A slope multi-dimensional early warning method based on an LSTM neural network and electronic equipment

Abstract

The application discloses a kind of based on the multi-dimensional early warning method of slope of LSTM neural network, it is related to geotechnical engineering technical field, the application includes collection data, data is preprocessed, forms time series dataset;LSTM neural network model is constructed and trained, with time series dataset as input, output each original probability vector of early warning level;Real-time calculation surface displacement and internal displacement rate and acceleration characteristics, construct deformation evolution stage discriminant factor, generate adaptive weight vector according to discriminant factor;Original probability vector and current normalized displacement value output by LSTM model are weighted and integrated using adaptive weight vector, obtain comprehensive risk score;Comprehensive risk score is compared with preset multilevel threshold, trigger corresponding level early warning and publish information.The application extracts multi-source time series characteristics by LSTM neural network, and adaptively generates weight in combination with deformation evolution stage identification mechanism, can continuously adjust the contribution of each parameter according to freeze-thaw evolution process.

Description

Technical Field

[0001] This invention belongs to the field of geotechnical engineering technology, and in particular relates to a multi-dimensional early warning method and electronic device for slopes based on LSTM neural network. Background Technology

[0002] Seasonal freeze-thaw zones refer to specific geographical areas where the surface soil freezes in winter and thaws in spring, exhibiting a distinct cyclical freeze-thaw process. A freeze-thaw cycle refers to the physical process by which pore water or fissure water within the soil undergoes a water-to-ice phase transition due to repeated alternating changes in ambient temperature, accompanied by frost heave and thaw settlement deformation of the soil volume. In each freeze-thaw cycle, the physical and mechanical properties, hydrogeological conditions, and internal stress field of the slope soil undergo irreversible, progressive deterioration. The slope deformation evolution stages are a technical classification method based on the characteristics of the displacement-time curve, dividing the entire process of slope instability into different development stages such as initial deformation, constant-rate creep, and accelerated failure, used to characterize the current stability state of the slope. Accurately identifying the slope deformation evolution stages under the influence of freeze-thaw cycles is a crucial prerequisite for judging its stability and issuing timely warnings.

[0003] Currently, various technical solutions have been developed in the industry for automated early warning of slope stability. The single-parameter threshold method is widely used, which sets fixed thresholds for parameters such as rainfall and surface displacement, triggering an alarm when real-time monitoring values ​​exceed these limits. While this method is logically simple, it fails to comprehensively consider the characteristics of slopes in freeze-thaw zones affected by the coupled effects of multiple factors such as temperature, moisture, and deformation. After multiple freeze-thaw cycles, the soil structure gradually deteriorates, and even small rainfall or displacement rates may indicate high risk, which fixed thresholds cannot reflect. The empirical formula method assigns fixed weights to parameters such as rainfall and displacement, calculating stability coefficients using predetermined formulas. However, the weighting is often based on historical experience in specific regions, limiting its applicability to slopes in freeze-thaw zones with varying geological conditions and climatic characteristics. Another method uses simple statistical analysis, defining fluctuation ranges by calculating the mean and variance of monitoring data, triggering an alarm when values ​​exceed these ranges. However, its depth of processing time-series data is limited, making it difficult to uncover the evolution and intrinsic correlations of parameters such as displacement and moisture content during the freeze-thaw process.

[0004] In addition to the conventional methods mentioned above, existing technologies have explored the introduction of neural networks to process multi-parameter time series data and combined them with weight adjustment mechanisms to improve early warning effectiveness. For example, by using a pre-set weight lookup table, fixed weight combinations can be selected based on the slope's soil and rock type and seasonal climate, allowing the contribution of each sensor to be switchable. However, this method relies on discrete, predefined static weight templates. For highway slopes in seasonal freeze-thaw zones, their deformation response does not correspond to a fixed dry or rainy season pattern, but rather exhibits continuous evolutionary characteristics of frost heave stability, thawing creep, and accelerated failure. Under different evolutionary stages, the early warning indicative value of parameters such as rainfall, water content, and displacement changes dynamically: during the thawing creep period, changes in water content are crucial for judging the stability trend; while during the accelerated failure period, displacement rate and acceleration are more direct indicators. Existing weight adjustment strategies based on discrete mode switching cannot adaptively and nonlinearly adjust to follow this continuous evolutionary physical process, making it difficult to accurately capture instability precursors, leading to delayed early warnings or inaccurate judgments. Therefore, the following solutions are proposed to address these issues. Summary of the Invention

[0005] The purpose of this invention is to provide a multi-dimensional early warning method and electronic device for slopes based on LSTM neural networks. By extracting multi-source temporal features through LSTM neural networks and combining them with a deformation evolution stage identification mechanism to adaptively generate weights, the contribution of each parameter can be continuously adjusted according to the freeze-thaw evolution process, thus solving the problem of insufficient ability of existing static model methods to identify gradual slope instability.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: This invention relates to a multi-dimensional early warning method for slopes based on LSTM neural networks, comprising the following steps: Rainfall, soil moisture content, surface displacement and internal displacement monitoring data were collected at the slope site. The data were cleaned and normalized to form a time series dataset. Construct and train an LSTM neural network model that takes the time series dataset as input and outputs the original probability vectors for each warning level. The velocity and acceleration characteristics of surface and internal displacements are calculated in real time, a discrimination factor for deformation evolution stages is constructed, and an adaptive weight vector is generated based on the discrimination factor. The original probability vector output by the LSTM model and the current normalized shift value are weighted and combined using the adaptive weight vector to obtain a comprehensive risk score. The comprehensive risk score is compared with preset multi-level thresholds to trigger the corresponding level of warning and release information.

[0007] Furthermore, the data cleaning includes using the Laida criterion to identify and remove outliers, and using linear interpolation to fill in short-term missing data; the normalization process uses Min-Max standardization to map each parameter to a uniform numerical range.

[0008] Furthermore, the LSTM neural network model includes an input layer, at least two stacked LSTM hidden layers, and a fully connected output layer connected to the hidden layers. The output layer uses a Softmax activation function to output probabilities corresponding to multiple warning levels.

[0009] Furthermore, the rates of the surface displacement and the internal displacement are calculated by the first-order difference of the corresponding displacement sequence, and the acceleration is calculated by the second-order difference of the corresponding displacement sequence.

[0010] Furthermore, the discrimination factor for the deformation evolution stage is constructed by weighted summation of the maximum values ​​among the surface displacement rate, internal displacement rate, surface displacement acceleration, and internal displacement acceleration, and its expression is: In the formula, As a discriminant factor for the deformation and evolution stages, , These are the average rates of surface displacement and internal displacement, respectively. , These are the average accelerations of surface displacement and internal displacement, respectively. , This is the adjustment coefficient.

[0011] Furthermore, the adaptive weight vector includes rainfall weight, moisture content weight, surface displacement weight, and internal displacement weight. Each weight value is determined by the discrimination factor of the deformation evolution stage after being mapped by an exponential function and normalized, so that when the discrimination factor increases, the displacement weight increases and the rainfall and moisture content weights decrease.

[0012] Furthermore, the comprehensive risk score is obtained by adding the products of surface displacement weight and current normalized surface displacement value, internal displacement weight and current normalized internal displacement value, rainfall weight and sum of high-level warning probabilities, and moisture content weight and sum of high-level warning probabilities.

[0013] Furthermore, the LSTM neural network model is trained in a supervised manner using historical monitoring data and corresponding slope status labels, with the training loss function being classification cross-entropy and the optimizer being the Adam algorithm.

[0014] Furthermore, the multi-level threshold includes at least three numerical ranges corresponding to different warning levels, and a warning of the corresponding level is triggered when the comprehensive risk score falls into the corresponding range.

[0015] The present invention also includes an electronic device comprising a memory and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors of any of the above-described slope multi-dimensional early warning method.

[0016] The present invention has the following beneficial effects: This invention introduces a deformation evolution stage identification mechanism into the dynamic weight allocation process, enabling the contribution of each monitoring parameter to the early warning decision to adaptively adjust according to the actual deformation development state of the slope, thereby enhancing the adaptability of the early warning system to complex environmental conditions such as freeze-thaw cycles. The deep feature extraction of multi-source time-series data by the LSTM neural network, combined with the weight generation algorithm based on real-time deformation rate and acceleration, helps to reflect the differences in the indicative role of each physical quantity under different evolution stages, thereby reducing the misjudgment that may be caused by single parameter evaluation or fixed weight mode. In addition, the multi-level early warning triggering method can output differentiated response measures based on the comprehensive risk score, which is conducive to relevant personnel taking targeted inspection and disposal arrangements.

[0017] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a multi-dimensional early warning method for slopes based on an LSTM neural network according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 As shown, this invention is a multi-dimensional early warning method for slopes based on LSTM neural networks. The early warning method includes the following steps: Rainfall, soil moisture content, surface displacement and internal displacement monitoring data were collected at the slope site. The data were cleaned and normalized to form a time series dataset. Construct and train an LSTM neural network model that takes a time series dataset as input and outputs the original probability vectors for each warning level. Real-time calculation of the rate and acceleration characteristics of surface and internal displacements; construction of discrimination factors for deformation evolution stages; generation of adaptive weight vectors based on discrimination factors. The original probability vector output by the LSTM model and the current normalized shift value are weighted and combined using an adaptive weight vector to obtain a comprehensive risk score. The comprehensive risk score is compared with preset multi-level thresholds to trigger the corresponding level of warning and release information.

[0022] Data cleaning includes using the Laida criterion to identify and remove outliers, and using linear interpolation to fill in short-term missing data; normalization processing uses Min-Max standardization to map each parameter to a uniform numerical range.

[0023] An LSTM neural network model consists of an input layer, at least two stacked LSTM hidden layers, and a fully connected output layer connected to the hidden layers. The output layer uses the Softmax activation function to output probabilities corresponding to multiple warning levels.

[0024] The rates of surface and internal displacements are calculated using the first-order difference of the corresponding displacement sequences, and the accelerations are calculated using the second-order difference of the corresponding displacement sequences.

[0025] The discrimination factor for deformation evolution stages is constructed by weighted summation of the maximum values ​​among surface displacement rate, internal displacement rate, surface displacement acceleration, and internal displacement acceleration. Its expression is: In the formula, As a discriminant factor for the deformation and evolution stages, , These are the average rates of surface displacement and internal displacement, respectively. , These are the average accelerations of surface displacement and internal displacement, respectively. , This is the adjustment coefficient.

[0026] The adaptive weight vector includes rainfall weight, water content weight, surface displacement weight, and internal displacement weight. Each weight value is determined by the discrimination factor of the deformation evolution stage through an exponential function mapping and normalization, so that the displacement weight increases and the rainfall and water content weights decrease when the discrimination factor increases.

[0027] The comprehensive risk score is obtained by adding the products of surface displacement weight and current normalized surface displacement value, internal displacement weight and current normalized internal displacement value, rainfall weight and sum of high-level warning probabilities, and moisture content weight and sum of high-level warning probabilities.

[0028] The LSTM neural network model is trained in a supervised manner using historical monitoring data and corresponding slope status labels. The training loss function is classification cross-entropy, and the optimizer is the Adam algorithm.

[0029] The multi-level threshold includes at least three numerical ranges corresponding to different warning levels. When the comprehensive risk score falls into the corresponding range, the corresponding level of warning is triggered.

[0030] The specific application of this embodiment is as follows: Step S1: Multi-source data acquisition and configurable preprocessing Step S11: Deploy an IoT monitoring network consisting of multiple sensors on the slope site, specifically including: A tipping bucket rain gauge is used to collect real-time rainfall intensity data; A tubular soil moisture sensor is buried at a representative depth in the slope to collect soil volumetric moisture content data. GNSS surface displacement monitoring stations are deployed at key locations on the slope surface to collect absolute surface displacement data. A fixed guide wheel inclinometer is installed inside the slope along the borehole to collect the relative changes in deep horizontal displacement. Each sensor collects data synchronously at a fixed sampling interval and transmits it back to the edge computing gateway or cloud server in real time via a wireless communication module. Step S12: Perform data cleaning operations, using the Raida criterion to detect outliers in the time series data of each parameter channel. Any values ​​exceeding the specified range will be removed. The numerical points are considered outliers and are removed. This is the mean value within the sliding window of that channel. The standard deviation is used; for missing values ​​at a single time step, if the missing interval is less than or equal to 3 sampling periods, linear interpolation is used to fill in the missing values ​​using the valid values ​​before and after; if there are more than 3 consecutive missing values, the entire time period is marked as an invalid time period and is automatically skipped by the system in subsequent model inference. Step S13: Normalize the cleaned data to eliminate the impact of differences in the dimensions of different physical quantities on neural network training and weight calculation; this application uses the Min-Max normalization method to linearly map the original values ​​of each parameter to... The normalization formula for the interval is: In the formula, This refers to the original sampled value of a certain parameter; and These are the maximum and minimum values ​​of the parameter in the training dataset, respectively, which remain unchanged after model deployment; after the above preprocessing, a well-organized four-dimensional time series dataset is finally formed. ,in These are the normalized rainfall, moisture content, surface displacement, and internal displacement, respectively. This corresponds to the slope status label. For time indexing.

[0031] Step S2: LSTM model construction, training, and deployment A multi-input, single-output LSTM neural network model is constructed to extract deep spatiotemporal features from historical multi-parameter time series and output the probabilities of early warning at various levels. The specific model structure is defined as follows: Input layer: Receives a time window with a length of [length missing]. A three-dimensional tensor with shape as ,in Take 24, which means it includes monitoring data from the past 24 hours. Corresponding to four characteristic channels: rainfall, moisture content, surface displacement, and internal displacement; Hidden Layer: Two stacked LSTM layers are used to enhance temporal feature learning capabilities; the first LSTM layer has 32 units, and the second layer has 16 units; each LSTM layer outputs a hidden state sequence, and the second LSTM layer outputs the hidden state at the last time step. As the extracted high-level temporal feature vector; Output layer: Connected to a fully connected layer that uses the Softmax activation function, the number of nodes... These correspond to four states: normal, blue alert, yellow alert, and red alert, respectively; the model output vector... satisfy In the formula, The slope predicted by the model belongs to the first... The probability of a class state; During the model training phase, a training set is constructed using historical monitoring data of this slope and adjacent slopes with similar geological conditions. The historical data needs to be labeled with status tags by geotechnical engineering experts based on actual inspection records and historical disaster reports. Supervised learning is employed during the training process, with the loss function being the classification cross-entropy. In the formula, The total number of training samples; This is the index number for the warning category; For the first The sample at the th Real labels in each category; The model predicts the first The sample belongs to the first The probability values ​​of each category are calculated; the optimizer uses the Adam algorithm, the initial learning rate is set to 0.001, the training cycle limit is set to 100, and an early stopping mechanism is introduced to terminate training early when the loss on the validation set no longer decreases for 5 consecutive cycles to prevent overfitting; the trained model parameters are serialized and stored, and deployed to the edge computing unit on the slope site or the cloud platform inference engine.

[0032] Step S3: Adaptive dynamic weight allocation and real-time analysis of the fusion deformation evolution stage identification Step S31: Calculate deformation rate and acceleration characteristics; at each real-time inference time step Using an LSTM model, normalized sequences of surface and internal displacements within an input time window are used to calculate the statistical characteristics of the first and second differences of the two types of displacements, respectively. Taking surface displacement as an example, its normalized sequence is: Define the average rate within the window. With average acceleration for: In the formula, In time step The average rate of change of surface displacement within the corresponding time window; In time step The average acceleration of the surface displacement within the corresponding time window; The time window length is the input to the LSTM model; For index variables; In the surface displacement sequence, at time index... The normalized displacement value at time t.

[0033] Similarly, calculate the average rate of internal displacement. With average acceleration When there are insufficient data points within the window, the mean is calculated using the effective difference count. Step S32: Construct the discrimination factor for deformation evolution stage; define the comprehensive deformation activity index. It integrates the rate and acceleration information of surface and internal displacements: In the formula, and The adjustment coefficient is determined by inversion calibration based on historical data of the slope. This represents the absolute value of the average surface displacement rate; is the absolute value of the average velocity of internal displacement; is the absolute value of the average acceleration of surface displacement. This represents the absolute value of the average acceleration of the internal displacement. The larger the value, the more active the slope is in a stage of accelerated deformation. Step S33: Generate an adaptive weight vector based on the deformation and evolution stages; define the weight generation function. Its output is a four-dimensional weight vector. The weighting coefficients for rainfall, moisture content, surface displacement, and internal displacement, respectively, satisfy the following conditions: The function is designed according to the following principles: during the stable or creeping phase... When the moisture content is relatively low, the weights of water content and rainfall are relatively high to capture environmental triggering factors; in the accelerated destructive stage, i.e. When the displacement is large, the displacement weight is increased to enhance the response to direct signs of instability; specifically, the following normalized mapping formula is used: In the formula, In time step Weighting coefficients assigned to surface displacement parameters; In time step Weighting coefficients assigned to internal displacement parameters; In time step Weighting coefficients assigned to rainfall parameters; In time step Weighting coefficients assigned to the moisture content parameter; Sensitivity factor; This is a normalization constant; this set of formulas ensures that when As the weight increases, the weights of surface displacement and internal displacement increase exponentially, while the weights of rainfall and moisture content decrease exponentially, and the sum of all weights is always 1. Step S34: Calculate the weighted comprehensive risk score; convert the original early warning probability vector output by the LSTM model in step S2 into a weighted comprehensive risk score. With adaptive weight vector The data should be integrated; considering that the warning probability itself already includes the model's confidence level for each state, the weight adjustment should focus on strengthening or suppressing the probability of the danger level; therefore, a comprehensive risk score is defined. for: In the formula, and The normalized displacement value at the current moment directly reflects the magnitude of the deformation; This is the sum of the high-level warning probabilities output by the LSTM. Through the above formula, the weight vector acts on both the original physical quantity and the model inference probability, realizing an adaptive warning decision that combines data-driven and physics-driven approaches. This step addresses the issue of seasonal freeze-thaw cycles affecting highway slopes, where deformation exhibits phased evolution characteristics: frost heave stabilization, thawing creep, and accelerated failure. It also addresses the changing indicative role of monitoring parameters at different evolutionary stages. An adaptive dynamic weight adjustment method based on real-time deformation rate and acceleration characteristics is proposed. This method no longer relies on pre-defined static geological and climatic models for table lookup. Instead, it quantifies the current deformation evolution stage of the slope and non-linearly generates the weight coefficients of each sensor at each time step, achieving continuous and adaptive evolution of the early warning contribution.

[0034] Step S4: Multi-level early warning triggering and information dissemination The system will calculate the comprehensive risk score obtained in step S3. The alert level is determined by comparing it with preset multi-level thresholds; the thresholds are set based on statistical analysis of historical cases, for example: when At that time, a red emergency alert was triggered; when At that time, a yellow alert was triggered; when At that time, a blue alert was triggered; when When the time is right, the status is determined to be normal; Once an early warning is triggered, the system automatically generates an early warning report, which includes the warning level, trigger time, slope location, and main contributing parameters (determined by comparing each weighted component with its corresponding physical quantity or probability value). Early warning information is pushed through multiple channels: red warnings send real-time SMS messages to at least three emergency response personnel and display a prominent pop-up window on the monitoring platform; yellow warnings send email notifications accompanied by a flashing yellow indicator on the platform; blue warnings are only displayed as a blue alert label on the platform interface for on-duty personnel to monitor. All early warning events and subsequent on-site handling feedback are recorded by the system, forming a historical case library for subsequent model retraining and threshold optimization, constituting a complete closed-loop learning mechanism.

[0035] The present invention also includes an electronic device comprising a memory and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors of any of the above-described slope multi-dimensional early warning method.

[0036] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0037] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-dimensional early warning method for slopes based on LSTM neural networks, characterized in that, The early warning method includes the following steps: Rainfall, soil moisture content, surface displacement and internal displacement monitoring data were collected at the slope site. The data were cleaned and normalized to form a time series dataset. Construct and train an LSTM neural network model that takes the time series dataset as input and outputs the original probability vectors for each warning level. The velocity and acceleration characteristics of surface and internal displacements are calculated in real time, a discrimination factor for deformation evolution stages is constructed, and an adaptive weight vector is generated based on the discrimination factor. The original probability vector output by the LSTM model and the current normalized shift value are weighted and combined using the adaptive weight vector to obtain a comprehensive risk score. The comprehensive risk score is compared with preset multi-level thresholds to trigger the corresponding level of warning and release information.

2. The slope multi-dimensional early warning method based on LSTM neural network according to claim 1, characterized in that, The data cleaning includes using the Laida criterion to identify and remove outliers, and using linear interpolation to fill in short-term missing data; the normalization process uses Min-Max standardization to map each parameter to a uniform numerical range.

3. The slope multi-dimensional early warning method based on LSTM neural network according to claim 1, characterized in that, The LSTM neural network model includes an input layer, at least two stacked LSTM hidden layers, and a fully connected output layer connected to the hidden layers. The output layer uses the Softmax activation function to output probabilities corresponding to multiple warning levels.

4. The slope multi-dimensional early warning method based on LSTM neural network according to claim 1, characterized in that, The rates of the surface displacement and the internal displacement are calculated by the first-order difference of the corresponding displacement sequence, and the acceleration is calculated by the second-order difference of the corresponding displacement sequence.

5. A multi-dimensional early warning method for slopes based on an LSTM neural network according to claim 4, characterized in that, The discrimination factor for the deformation evolution stage is constructed by weighted summation of the maximum values ​​among surface displacement rate, internal displacement rate, surface displacement acceleration, and internal displacement acceleration. Its expression is as follows: In the formula, As a discriminant factor for the deformation and evolution stages, , These are the average rates of surface displacement and internal displacement, respectively. , These are the average accelerations of surface displacement and internal displacement, respectively. , This is the adjustment coefficient.

6. A multi-dimensional early warning method for slopes based on an LSTM neural network according to claim 5, characterized in that, The adaptive weight vector includes rainfall weight, water content weight, surface displacement weight, and internal displacement weight. Each weight value is determined by the discrimination factor of the deformation evolution stage after being mapped by an exponential function and normalized, so that when the discrimination factor increases, the displacement weight increases and the rainfall and water content weights decrease.

7. A multi-dimensional early warning method for slopes based on an LSTM neural network according to claim 6, characterized in that, The comprehensive risk score is obtained by adding the products of surface displacement weight and current normalized surface displacement value, internal displacement weight and current normalized internal displacement value, rainfall weight and sum of high-level warning probabilities, and moisture content weight and sum of high-level warning probabilities.

8. A multi-dimensional early warning method for slopes based on an LSTM neural network according to claim 1, characterized in that, The LSTM neural network model is trained in a supervised manner using historical monitoring data and corresponding slope status labels. The training loss function is classification cross-entropy, and the optimizer is the Adam algorithm.

9. A multi-dimensional early warning method for slopes based on an LSTM neural network according to claim 1, characterized in that, The multi-level threshold includes at least three numerical ranges corresponding to different warning levels. When the comprehensive risk score falls into the corresponding range, the corresponding level of warning is triggered.

10. An electronic device, characterized in that, It includes a memory, and one or more instructions, wherein one or more instructions are stored in the memory and configured to be executed by one or more processors as described in any one of claims 1-9.

Images