more complicated.

[0004] Therefore, understanding and monitoring these dynamic changes are of great significance for the prevention and mitigation of landslide hazards. Summary of the invention

[0005] The embodiments of the present disclosure provide a multi-time-scale landslide deformation prediction method, device, electronic device, computer-readable storage medium, and computer program product.

[0006] In the first aspect, the embodiments of the present disclosure propose a multi-time-scale landslide deformation prediction method, including: downsampling the original landslide deformation time series data observed from the target area to obtain a time series data set composed of landslide deformation time series data of different sequence lengths; decomposing the time series data set to obtain a trend information set and a period information set; according to the characteristics of the trend information, supplementing part of the trend information with a smaller time scale in the trend information set to obtain a supplemented trend information set; according to the characteristics of the period information, supplementing part of the period information with a larger time scale in the period information set to obtain a supplemented period information set; fusing the trend information and period information at the same time scale in the supplemented trend information set and the supplemented period information set to obtain a fused time series data set; using a preset time series prediction model to process the fused time series data set to obtain a landslide deformation prediction result for the target area.

[0007] In a second aspect, an embodiment of the present disclosure proposes a multi-time-scale landslide deformation prediction device, comprising: a downsampling processing unit, configured to perform downsampling processing on the original landslide deformation time series data observed from the target area, and obtain a time series data set composed of landslide deformation time series data of different sequence lengths; a decomposition processing unit, configured to decompose the time series data set, and obtain a trend information set and a period information set; a trend information supplementing unit, configured to supplement part of the trend information with a smaller time scale in the trend information set according to the characteristics of the trend information, and obtain a supplemented trend information set; a period information supplementing unit, configured to supplement part of the period information with a larger time scale in the period information set according to the characteristics of the period information, and obtain a supplemented period information set; a fusion unit, configured to fuse the trend information and period information at the same time scale in the supplemented trend information set and the supplemented period information set, and obtain a fused time series data set; a prediction unit, configured to process the fused time series data set using a preset time series prediction model, and obtain a landslide deformation prediction result for the target area.

[0008] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that when the at least one processor executes, the multi-time-scale landslide deformation prediction method described in the first aspect can be implemented.

[0009] In a fourth aspect, an embodiment of the present disclosure provides a non-transitory computer-readable storage medium storing computer instructions, which are used to enable a computer to implement the multi-time-scale landslide deformation prediction method described in the first aspect when executed.

[0010] In a fifth aspect, an embodiment of the present disclosure provides a computer program product including a computer program, which, when executed by a processor, can implement the steps of the multi-time-scale landslide deformation prediction method as described in the first aspect.

[0011] The multi-time-scale landslide deformation prediction scheme provided by the present disclosure first obtains a landslide deformation time series data set containing different time scales by downsampling the original landslide deformation time series data, and then decomposes it into a trend part characterizing the overall trend of the landslide deformation and a period part reflecting the periodic change law, and then supplements the trend information or period information of the corresponding time scale by combining the different characteristics of the trend information and the period information, so as to obtain a more comprehensive and accurate trend information set and period information set, and then fuses the parts of the same time scale in the supplemented trend information set and the period information set to further highlight the period and trend information, and finally analyzes and processes the fused time series data set by using a time series prediction model, so as to obtain a more accurate landslide deformation prediction model.

[0012] That is, the multi-time scale landslide deformation prediction scheme provided by the present invention not only improves the accuracy and reliability of prediction by making full use of the landslide deformation time series information at different time scales, but also innovatively performs collaborative modeling of cycles and trends based on the characteristics of deformation cycles and trends, thereby improving the model's ability to capture landslide deformation data while also enhancing the accuracy of prediction results.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Other features, objects and advantages of the present disclosure will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0015] Figure 1 is an exemplary system architecture in which the present disclosure may be applied;

[0016] Figure 2 A flowchart of a multi-time scale landslide deformation prediction method provided by an embodiment of the present disclosure;

[0017] Figure 3 A flowchart of a method for constructing a time series data set provided in an embodiment of the present disclosure;

[0018] Figure 4 A flow chart of a trend information supplement method provided by an embodiment of the present disclosure;

[0019] Figure 5 A flowchart of a period information supplement method provided by an embodiment of the present disclosure;

[0020] Figure 6A flowchart of a method for fusing periodic information and trend information on the same time scale provided by an embodiment of the present disclosure;

[0021] Figure 7 A structural block diagram of a multi-time scale landslide deformation prediction device provided by an embodiment of the present disclosure;

[0022] Figure 8 A schematic diagram of the structure of an electronic device suitable for executing a multi-time-scale landslide deformation prediction method provided in an embodiment of the present disclosure;

[0023] Fig. 9 A schematic block diagram of an exemplary electronic device 900 provided according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0024] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description. It should be noted that the embodiments in the present disclosure and the features in the embodiments may be combined with each other without conflict.

[0025] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0026] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the multi-time-scale landslide deformation prediction method, apparatus, electronic device, and computer-readable storage medium of the present disclosure can be applied.

[0027] like Figure 1 As shown, the system architecture 100 may include information collection devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the information collection devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0028] The information collection devices 101, 102, and 103 are used to be set up in areas where landslide geological disasters may occur, and collect landslide deformation data in the area through various information collection means, and can send the collected landslide deformation data to the server 105 through the network 104, or receive control instructions sent by the server 105 through the network 104. Various applications for realizing information communication between the information collection devices 101, 102, 103 and the server 105 can be installed, such as information collection applications, landslide deformation prediction applications, landslide geological disaster prediction applications, etc.

[0029] The information collection devices 101, 102, 103 and the server 105 can be hardware or software. When the information collection devices 101, 102, 103 are hardware, they can be various electronic devices with information collection capabilities, including but not limited to sensors, cameras, scanners or radars, etc.; when the information collection devices 101, 102, 103 are software, they can be installed in the electronic devices listed above, which can be implemented as multiple software or software modules, or as a single software or software module, and no specific limitation is made here. When the server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server; when the server is software, it can be implemented as multiple software or software modules, or as a single software or software module, and no specific limitation is made here.

[0030] The server 105 can provide various services through various built-in applications. Taking the landslide geological disaster prediction application that can provide landslide deformation prediction service as an example, the server 105 can achieve the following effects when running the landslide geological disaster prediction application: first, the original landslide deformation time series data obtained by the information collection devices 101, 102, and 103 through continuous observation and collection of the target area is received through the network 104; then, the original landslide deformation time series data is downsampled to obtain a time series data set composed of landslide deformation time series data of different sequence lengths; then, the time series data set is decomposed to obtain a trend information set and periodic information set; next, according to the characteristics of trend information, some trend information with a smaller time scale in the trend information set is supplemented to obtain a supplemented trend information set; at the same time, according to the characteristics of periodic information, some periodic information with a larger time scale in the periodic information set is supplemented to obtain a supplemented periodic information set; then, the trend information and periodic information at the same time scale in the supplemented trend information set and the supplemented periodic information set are fused to obtain a fused time series data set; finally, the preset time series prediction model is used to process the fused time series data set to obtain the landslide deformation prediction result in the target area.

[0031] It should be noted that, in addition to being continuously obtained from the information collection devices 101, 102, and 103 through the network 104, the original landslide deformation time series data can also be pre-stored locally in the server 105 in various ways. Therefore, when the server 105 detects that these data have been stored locally (for example, starting to process the prediction tasks to be processed before), it can choose to directly obtain these data locally. In this case, the exemplary system architecture 100 may also not include the information collection devices 101, 102, 103 and the network 104.

[0032] Since landslide deformation prediction requires more computing resources and stronger computing power, the multi-time scale landslide deformation prediction method provided in the subsequent embodiments of the present disclosure is generally executed by a server 105 with stronger computing power and more computing resources. Correspondingly, the multi-time scale landslide deformation prediction device is generally also set in the server 105.

[0033] It should be understood that Figure 1 The number of information collection devices, networks and servers in the embodiment is only for illustration. Any number of information collection devices, networks and servers may be provided according to the implementation requirements.

[0034] Please refer to Figure 2 , Figure 2 A flowchart of a multi-time scale landslide deformation prediction method provided by an embodiment of the present disclosure, wherein process 200 includes the following steps: Step 201: downsampling the original landslide deformation time series data observed in the target area to obtain a time series data set consisting of landslide deformation time series data of different sequence lengths; This step is intended to be performed by the execution body of the multi-time scale landslide deformation prediction method (e.g. Figure 1 The server 105 shown in the figure performs downsampling processing on the original landslide deformation time series data of the target area, thereby obtaining landslide deformation time series data with a larger (or coarser) time granularity, and finally forming a time series data set consisting of landslide deformation time series data of at least two different time scales.

[0035] The target area refers to a geographical area where landslide geological disasters are likely to occur, such as a mountain area, some valleys or rock and soil layers near lakes and reservoirs, etc., and the original landslide deformation data is collected by information collection equipment (such as Figure 1The information collection devices 101, 102, and 103 in the data acquisition system 101, 102, and 103) are collected through the continuous observation process, and the observation unit corresponding to the original landslide deformation time series data can be the first time length, specifically, it can be 1 day, that is, the landslide deformation data observed in the target area every day is taken as one portion. If the observation is continued for 30 days, then the original landslide deformation time series data is a time series data with a length of 30. In addition, multiple downsampling can be performed continuously, and the number of downsampling and the downsampling rate can be determined or adjusted according to actual needs, and are not specifically limited here.

[0036] Downsampling is a method of reducing the number of samples in a signal or data set, which effectively reduces the amount of data and improves processing efficiency while retaining key trends and patterns. The use of downsampling technology to process time series data can be achieved mainly through the following steps: 1) Determine the downsampling rate: According to the analysis requirements, select an appropriate time interval, such as extracting a data point every minute, hour, day, every two days, every four days, or even every week; 2) Select the downsampling method: Average downsampling: Calculate the average value in each time period to obtain representative data for that period; Maximum / minimum downsampling: extract the maximum or minimum value in each time period, suitable for capturing extreme value changes; Random sampling: randomly select data points within a time period, suitable for irregular data; 3) Implement downsampling: Reorganize the data according to the selected method to generate a new time series.

[0037] Step 202: Decompose the time series data set to obtain a trend information set and a period information set; Based on step 201, this step aims to decompose the time series data set by the above-mentioned execution subject to obtain the trend part (i.e., trend information set) representing the overall cumulative deformation and the periodic part (i.e., periodic information set) representing the periodic change law.

[0038] Among them, separating the trend part from the time series data can be achieved through the sliding average method (using a sliding window to calculate the moving average of the data, smoothing the time series, and extracting the long-term trend), the addition / multiplication model (assuming that the data can be decomposed into trend, seasonality, and residual, the addition model or the multiplication model can be used for decomposition) and linear regression (establishing a linear regression model, fitting the trend part, and extracting the trend component); extracting the periodic part from the time series can be achieved through Fourier transform (converting data from the time domain to the frequency domain, identifying the frequency components, and extracting the periodic change part), wavelet transform (capable of capturing periodic components of different frequencies, suitable for processing non-stationary time series), seasonal decomposition (decomposing the time series into trend, seasonality, and residual parts) and other methods.

[0039] In order to separate the periodic variation law and the trend of the overall cumulative deformation in the landslide deformation time series data, an implementation method including but not limited to the following may also be used: The time series data set is decomposed using a Fast Fourier Transform (FFT) algorithm to obtain a trend information set as a first decomposition result and a period information set as a second decomposition result.

[0040] This can be achieved through the following steps: 1) Apply Fast Fourier Transform Calculate FFT: Apply FFT algorithm to the landslide deformation time series data to convert the time domain signal into the frequency domain signal; Spectral analysis: Analyze the obtained frequency domain data to identify significant frequency components, especially frequencies that change periodically.

[0041] 2) Extracting periodic components Identify the main frequencies: By observing the amplitude spectrum in the FFT result, identify the main periodic components. This usually appears as a spike in the spectrum, representing the periodic changes in the signal.

[0042] Constructing a periodic model: Using the identified frequencies, reconstruct the periodic components. These frequency components can be converted back to the time domain using an inverse FFT (IFFT).

[0043] 3) Extract trend components Filtering: Compare the original signal with the extracted periodic component. Usually, the trend component can be extracted by a low-pass filter, that is, the periodic component is removed from the original signal; Calculate the trend: The trend part can be obtained from the original signal through simple difference, moving average and other methods, or by fitting the trend through regression analysis.

[0044] Furthermore, in order to improve the separation effect, the landslide deformation time series data can be preprocessed before step 1), for example, filtering techniques (such as moving average, low-pass filtering, etc.) can be used to remove noise in the time series to ensure the accuracy of subsequent analysis, and standardization: standardize the data to eliminate magnitude differences for subsequent processing.

[0045] Step 203: supplementing part of the trend information set with a smaller time scale according to the characteristics of the trend information, to obtain a supplemented trend information set; On the basis of step 202, this step aims to have the above-mentioned execution subject supplement the trend information with a smaller time scale in the trend information set according to the characteristics of the trend information, which can be understood as: using the "macro" trend to supplement the "micro" trend (i.e., the coarse-grained trend supplements the fine-grained trend), thereby obtaining the supplemented trend information set. That is, the supplemented trend information set is based on the trend information set before supplementation, and mainly supplements the trend information with a smaller time scale. It can be considered that the trend information with a larger time scale does not need to be supplemented, or the supplementation is less.

[0046] Step 204: supplementing part of the periodic information with a larger time scale in the periodic information set according to the characteristics of the periodic information to obtain a supplemented periodic information set; On the basis of step 202, this step aims to have the above-mentioned execution subject supplement the partial periodic information with a larger time scale in the periodic information set according to the characteristics of the periodic information, that is, to use the deformation law reflected by the small period to supplement the periodic information of the large period, so as to obtain the supplemented periodic information set. That is, the supplemented periodic information set is based on the periodic information set before supplementation, and mainly supplements the partial periodic information with a larger time scale. It can be considered that the partial periodic information with a smaller time scale does not need to be supplemented, or the supplementation is less.

[0047] It should be noted that there is no causal or dependency relationship between step 203 and step 204, and they can be executed in parallel. This embodiment only expresses the two steps in a serial manner for the convenience of expression, but in fact, the two steps can be executed in parallel at the same time.

[0048] Step 205: fusing the trend information and the period information at the same time scale in the supplemented trend information set and the supplemented period information set to obtain a fused time series data set; On the basis of step 203 and step 204, this step aims to fuse the trend information and periodic information at the same time scale in the supplemented trend information set and the supplemented periodic information set by the above-mentioned execution subject, and then obtain the fused time series data at each time scale, and summarize them to obtain the fused time series data set.

[0049] Step 206: Use a preset time series prediction model to process the fused time series data set to obtain a landslide deformation prediction result for the target area.

[0050] Based on step 205, this step aims to obtain the landslide deformation prediction result of the target area by processing the fused time series data set by the above-mentioned execution subject using the preset time series prediction model. Specifically, the time series prediction model can be a Transformer (literally translated as: converter) model of a specific multi-head self-attention mechanism.

[0051] The multi-time-scale landslide deformation prediction method provided by the embodiment of the present disclosure first obtains a landslide deformation time series data set containing different time scales by downsampling the original landslide deformation time series data, and then decomposes it into a trend part characterizing the overall trend of the landslide deformation and a period part reflecting the periodic change law, and then supplements the trend information or period information of the corresponding time scale by combining the different characteristics of the trend information and the period information, so as to obtain a more comprehensive and accurate trend information set and period information set, and then fuses the parts of the same time scale in the supplemented trend information set and the period information set to further highlight the period and trend information, and finally analyzes and processes the fused time series data set by using a time series prediction model, so as to obtain a more accurate landslide deformation prediction model.

[0052] That is, the multi-time scale landslide deformation prediction scheme provided by the present invention not only improves the accuracy and reliability of prediction by making full use of the landslide deformation time series information at different time scales, but also innovatively performs collaborative modeling of cycles and trends based on the characteristics of deformation cycles and trends, thereby improving the model's ability to capture landslide deformation data while also enhancing the accuracy of prediction results.

[0053] In order to deepen the understanding of how to obtain a time series data set by downsampling, this embodiment also Figure 3 A method for constructing a time series data set is provided, wherein process 300 includes the following steps: Step 301: obtaining original landslide deformation time series data obtained by continuously observing a target area with a first preset time length as an observation unit; Among them, the first preset time length can be 1 day, and the continuous observation time can be 120 days. Then the original landslide deformation time series data at this time is a time series with a length of 120.

[0054] The landslide deformation data for each observation time may only include data obtained by a single information collection device observing the target area, or may be a summary or integration of data collected by multiple information collection devices using the same or different observation methods.

[0055] Step 302: downsampling the original landslide deformation time series data by using an average pooling downsampling technique to obtain downsampled landslide deformation time series data; On the basis of step 301, this step aims to perform downsampling processing on the original landslide deformation time series data by the above-mentioned execution subject through the average pooling downsampling technology to obtain the downsampled landslide deformation time series data.

[0056] The number of downsampling processes may be one or more times. Taking two consecutive downsampling processes as an example, it may be specifically performed in the following steps: Firstly, the original landslide deformation time series data is downsampled once by using the average pooling downsampling technique to obtain the downsampled time series data. Then, the time series data after the first downsampling is downsampled twice through the average pooling downsampling technology to obtain the time series data after the second downsampling.

[0057] Generally, the same pooling window can be used for the first downsampling process and the second downsampling process. In special cases, it is not ruled out that the two downsampling processes use different pooling windows.

[0058] Taking the pooling window of 2 as an example (which means reducing the original data per day to data every two days), the time series data obtained after the first downsampling process is a time series with a length of 60, and the time series data obtained after the second downsampling process is a time series with a length of 30.

[0059] Step 303: construct a time series data set based on the original landslide deformation time series data and the downsampled landslide deformation time series data.

[0060] Correspondingly, if two consecutive downsampling processes are performed, then this step will adjust the adaptability to: construct a time series data set based on the original landslide deformation time series data, the time series data after the first downsampling, and the time series data after the second downsampling.

[0061] In order to deepen the understanding of how to supplement part of the trend information in the trend information set according to the characteristics of the trend information, this embodiment also Figure 4 A trend information supplement method is shown, wherein the process 400 includes the following steps: Step 401: determining the part of trend information with the largest time scale in the trend information set as coarse-grained trend information; Trend information is the trend information used to characterize the overall cumulative deformation in the landslide deformation process. Therefore, this step aims to determine the trend information with the largest time scale in the trend information set as coarse-grained trend information, namely "macro" trend information, by the above-mentioned execution subject.

[0062] Step 402: Mapping the coarse-grained trend information into new fine-grained trend information using a fully connected layer and a Gaussian error linear activation function; On the basis of step 401, this step aims to map the coarse-grained trend information into new fine-grained trend information by the above-mentioned execution subject using a fully connected layer and a Gaussian error linear activation function.

[0063] The fully connected layer plays an important role in neural networks and is mainly used for feature combination, dimensionality reduction, classification, and regression tasks. It increases the nonlinear ability of the network by combining it with the activation function, allowing deep learning models to process complex data more effectively. Although the fully connected layer is very effective in many applications, due to its large number of parameters and high computational complexity, it is usually necessary to combine it with other types of layers (such as convolutional layers) on large-scale data sets to improve efficiency and effectiveness.

[0064] The Gaussian Error Linear Unit (GELU) activation function is a nonlinear activation function. Its main function is to introduce nonlinear characteristics while maintaining certain linear characteristics, so that the model can better capture complex patterns. It has the characteristics and effects of smoothness (smooth transition at zero point, which helps to reduce instability during model training), based on Gaussian distribution (its shape is similar to the cumulative distribution function of Gaussian distribution, which makes it have a certain guiding effect on the randomness of input in forward propagation), and improved performance (in a variety of tasks, GELU usually brings better model performance and convergence speed compared with ReLU and other activation functions).

[0065] By combining the fully connected layer with the GELU activation function, it is possible to handle complex tasks more effectively and enhance its performance in practical applications.

[0066] Step 403: supplement the original fine-grained trend information in the trend information set with the new fine-grained trend information to obtain supplemented fine-grained trend information; Based on step 402, this step aims to use the new fine-grained trend information to supplement the original fine-grained trend information in the trend information set by the above-mentioned execution subject to obtain supplemented fine-grained trend information. The fine-grained trend information is the partial trend information with the smallest time scale, and the new fine-grained trend information has the same time scale as the fine-grained trend information.

[0067] Step 404: construct a supplemented trend information set based on the coarse-grained trend information and the supplemented fine-grained trend information.

[0068] Based on step 403, this step aims to construct a supplemented trend information set based on the coarse-grained trend information and the supplemented fine-grained trend information. Assuming that there is also intermediate granularity trend information between the coarse-grained trend information and the fine-grained trend information, in this embodiment, it can be inherited to the supplemented trend information set without modification, and in some other embodiments, it can also be supplemented less than the fine-grained trend information.

[0069] and Figure 4 Similar to the embodiment shown, this embodiment is Figure 5 A supporting information supplement scheme is provided for periodic information, and its process 500 includes the following steps: Step 501: determining the part of periodic information with the smallest time scale in the periodic information set as fine-grained periodic information; The periodic information is used to characterize the periodic inertia in the landslide deformation process. Therefore, this step aims to determine the periodic information with the smallest time scale in the periodic information set as fine-grained trend information, that is, small periodic inertia information, by the above-mentioned execution subject.

[0070] Step 502: Map the fine-grained periodic information to new coarse-grained periodic information using a fully connected layer and a Gaussian error linear activation function; The technology used in this step can be found in the relevant description of step 402, and will not be repeated here.

[0071] Step 503: supplement the original coarse-grained periodic information in the periodic information set with the new coarse-grained periodic information to obtain supplemented coarse-grained periodic information; On the basis of step 502, this step aims to use the new coarse-grained periodic information to supplement the original coarse-grained periodic information in the periodic information set by the above-mentioned execution subject to obtain supplemented coarse-grained periodic information. The coarse-grained periodic information is the partial periodic information with the largest time scale, and the new coarse-grained periodic information has the same time scale as the coarse-grained periodic information.

[0072] Step 504: construct a supplemented periodic information set based on the fine-grained periodic information and the supplemented coarse-grained periodic information.

[0073] On the basis of step 503, this step aims to construct a supplemented periodic information set based on the fine-grained periodic information and the supplemented coarse-grained periodic information. Assuming that there is intermediate periodic information between the coarse-grained periodic information and the fine-grained periodic information, in this embodiment, it can be inherited to the supplemented periodic information set without modification, and in some other embodiments, it can also be supplemented less than the coarse-grained periodic information.

[0074] In order to deepen the understanding of how to fuse the trend information and cycle information under the same time scale, this embodiment also uses Figure 6 A specific implementation is shown, and the process 600 includes the following steps: Step 601: determining trend information and cycle information at the same time scale in the supplemented trend information set and the supplemented cycle information set; Step 602: using a feedforward neural network to interact with trend information and period information at each same time scale to obtain interactive time series data; This step aims to allow the above-mentioned execution subject to interact with the cycle and trend information of each time scale fusion through a feedforward neural network, and add it to the original landslide deformation time series data to achieve information extraction and information interaction of the original observation sequence. The feedforward neural network can contain two fully connected layers, and the activation function uses GELU.

[0075] Step 603: Based on the interacted time series data at different time scales, a fused time series data set is constructed.

[0076] This embodiment provides a method for using a feedforward neural network to interactively fuse period information and trend information through steps 601 to 603 to improve the fusion effect.

[0077] On the basis of any of the above embodiments, after obtaining the landslide deformation prediction result, the predicted probability of landslide disasters occurring in the target area in different time periods can be further determined based on the landslide deformation prediction result. For example, the corresponding disaster occurrence probability can be predicted 2 hours, 8 hours, 12 hours, 24 hours, 3 days, 5 days or even longer after the last observation time point in the time series data. Disaster prevention measures can be further determined based on the predicted probability, such as reinforcement, early blasting or evacuation of people within the surrounding affected area, etc.

[0078] To deepen understanding, this disclosure also provides a complete and specific implementation solution in combination with a specific application scenario: The network model architecture specifically used in this embodiment is as follows: Figure 7 As shown, it is mainly divided into three parts: The first part is to downsample the landslide deformation data and extract deformation data of multiple time scales; The second part is to decompose and fuse deformation data at multiple time scales; The third part inputs the deformation data of multiple time scales into the Transformer module for the final landslide prediction.

[0079] The method proposed in this embodiment takes the landslide in the reservoir area of ​​Reservoir A as an example. It is assumed that the data set used is collected from the field landslide monitoring station. The collection unit is day, which includes multiple GNSS (Global Navigation Satellite System) surface displacement measurement points, 2 deep displacement measurement holes and 1 meteorological station. The data set consists of two covariates (precipitation and reservoir water level) and one variable to be predicted (landslide displacement), that is, historical precipitation, historical reservoir water level and historical landslide displacement are used as features, and the current displacement of the landslide is used as a label. The specific implementation steps are as follows: Step 1: Preprocess the acquired data and divide the landslide deformation data set into training set, validation set and test set in a ratio of 8:1:1. Then standardize the training data and apply it to the validation set and test set. Finally, divide the sliding window and set the sliding window size to 120.

[0080] Step 2: Build the model and train it. The specific process is as follows: 1) Downsampling of landslide deformation time series Landslide deformation data usually show various time scale characteristics, and each time scale has different effects on the effect of landslide prediction. In order to balance the effects of different time scales on landslide deformation prediction, this step first uses downsampling technology to extract multiple time scale sequences of the original deformation data.

[0081] That is, first, input the original deformation time series data, recorded as , which contains the most detailed time changes with days as the observation unit, and then uses the average pooling downsampling technique, setting the pooling window to 2, indicating that every 2 days of data points are used as a window, and the number of downsampling layers 𝑀 is taken as 2 times, and multiple downsampling is performed. Each downsampling reduces the length of the time series to half of the original, and the sequence set is obtained , where is the original time series data, containing the finest-grained temporal changes, providing richer temporal information and details, is The time series obtained after downsampling based on is of length , is The coarsest-grained time series obtained on the basis of , providing the model with long-term time changes, that is, obtaining time series with time scales of 120, 60, and 30 respectively, and saving the results as a list, which contains the landslide deformation data of three time scales.

[0082] 2) Decomposition of landslide deformation sequences at multiple time scales In the landslide displacement time series analysis, it is necessary to identify and separate the trend and period parts in the data. Specifically, the fast Fourier transform algorithm is used to decompose the period and trend of multiple scales respectively, and finally the results of the multiple scale decomposition are saved in the period list and trend list, that is, each list contains three time scale information.

[0083] .3) Fusion of landslide deformation sequences at multiple time scales.

[0084] .a) For period (periodic information), the fine-grained period information is used as the supplementary information of the coarse-grained period, and the fine-grained scale is mapped to the coarse-grained scale through the fully connected layer.

[0085] ;in, For fine-grained periodic information, the module realizes the mapping from fine-grained to coarse-grained. It contains two fully connected layers and a GELU activation function. The input dimension and output dimension of the fully connected layer depend on the original sequence length and the downsampling window size. and the number of downsampling layers M, where the downsampling window size is 2, and the number of downsampling layers M is 2. The final output is a list containing three elements, each of which represents the periodic fusion information of a time scale. The formulas for the input dimension and output dimension of the fully connected layer are as follows: ; .

[0086] b) For trend, macro trend information provides the model with long-term deformation patterns. This macro information is integrated into fine-grained trend information to help the model capture the overall changes in trends, that is, coarse-grained information is used as supplementary information to fine-grained information.

[0087] .

[0088] same, The module realizes coarse-grained to fine-grained mapping, which is achieved through two fully connected layers and a GELU activation function. Downsampling window in module The size is set to 2, and the number of downsampling layers M is 2. The final output is a list containing three elements, each of which represents the trend fusion information of a time scale. The formulas for the input dimension and output dimension of the fully connected layer are as follows: ; , c) Then, the cycle and trend information of each time scale fusion are interacted through the feedforward neural network and added to the original sequence to achieve information extraction and information interaction of the original observation sequence. The feedforward neural network contains two fully connected layers, and the activation function uses GELU, and finally obtains , Indicates Layer The shape is (B, T, N) where B is the batch size, T is the time series length, and N is the feature dimension.

[0089] .

[0090] 4) Transformer module.

[0091] a) The shape is changed to (B, N, T) through the inversion operation, and then the Embedding operation is performed. Here, the complete time series of each feature is encoded into a token to obtain a comprehensive representation of each feature. , where N is the number of features and the D mapping dimension is set to 128; b) Set up the projection layers (query, key, and value) and output projection layers required by the multi-head attention layer, with the input embedding dimension set to 128 and the number of attention heads set to 8; c) Stack the Transformer modules of the multi-head attention mechanism, layer normalization, and feed-forward neural network multiple times, setting the number of Transformer layers to 3, with 128 hidden units in each layer; d) Repeat the above steps until all scales are traversed and the prediction results of all scales are obtained. Finally, the prediction results of all scales are concatenated on a new dimension, all elements on this new dimension are summed, and the prediction results are output. Step 3: Model validation. Set the training to 20 epochs, validate the model after each epoch, and select the model with the smallest loss on the validation set as the final model for subsequent model testing.

[0092] Step 4: Model testing. Use the pre-divided test set to test the trained model. Compare the model's prediction results with the true observations, and use the mean square error (MSE), mean absolute error (MAE), and determination coefficient (R²) of the three regression task evaluation indicators to evaluate the model's prediction performance.

[0093] Further references Figure 8 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a multi-time scale landslide deformation prediction device. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0094] like Figure 8 As shown, the multi-time-scale landslide deformation prediction device 800 of this embodiment may include: a downsampling processing unit 801, a decomposition processing unit 802, a trend information supplementing unit 803, a period information supplementing unit 804, a fusion unit 805, and a prediction unit 806. Among them, the downsampling processing unit 801 is configured to perform downsampling processing on the original landslide deformation time series data observed from the target area to obtain a time series data set composed of landslide deformation time series data of different sequence lengths; the decomposition processing unit 802 is configured to decompose the time series data set to obtain a trend information set and a period information set; the trend information supplementation unit 803 is configured to supplement part of the trend information with a smaller time scale in the trend information set according to the characteristics of the trend information to obtain a supplemented trend information set; the period information supplementation unit 804 is configured to supplement part of the period information with a larger time scale in the period information set according to the characteristics of the period information to obtain a supplemented period information set; the fusion unit 805 is configured to fuse the supplemented trend information set and the trend information and period information at the same time scale in the supplemented period information set to obtain a fused time series data set; the prediction unit 806 is configured to process the fused time series data set using a preset time series prediction model to obtain a landslide deformation prediction result of the target area.

[0095] In this embodiment, in the multi-time scale landslide deformation prediction device 800, the specific processing of the downsampling processing unit 801, the decomposition processing unit 802, the trend information supplementing unit 803, the period information supplementing unit 804, the fusion unit 805, and the prediction unit 806 and the technical effects thereof can be referred to respectively. Figure 2 The relevant descriptions of steps 201 - 206 in the corresponding embodiment are not repeated here.

[0096] In some other implementations of this embodiment, the multi-time scale landslide deformation prediction device 800 may further include: An original landslide deformation time series data acquisition unit is configured to acquire original landslide deformation time series data obtained by continuously observing the target area with a first preset time length as an observation unit; Correspondingly, the downsampling processing unit 801 may include: The average pooling downsampling subunit is configured to downsample the original landslide deformation time series data by using the average pooling downsampling technique to obtain downsampled landslide deformation time series data; The time series data combination construction subunit is configured to construct the time series data set based on the original landslide deformation time series data and the downsampled landslide deformation time series data.

[0097] In some other implementations of this embodiment, the average pooling downsampling subunit may be further configured as follows: Performing a downsampling process on the original landslide deformation time series data by using the average pooling downsampling technology to obtain time series data after a downsampling process; Performing a secondary downsampling process on the once-downsampled time series data by using the average pooling downsampling technology to obtain the twice-downsampled time series data; Correspondingly, the time series data combination construction subunit may be further configured as follows: The time series data set is constructed based on the original landslide deformation time series data, the time series data after the first downsampling, and the time series data after the second downsampling.

[0098] In some other implementations of this embodiment, the primary downsampling process and the secondary downsampling process use the same pooling window.

[0099] In some other implementations of this embodiment, the decomposition processing unit 802 may be further configured to: The time series data set is decomposed by using a fast Fourier transform algorithm to obtain a trend information set as a first decomposition result and a period information set as a second decomposition result.

[0100] In some other implementations of this embodiment, the trend information supplementing unit 803 may be further configured to: Determine the part of trend information with the largest time scale in the trend information set as coarse-grained trend information; Mapping the coarse-grained trend information into new fine-grained trend information using a fully connected layer and a Gaussian error linear activation function; The new fine-grained trend information is used to supplement the original fine-grained trend information in the trend information set to obtain supplemented fine-grained trend information; wherein the fine-grained trend information is part of the trend information with the smallest time scale, and the new fine-grained trend information and the fine-grained trend information have the same time scale; Based on the coarse-grained trend information and the supplemented fine-grained trend information, the supplemented trend information set is constructed.

[0101] In some other implementations of this embodiment, the period information supplementing unit 804 may be further configured to: Determine the part of periodic information with the smallest time scale in the periodic information set as fine-grained periodic information; Mapping the fine-grained periodic information into new coarse-grained periodic information using a fully connected layer and a Gaussian error linear activation function; The new coarse-grained periodic information is used to supplement the original coarse-grained periodic information in the periodic information set to obtain supplemented coarse-grained periodic information; wherein the coarse-grained periodic information is the partial periodic information with the largest time scale, and the new coarse-grained periodic information and the coarse-grained periodic information have the same time scale; The supplemented periodic information set is constructed based on the fine-grained periodic information and the supplemented coarse-grained periodic information.

[0102] In some other implementations of this embodiment, the fusion unit 805 may be further configured to: Determine the trend information and the period information in the same time scale in the supplemented trend information set and the supplemented period information set; Using a feedforward neural network to interact with each trend information and period information under the same time scale to obtain interactive time series data; Based on the interacted time series data of different time scales, the fused time series data set is constructed.

[0103] In some other implementations of this embodiment, the multi-time scale landslide deformation prediction device 800 may further include: A prediction probability determination unit is configured to determine the prediction probability of landslide disasters occurring in the target area in different time periods based on the landslide deformation prediction result; The disaster prevention test determination unit is configured to determine disaster prevention measures based on the predicted probability.

[0104] This embodiment exists as a device embodiment corresponding to the above method embodiment. The multi-time scale landslide deformation prediction device provided by this embodiment first obtains a landslide deformation time series data set containing different time scales by downsampling the original landslide deformation time series data, and then decomposes it into a trend part that characterizes the overall trend of the landslide deformation and a period part that reflects the periodic change law. Then, the trend information or period information of the corresponding time scale is supplemented by combining the different characteristics of the trend information and the period information, so as to obtain a more comprehensive and accurate trend information set and period information set. Next, the same time scale parts of the supplemented trend information set and the period information set are fused to further highlight the period and trend information. Finally, the fused time series data set is analyzed and processed by using a time series prediction model, so as to obtain a more accurate landslide deformation prediction model. The device not only improves the accuracy and reliability of the prediction by making full use of the landslide deformation time series information at different time scales, but also innovatively performs collaborative modeling of the period and trend based on the characteristics of the deformation period and trend, which improves the model's ability to capture landslide deformation data while also enhancing the accuracy of the prediction results.

[0105] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the multi-time-scale landslide deformation prediction method described in any of the above embodiments can be implemented when the at least one processor executes.

[0106] According to an embodiment of the present disclosure, the present disclosure further provides a readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to implement the multi-time scale landslide deformation prediction method described in any of the above embodiments when executed.

[0107] According to an embodiment of the present disclosure, the present disclosure further provides a computer program product, which, when executed by a processor, can implement the multi-time-scale landslide deformation prediction method described in any of the above embodiments.

[0108] Fig. 9A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0109] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0110] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0111] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as the multi-time scale landslide deformation prediction method. For example, in some embodiments, the multi-time scale landslide deformation prediction method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the multi-time scale landslide deformation prediction method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the multi-time-scale landslide deformation prediction method in any other appropriate manner (eg, by means of firmware).

[0112] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0113] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0114] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0115] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0116] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0117] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client-server relationship is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and virtual private server (VPS) services.

[0118] According to the technical solution of the embodiment of the present disclosure, firstly, the original landslide deformation time series data is downsampled to obtain a landslide deformation time series data set containing different time scales, and then it is decomposed into a trend part that characterizes the overall trend of the landslide deformation and a period part that reflects the periodic change law, and then the trend information or period information of the corresponding time scale is supplemented by combining the different characteristics of the trend information and the period information, so as to obtain a more comprehensive and accurate trend information set and period information set, and then the parts of the same time scale in the supplemented trend information set and the period information set are fused to further highlight the period and trend information, and finally the fused time series data set is analyzed and processed by using a time series prediction model, so as to obtain a more accurate landslide deformation prediction model. That is, the scheme not only improves the accuracy and reliability of the prediction by making full use of the landslide deformation time series information at different time scales, but also innovatively performs collaborative modeling of the period and trend based on the characteristics of the deformation period and trend, and improves the model's ability to capture landslide deformation data while also enhancing the accuracy of the prediction results.

[0119] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0120] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A multi-time scale landslide deformation prediction method, characterized in that: include: The original landslide deformation time series data observed in the target area are downsampled to obtain a time series data set consisting of landslide deformation time series data of different sequence lengths; Decomposing the time series data set to obtain a trend information set and a period information set; According to the characteristics of the trend information, part of the trend information with a smaller time scale in the trend information set is supplemented to obtain a supplemented trend information set; According to the characteristics of the periodic information, part of the periodic information with a larger time scale in the periodic information set is supplemented to obtain a supplemented periodic information set; Fusion of the trend information and the period information at the same time scale in the supplemented trend information set and the supplemented period information set to obtain a fused time series data set; The fused time series data set is processed using a preset time series prediction model to obtain a landslide deformation prediction result in the target area.

2. The method according to claim 1, characterized in that Also includes: Acquire original landslide deformation time series data obtained by continuously observing the target area with a first preset time length as an observation unit; Correspondingly, the original landslide deformation time series data observed from the target area is downsampled to obtain a time series data set consisting of landslide deformation time series data of different sequence lengths, including: Downsampling the original landslide deformation time series data by using an average pooling downsampling technique to obtain downsampled landslide deformation time series data; The time series data set is constructed based on the original landslide deformation time series data and the downsampled landslide deformation time series data.

3. The method according to claim 2, characterized in that The downsampling process of the original landslide deformation time series data by the average pooling downsampling technology is performed to obtain the downsampled landslide deformation time series data, including: Performing a downsampling process on the original landslide deformation time series data by using the average pooling downsampling technology to obtain time series data after a downsampling process; Performing a secondary downsampling process on the once-downsampled time series data by using the average pooling downsampling technology to obtain the twice-downsampled time series data; Correspondingly, the time series data set is constructed based on the original landslide deformation time series data and the downsampled landslide deformation time series data, including: The time series data set is constructed based on the original landslide deformation time series data, the time series data after the first downsampling, and the time series data after the second downsampling.

4. The method according to claim 3, characterized in that The primary downsampling process and the secondary downsampling process use the same pooling window.

5. The method according to claim 1, characterized in that Decomposing the time series data set to obtain a trend information set and a period information set includes: The time series data set is decomposed by using a fast Fourier transform algorithm to obtain a trend information set as a first decomposition result and a period information set as a second decomposition result.

6. The method according to claim 1, characterized in that According to the characteristics of the trend information, part of the trend information with a smaller time scale in the trend information set is supplemented to obtain a supplemented trend information set, including: Determine the part of trend information with the largest time scale in the trend information set as coarse-grained trend information; Mapping the coarse-grained trend information into new fine-grained trend information using a fully connected layer and a Gaussian error linear activation function; The new fine-grained trend information is used to supplement the original fine-grained trend information in the trend information set to obtain supplemented fine-grained trend information; wherein the fine-grained trend information is part of the trend information with the smallest time scale, and the new fine-grained trend information and the fine-grained trend information have the same time scale; Based on the coarse-grained trend information and the supplemented fine-grained trend information, the supplemented trend information set is constructed.

7. A multi-time scale landslide deformation prediction device, characterized in that: include: A downsampling processing unit is configured to perform downsampling processing on the original landslide deformation time series data observed from the target area to obtain a time series data set consisting of landslide deformation time series data of different sequence lengths; A decomposition processing unit is configured to decompose the time series data set to obtain a trend information set and a period information set; A trend information supplementing unit is configured to supplement part of the trend information set with a smaller time scale according to the characteristics of the trend information, so as to obtain a supplemented trend information set; A period information supplementing unit is configured to supplement part of the period information set with a larger time scale according to the characteristics of the period information, so as to obtain a supplemented period information set; a fusion unit configured to fuse the trend information and the period information at the same time scale in the supplemented trend information set and the supplemented period information set to obtain a fused time series data set; The prediction unit is configured to process the fused time series data set using a preset time series prediction model to obtain a landslide deformation prediction result of the target area.

8. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the multi-time-scale landslide deformation prediction method according to any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to execute the multi-time-scale landslide deformation prediction method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the steps of the multi-time scale landslide deformation prediction method according to any one of claims 1 to 6 are implemented.

Images