A vehicle-mounted geological disaster early warning system and method based on environmental perception

Abstract

The present invention relates to a vehicle-mounted geological disaster early warning system and method based on environmental perception, and relates to the field of vehicle early warning technology. The system includes: an information acquisition module, which is used to obtain geological disaster information, vehicle information and external environment information in real time; a judgment and analysis module, which predicts the vehicle driving path through a driving path prediction unit; the geological disaster prediction unit inputs the geological disaster information into a pre-trained geological disaster early warning model to predict the probability value of geological disasters at each monitoring point or area; the risk assessment unit assesses the risk level of the impact of geological disasters on the vehicle; and an early warning execution module, which is used to match safety recommendations according to the risk level and provide early warning prompts. The present invention can effectively improve driving safety by comprehensively acquiring information, using a geological disaster prediction model to assess risks in real time, and providing early warnings and safety recommendations in a timely manner.

Description

Technical Field

[0001] The present invention relates to the field of vehicle early warning technology, and in particular to a vehicle-mounted geological disaster early warning system and method based on environmental perception. Background Art

[0002] When geological disasters such as earthquakes, landslides, mudslides and collapses occur, drivers cannot get timely information through the on-board system and can only learn about it through mobile phones and other channels, and are unable to assess the degree of danger of the current situation.

[0003] Traditional geological disaster early warning systems mostly rely on a single type of monitoring data, have limitations when processing complex spatiotemporal data, and find it difficult to accurately capture the correlation and characteristics between different disasters.

[0004] With the development of sensor technology and big data analysis, more and more real-time data can be collected and utilized, providing new opportunities for geological disaster early warning. However, how to effectively integrate multi-source data and build a comprehensive geological disaster early warning system through advanced machine learning and deep learning technologies to improve the accuracy and timeliness of early warning remains a challenge. Summary of the invention

[0005] In view of the lack of geological disaster warning during driving in the prior art, the present invention proposes a vehicle-mounted geological disaster warning system and method based on environmental perception.

[0006] In order to achieve the above object, the present invention is implemented by the following technical solutions:

[0007] A vehicle-mounted geological disaster early warning system based on environmental perception, the system comprising:

[0008] An information acquisition module is used to acquire geological disaster information, vehicle information and external environment information in real time, wherein the geological disaster information includes seismic waveform data and terrain image data, the vehicle information includes navigation route, vehicle real-time position, real-time vehicle speed, vehicle historical driving trajectory data and driver preference data, and the external environment information includes historical traffic data, real-time traffic information and weather data;

[0009] Judgment and analysis module, including driving path prediction unit, geological disaster prediction unit and risk assessment unit;

[0010] The driving path prediction unit is used to predict the vehicle driving path based on vehicle information and external environment information;

[0011] The geological disaster prediction unit is used to input geological disaster information into a pre-trained geological disaster early warning model to predict the probability value of geological disasters occurring at each monitoring point or area;

[0012] The geological disaster warning model adopts a multi-task learning architecture, which can simultaneously learn multiple goals such as earthquake intensity prediction, landslide risk assessment, mudslide warning and collapse warning, and promote knowledge transfer by sharing the underlying feature layer; the geological disaster warning model is specifically a convolutional long expression memory model, which integrates convolution operations into the long expression memory architecture, combines graph neural networks and self-supervised contrastive learning, and introduces DCT transformation in time series prediction;

[0013] The risk assessment unit is used to assess the risk level of the impact of geological disasters on vehicles based on the probability value of geological disasters occurring at each monitoring point or area;

[0014] The early warning execution module is used to match security recommendations according to risk levels and provide early warning prompts.

[0015] As a preferred solution of the present invention, the method for predicting a vehicle driving path based on vehicle information and external environment information includes:

[0016] Perform data cleaning, denoising and feature extraction on vehicle information and external environment information to generate a structured data set for training;

[0017] The vehicle's historical driving trajectory data and real-time traffic information are trained through a recurrent neural network or a long short-term memory network to build a time series prediction model and predict the vehicle's driving path based on the time series data;

[0018] Use federated learning technology to train data locally and share model parameters without directly sharing the original data;

[0019] A deep Q-learning algorithm is used to optimize the prediction results and simulate the feedback mechanism of different path selections;

[0020] Based on real-time traffic data and weather data, the self-attention mechanism is used to dynamically adjust and update the prediction results in real time, and timely feedback is given to the driver to adapt to the real-time changing traffic conditions.

[0021] As a preferred solution of the present invention, the method for constructing the geological disaster early warning model includes:

[0022] Data preprocessing: normalize the geological disaster information from multiple sources, build an adjacency matrix based on the geographical location relationship, and indicate the connection strength between each monitoring point;

[0023] Feature extraction: 1D CNN is used to extract features from seismic waveform data. For terrain image data, 2D CNN is used to extract terrain information.

[0024] The input feature map is transformed by DCT It is divided into multiple feature vectors along the channel dimension, and DCT transform of different frequency components is applied to each feature vector. The formula is:

[0025] ;

[0026] In the formula, It is The frequency representation of the feature vector after DCT processing; represents the feature vector index, , represents the number of eigenvectors; Representation feature map Middle feature vectors; represents the frequency index, , is the length of DCT transformation; is the element index in the feature vector; Representation feature map Middle The eigenvector elements; is the basis function of DCT;

[0027] Graph neural network block operation: Using ChebConv and GCSConv layers, the spatial and temporal information is transferred according to the adjacency matrix to encode the spatiotemporal relationship in the graph structure. The graph neural network block operation is expressed as:

[0028] ;

[0029] In the formula, For the The feature vector of each node; is the number of nodes; is the adjacency matrix; Represents the input feature vector and the adjacency matrix The output of the graph neural network block;

[0030] Self-supervised contrastive learning: By constructing positive and negative samples and defining a contrastive loss function, the model can learn the intrinsic relationship between data without relying on a large amount of labeled data. The contrastive loss function is expressed as:

[0031] ;

[0032] In the formula, is the contrast loss; is a positive sample pair and The similarity function between is the negative sample set, is the negative sample set Any element in It is a positive sample and negative samples The similarity function between is the temperature parameter, which is used to adjust the output of the similarity function;

[0033] State update of the convolutional long expression memory model: By integrating the convolution operation into the long expression memory architecture to capture multi-scale spatiotemporal structures, the state update equation is as follows:

[0034] ;

[0035] ;

[0036] In the formula, and Represent the fast hidden state and the slow hidden state respectively; is the time step Observed value of and are gain factors, used to adjust and Rate of change and They are and The update function of and The update functions are and Parameters to adjust the behavior of the function; represents the Hadamard product;

[0037] An implicit-explicit time stepping scheme is adopted in the convolutional long expression memory model, and the state update equation is solved using an implicit Euler method or a Runge-Kutta method;

[0038] Multi-task learning: Introduce a multi-task learning mechanism to simultaneously learn multiple objectives such as earthquake intensity prediction, landslide risk assessment, and mudslide warning, and promote knowledge transfer by sharing the underlying feature layer.

[0039] As a preferred solution of the present invention, in the multi-task learning mechanism, a softmax function is used to convert the output into the probability of each category, which is expressed as:

[0040] ;

[0041] In the formula, For the The scores of the categories; It is The original output value of each category (logits, that is, the unnormalized value output by the model in the last layer); is the total number of predicted categories.

[0042] As a preferred solution of the present invention, in the risk assessment unit, the probability values ​​of geological disasters occurring at each monitoring point or area on the vehicle driving path are weighted averaged to obtain the comprehensive risk probability of the entire vehicle driving path, and the risks are divided into different levels according to the comprehensive risk probability:

[0043] When the comprehensive risk probability is between 0 and 0.2, the risk level is low risk;

[0044] When the comprehensive risk probability is between 0.2 and 0.6, the risk level is medium risk;

[0045] When the comprehensive risk probability is between 0.6 and 1.0, the risk level is high risk;

[0046] The weights are determined based on the historical disaster frequency, terrain, and distance from the vehicle's travel path of each monitoring point or area.

[0047] As a preferred solution of the present invention, the risk assessment unit also includes collecting images from a vehicle-mounted camera, and using a target detection algorithm to determine whether there are trees, billboards, tall buildings, mountains, boulders or other objects with a risk of falling around the vehicle. If so, a falling risk weight is added to the comprehensive risk probability to further divide the risk level.

[0048] As a preferred solution of the present invention, the early warning execution module includes:

[0049] The geological information display unit is used to display geological disaster information through the central control screen or voice broadcast, and generate early warning information to remind users;

[0050] The safety suggestion matching unit is used to match safety suggestions according to the current risk level of the vehicle.

[0051] A vehicle-mounted geological disaster early warning method based on environmental perception, based on the above-mentioned vehicle-mounted geological disaster early warning system based on environmental perception, the method comprises:

[0052] Real-time acquisition of geological disaster information, vehicle information and external environment information, wherein the geological disaster information includes seismic waveform data and terrain image data, the vehicle information includes navigation route, vehicle real-time position, real-time vehicle speed, vehicle historical driving trajectory data and driver preference data, and the external environment information includes historical traffic data, real-time traffic information and weather data;

[0053] Predicting the vehicle's driving path based on vehicle information and external environment information;

[0054] Input geological disaster information into the pre-trained geological disaster early warning model to predict the probability value of geological disasters occurring at each monitoring point or area;

[0055] According to the probability value of geological disasters at each monitoring point or area, the risk level of the impact of geological disasters on vehicles is assessed;

[0056] Match security recommendations based on risk levels and provide early warning prompts.

[0057] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the above-mentioned vehicle-mounted geological disaster early warning method based on environmental perception.

[0058] Compared with the prior art, the present invention has the following beneficial effects: it can obtain diversified information in real time, so that the system can more accurately predict the risk of geological disasters and respond in time; it can predict the driving path of the vehicle by combining vehicle information and external environment information, which helps the system to identify potential geological disaster risk areas in advance, so as to carry out targeted early warning; the geological disaster early warning model using a multi-task learning architecture can simultaneously learn multiple goals such as earthquake intensity prediction, landslide risk assessment, mudslide warning and collapse warning, thereby improving the generalization performance and prediction accuracy of the model, so that the system can more accurately predict the probability of occurrence of geological disasters; according to the probability value of the occurrence of geological disasters, the risk level of the impact of geological disasters on vehicles is evaluated in real time. This real-time risk assessment capability enables the system to respond in time before the disaster occurs, provide safety advice to the driver, and help the driver make correct decisions when facing potential geological disaster risks, thereby significantly improving driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0060] Figure 1 It is a system modular structure diagram of the present invention;

[0061] Figure 2 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the embodiment of the present invention clearer, the technical solution of the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings of the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all of the embodiments. Based on the described embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention.

[0063] like Figure 1 As shown in FIG. 1 , an embodiment of the present invention is provided, which provides a vehicle-mounted geological disaster early warning system based on environmental perception, including:

[0064] (1) Information acquisition module

[0065] The information acquisition module is used to obtain geological disaster information, vehicle information and external environment information in real time. The geological disaster information includes seismic waveform data, terrain image data, etc. The vehicle information includes navigation route, real-time vehicle position, real-time vehicle speed, historical vehicle driving trajectory data, driver preference data (driving habits, preferences, etc.), etc. The external environment information includes historical traffic data, real-time traffic information (such as traffic congestion, road construction, traffic accidents, etc.), weather data, etc.;

[0066] (2) Judgment analysis module

[0067] It includes driving route prediction unit, geological disaster prediction unit and risk assessment unit;

[0068] 1) The driving path prediction unit is used to predict the vehicle driving path based on vehicle information and external environment information, and the method includes:

[0069] Perform data cleaning, denoising and feature extraction on vehicle information and external environment information to generate a structured data set for training;

[0070] The vehicle's historical driving trajectory data and real-time traffic information are trained through a recurrent neural network or a long short-term memory network to build a time series prediction model and predict the vehicle's driving path based on the time series data;

[0071] Federated learning technology is used to train data locally and share model parameters instead of directly sharing raw data, thereby enhancing data privacy protection and improving the generalization ability of the path prediction model through collaborative training of different data sources;

[0072] The deep Q-network (DQN) algorithm is used to optimize the prediction results. By simulating the feedback mechanism of different path selections, the prediction results are more accurate and adaptable to complex traffic environments.

[0073] Based on real-time traffic data and weather data, the self-attention mechanism is used to dynamically adjust and update the prediction results in real time, and timely feedback is given to the driver to adapt to the real-time changing traffic conditions.

[0074] 2) The geological disaster prediction unit is used to input geological disaster information into the pre-trained geological disaster early warning model to predict the probability value of geological disasters occurring at each monitoring point or area;

[0075] The geological disaster warning model adopts a multi-task learning architecture, which can simultaneously learn multiple goals such as earthquake intensity prediction, landslide risk assessment, mudslide warning and collapse warning, and promote knowledge transfer by sharing the underlying feature layer; the geological disaster warning model is specifically a convolutional long expression memory model, which integrates convolution operations into the long expression memory (Long Expressive Memory, LEM) architecture, combines graph neural network (Graph Neural Network, GNN) and self-supervised contrastive learning, and introduces DCT transform (Discrete Cosine Transform) in time series prediction;

[0076] The long expression memory architecture is a recurrent neural network unit designed to process data with complex spatiotemporal dependencies. The LEM architecture enhances the capabilities of traditional recurrent units (such as LSTM or GRU) by introducing multi-scale temporal dynamics, thereby better capturing long-term dependencies and multi-scale patterns in the data. By integrating convolution operations into the LEM architecture, the model's ability to process spatial information is further enhanced. This enables the model to better capture multi-scale spatial patterns in the input data.

[0077] The construction methods of geological disaster early warning model include:

[0078] 1. Data preprocessing:

[0079] Normalize multi-source geological disaster information (such as seismic waveform data, terrain image data, meteorological data, etc.), and construct an adjacency matrix based on geographical location relationships to represent the connection strength between monitoring points;

[0080] 2. Feature extraction:

[0081] Convolutional feature extraction: 1D CNN (one-dimensional convolutional neural network) is used to extract features from seismic waveform data. For terrain image data, 2D CNN (two-dimensional convolutional neural network) is used to extract terrain information.

[0082] Frequency enhanced channel attention mechanism: The input feature map is transformed through DCT It is divided into multiple feature vectors along the channel dimension, and DCT transform of different frequency components is applied to each feature vector to capture more frequency information and avoid the Gibbs phenomenon in Fourier transform. The formula is:

[0083] ;

[0084] In the formula, It is The frequency representation of the feature vector after DCT processing; represents the feature vector index, , represents the number of eigenvectors; Representation feature map Middle feature vectors; represents the frequency index, , is the length of DCT transformation; is the element index in the feature vector; Representation feature map Middle The eigenvector elements; is the basis function of DCT;

[0085] 3. Graph Neural Network (GNN) block operations:

[0086] Using ChebConv and GCSConv layers, spatial and temporal information is transferred according to the adjacency matrix to encode the spatiotemporal relationship in the graph structure. The graph neural network block operation is expressed as:

[0087] ;

[0088] In the formula, For the The feature vector of each node; is the number of nodes; is the adjacency matrix; Represents the input feature vector and the adjacency matrix The output of the graph neural network block;

[0089] ChebConv (Chebyshev Convolution) is a convolution operation used in graph neural networks (GNNs). It approximates graph filters based on Chebyshev polynomials. This convolution method can effectively process local and non-local information in graph-structured data, thereby better capturing the complex relationships between nodes in the graph.

[0090] GCSConv (Graph Skip Convolution) can be expressed as "graph skip convolution". "Jump" refers to the ability of this layer to enhance local structure and feature diffusion through trainable skip connections. Specifically, the GCSConv layer combines spectral domain diffusion (focusing on the global structure of the graph) and spatial domain diffusion (emphasizing the local neighborhood structure of each node) to better capture local and global information in graph data. In the application of earthquake early warning and other geological disaster prediction, the GCSConv layer can help the model more effectively encode the complex spatiotemporal relationship between monitoring points and improve the accuracy and reliability of predictions.

[0091] 4. Self-supervised contrastive learning:

[0092] Self-supervised contrastive learning is introduced to enhance the generalization ability of the model. By constructing positive and negative samples and defining a suitable contrastive loss function, the model can learn the intrinsic relationship between data without relying on a large amount of labeled data. The contrastive loss function is expressed as:

[0093] ;

[0094] In the formula, It is the contrast loss, which is used to measure the similarity difference between positive sample pairs and negative sample pairs; is a positive sample pair and Similarity function between them. Common similarity functions include cosine similarity or dot product. is the negative sample set, is the negative sample set Any element in It is a positive sample and negative samples The similarity function between is the temperature parameter used to adjust the output of the similarity function. It is usually a small positive number used to control the sharpness of the similarity function.

[0095] 5. State update of convolutional long expression memory model:

[0096] By integrating convolution operations into a long representation memory architecture to capture multi-scale spatiotemporal structures, the state update equation is as follows:

[0097] ;

[0098] ;

[0099] In the formula, and Represent the fast hidden state and the slow hidden state respectively; is the time step Observed value of and are gain factors, used to adjust and The rate of change of and They are and The update function of and The update functions are and Parameters to adjust the behavior of the function; represents the Hadamard product;

[0100] In the convolutional long expression memory model, an implicit-explicit (IMEX) time stepping scheme is adopted to improve numerical stability. For example, the implicit Euler method or the Runge-Kutta method can be used to solve the above state update equation.

[0101] 6. Multi-task learning:

[0102] The introduction of a multi-task learning mechanism enables the model to simultaneously learn multiple objectives such as earthquake intensity prediction, landslide risk assessment, and debris flow warning, and promotes knowledge transfer by sharing the underlying feature layer;

[0103] 7. Output the predicted probability value:

[0104] The final output of the model can be a continuous value (such as earthquake intensity), a classification label (such as risk level), or other forms of prediction results. For multi-classification tasks, the softmax function is directly used to convert the output into the probability of each category, expressed as:

[0105] ;

[0106] In the formula, For the The score of each geological hazard category; It is The original output value of each geological disaster category (logits, i.e. the unnormalized value output by the model at the last layer); is the total number of predicted geological disaster categories.

[0107] 3) The risk assessment unit is used to assess the risk level of the impact of geological disasters on vehicles based on the probability value of geological disasters occurring at each monitoring point or area;

[0108] In the risk assessment unit, the probability values ​​of geological disasters occurring at each monitoring point or area on the vehicle's driving path are weighted averaged to obtain the comprehensive risk probability of the entire vehicle's driving path. According to the comprehensive risk probability, the risk is divided into different levels:

[0109] When the comprehensive risk probability is between 0 and 0.2, the risk level is low risk;

[0110] When the comprehensive risk probability is between 0.2 and 0.6, the risk level is medium risk;

[0111] When the comprehensive risk probability is between 0.6 and 1.0, the risk level is high risk;

[0112] Among them, the weights are determined based on the historical disaster frequency, terrain, and distance from the vehicle's travel path at each monitoring point or area;

[0113] The risk assessment unit also includes collecting images from the on-board camera and using the target detection algorithm to determine whether there are trees, billboards, tall buildings, mountains, boulders or other objects with a risk of falling around the vehicle. If so, the falling risk weight is added to the comprehensive risk probability to further divide the risk level.

[0114] Optionally, the earthquake risk assessment criteria may also be set as:

[0115] High risk: earthquake magnitude greater than 6, focal depth 0-60 km, within 80 km from the epicenter, with tall buildings, mountains, boulders, bridges, electric poles, etc. nearby (if the earthquake is within 50 km of the seabed and the magnitude is higher than 6, the risk will be increased: within 200 meters from the coastline and within 10 meters above sea level);

[0116] Medium risk: Earthquake magnitude 4 to 6, focal depth 60-300 km, distance from the epicenter 80-160 km, nearby street lights, trees, high walls, fences, billboards, etc. (If the earthquake occurs in the sea, the risk will increase: 200-1000 meters from the coastline, 10-30 meters above sea level);

[0117] Low risk: Earthquake magnitude is less than 4, focal depth is more than 300 kilometers, and distance from the epicenter is more than 160 kilometers, etc.

[0118] (3) Early warning execution module

[0119] The warning execution module is used to match security recommendations according to risk levels and provide warning prompts, including:

[0120] The geological information display unit is used to display geological disaster information through the central control screen or voice broadcast, and generate early warning information to remind users;

[0121] The warning prompt mode can be: red for high risk, yellow for medium risk, and blue for low risk, making the risk prompt more intuitive;

[0122] The safety suggestion matching unit is used to match safety suggestions according to the current risk level of the vehicle.

[0123] like Figure 2 As shown, another embodiment of the present invention provides a vehicle-mounted geological disaster early warning method based on environmental perception, based on the above-mentioned vehicle-mounted geological disaster early warning system based on environmental perception, comprising the following steps:

[0124] S1: Real-time acquisition of geological disaster information, vehicle information and external environment information. Geological disaster information includes seismic waveform data and terrain image data. Vehicle information includes navigation route, vehicle real-time location, real-time vehicle speed, vehicle historical driving trajectory data and driver preference data. External environment information includes historical traffic data, real-time traffic information and weather data.

[0125] S2: predicting the vehicle's driving path based on vehicle information and external environment information;

[0126] S3: Input the geological disaster information into the pre-trained geological disaster early warning model to predict the probability value of geological disasters occurring at each monitoring point or area;

[0127] S4: Assess the risk level of the impact of geological disasters on vehicles based on the probability value of geological disasters occurring at each monitoring point or area;

[0128] S5: Match security recommendations based on risk levels and provide early warning prompts.

[0129] Each part of the present application may be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiment, a plurality of steps or methods may be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above-mentioned embodiment method may be completed by instructing the relevant hardware through a program, and the program may be stored in a computer-readable storage medium, and the program, when executed, includes one of the steps of the method embodiment or a combination thereof. Therefore, an embodiment of the present invention further provides a computer-readable storage medium, which may be a read-only memory, a disk or an optical disk, etc., on which a computer program is stored, and when the computer program is executed by a processor, a vehicle-mounted geological disaster early warning method based on environmental perception as described above is implemented.

[0130] In summary, the present invention can obtain diversified information in real time, so that the system can more accurately predict the risk of geological disasters and respond in time; combining vehicle information and external environment information to predict the vehicle's driving path helps the system to identify potential geological disaster risk areas in advance, so as to carry out targeted early warning; the geological disaster early warning model using a multi-task learning architecture can simultaneously learn multiple goals such as earthquake intensity prediction, landslide risk assessment, mudslide warning and collapse warning, thereby improving the generalization performance and prediction accuracy of the model, so that the system can more accurately predict the probability of occurrence of geological disasters; according to the probability value of the occurrence of geological disasters, the risk level of the impact of geological disasters on vehicles is evaluated in real time. This real-time risk assessment capability enables the system to respond in time before the disaster occurs, provide safety advice to the driver, and help the driver make correct decisions when facing potential geological disaster risks, thereby significantly improving driving safety.

[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A vehicle-mounted geological disaster early warning system based on environmental perception, characterized in that: The system comprises: An information acquisition module is used to acquire geological disaster information, vehicle information and external environment information in real time, wherein the geological disaster information includes seismic waveform data and terrain image data, the vehicle information includes navigation route, vehicle real-time position, real-time vehicle speed, vehicle historical driving trajectory data and driver preference data, and the external environment information includes historical traffic data, real-time traffic information and weather data; Judgment and analysis module, including driving path prediction unit, geological disaster prediction unit and risk assessment unit; The driving path prediction unit is used to predict the vehicle driving path based on vehicle information and external environment information; The geological disaster prediction unit is used to input geological disaster information into a pre-trained geological disaster early warning model to predict the probability value of geological disasters occurring at each monitoring point or area; The geological disaster warning model adopts a multi-task learning architecture, which can simultaneously learn multiple goals such as earthquake intensity prediction, landslide risk assessment, mudslide warning and collapse warning, and promote knowledge transfer by sharing the underlying feature layer; the geological disaster warning model is specifically a convolutional long expression memory model, which integrates convolution operations into the long expression memory architecture, combines graph neural networks and self-supervised contrastive learning, and introduces DCT transformation in time series prediction; The risk assessment unit is used to assess the risk level of the impact of geological disasters on vehicles based on the probability value of geological disasters occurring at each monitoring point or area; In the risk assessment unit, the probability values ​​of geological disasters occurring at each monitoring point or area on the vehicle's driving path are weighted averaged to obtain a comprehensive risk probability of the entire vehicle's driving path; The early warning execution module is used to match security recommendations according to risk levels and provide early warning prompts.

2. The vehicle-mounted geological disaster early warning system based on environmental perception according to claim 1 is characterized in that: The method for predicting a vehicle driving path based on vehicle information and external environment information includes: Perform data cleaning, denoising and feature extraction on vehicle information and external environment information to generate a structured data set for training; The vehicle's historical driving trajectory data and real-time traffic information are trained through a recurrent neural network or a long short-term memory network to build a time series prediction model and predict the vehicle's driving path based on the time series data; Use federated learning technology to train data locally and share model parameters without directly sharing the original data; A deep Q-learning algorithm is used to optimize the prediction results and simulate the feedback mechanism of different path selections; Based on real-time traffic data and weather data, the self-attention mechanism is used to dynamically adjust and update the prediction results in real time, and timely feedback is given to the driver to adapt to the real-time changing traffic conditions.

3. The vehicle-mounted geological disaster early warning system based on environmental perception according to claim 1 is characterized in that: The method for constructing the geological disaster early warning model includes: Data preprocessing: normalize the geological disaster information from multiple sources, build an adjacency matrix based on the geographical location relationship, and indicate the connection strength between each monitoring point; Feature extraction: 1D CNN is used to extract features from seismic waveform data. For terrain image data, 2D CNN is used to extract terrain information. The input feature map is transformed by DCT It is divided into multiple feature vectors along the channel dimension, and DCT transform of different frequency components is applied to each feature vector. The formula is: ； In the formula, It is The frequency representation of the feature vector after DCT processing; represents the feature vector index, , represents the number of eigenvectors; Representation feature map Middle feature vectors; represents the frequency index, , is the length of DCT transformation; is the index of the element in the feature vector; Representation feature map Middle The eigenvector elements; is the basis function of DCT; Graph neural network block operation: Using ChebConv and GCSConv layers, the spatial and temporal information is transferred according to the adjacency matrix to encode the spatiotemporal relationship in the graph structure. The graph neural network block operation is expressed as: ； In the formula, For the The feature vector of each node; is the number of nodes; is the adjacency matrix; Represents the input feature vector and the adjacency matrix The output of the graph neural network block; Self-supervised contrastive learning: By constructing positive and negative samples and defining a contrastive loss function, the model can learn the intrinsic relationship between data without relying on a large amount of labeled data. The contrastive loss function is expressed as: ； In the formula, is the contrast loss; is a positive sample pair and The similarity function between is the negative sample set, is the negative sample set Any element in It is a positive sample and negative samples The similarity function between is the temperature parameter, which is used to adjust the output of the similarity function; State update of the convolutional long expression memory model: By integrating the convolution operation into the long expression memory architecture to capture multi-scale spatiotemporal structures, the state update equation is as follows: ； ； In the formula, and Represent the fast hidden state and the slow hidden state respectively; is the time step Observed value of and are gain factors, used to adjust and The rate of change of and They are and The update function of and The update functions are and Parameters to adjust the behavior of the function; represents the Hadamard product; An implicit-explicit time stepping scheme is adopted in the convolutional long expression memory model, and the state update equation is solved using an implicit Euler method or a Runge-Kutta method; Multi-task learning: Introduce a multi-task learning mechanism to simultaneously learn multiple objectives such as earthquake intensity prediction, landslide risk assessment, and mudslide warning, and promote knowledge transfer by sharing the underlying feature layer.

4. The vehicle-mounted geological disaster early warning system based on environmental perception according to claim 3 is characterized in that: In the multi-task learning mechanism, the softmax function is used to convert the output into the probability of each category, expressed as: ； In the formula, For the The scores of the categories; It is The original output value of each category; is the total number of predicted categories.

5. The vehicle-mounted geological disaster early warning system based on environmental perception according to claim 1 is characterized in that: In the risk assessment unit, risks are divided into different levels according to the comprehensive risk probability: When the comprehensive risk probability is between 0 and 0.2, the risk level is low risk; When the comprehensive risk probability is between 0.2 and 0.6, the risk level is medium risk; When the comprehensive risk probability is between 0.6 and 1.0, the risk level is high risk; The weights are determined based on the historical disaster frequency, terrain, and distance from the vehicle's travel path of each monitoring point or area.

6. The vehicle-mounted geological disaster early warning system based on environmental perception according to claim 5 is characterized in that: The risk assessment unit also includes collecting images from the vehicle camera and using a target detection algorithm to determine whether there are trees, billboards, tall buildings, mountains, boulders or other objects with a risk of falling around the vehicle. If so, a falling risk weight is added to the comprehensive risk probability to further divide the risk level.

7. The vehicle-mounted geological disaster early warning system based on environmental perception according to claim 1 is characterized in that: The early warning execution module includes: The geological information display unit is used to display geological disaster information through the central control screen or voice broadcast, and generate early warning information to remind users; The safety suggestion matching unit is used to match safety suggestions according to the current risk level of the vehicle.

8. The early warning method of the vehicle-mounted geological disaster early warning system based on environment perception according to any one of claims 1 to 7, characterized in that: The method comprises: Real-time acquisition of geological disaster information, vehicle information and external environment information, wherein the geological disaster information includes seismic waveform data and terrain image data, the vehicle information includes navigation route, vehicle real-time position, real-time vehicle speed, vehicle historical driving trajectory data and driver preference data, and the external environment information includes historical traffic data, real-time traffic information and weather data; Predicting the vehicle's driving path based on vehicle information and external environment information; Input geological disaster information into the pre-trained geological disaster early warning model to predict the probability value of geological disasters occurring at each monitoring point or area; According to the probability value of geological disasters at each monitoring point or area, the risk level of the impact of geological disasters on vehicles is assessed; Match security recommendations based on risk levels and provide early warning prompts.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the warning method of the vehicle-mounted geological disaster warning system based on environmental perception as described in claim 8.

Images