Geological disaster automatic identification system and method based on multi-source remote sensing data

Abstract

The invention discloses an automatic geological disaster recognition system and method based on multi-source remote sensing data, and particularly relates to the field of geological disaster recognition, and the system comprises a multi-modal remote sensing data acquisition module, a cross-domain physical fusion module, a spatio-temporal evolution decision module, a multi-cascade early warning decision module, an optimization control module and a visualization module. According to the geological disaster automatic identification system and method based on the multi-source remote sensing data, virtual features are generated through a cross-domain physical fusion module by using a domain adversarial network, the model generalization ability during cross-domain application is improved, physical association among the multi-source remote sensing data is deeply mined, and dependence on manual design rules is eliminated; through a three-layer processing chain technology composed of a spatial-temporal feature extraction layer, a dynamic graph evolution layer and a critical recognition layer, the capability of capturing disaster features in a complex geological environment is effectively improved, especially the recognition precision of precursor tiny deformation is improved, and the risk of missing report is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of geological disaster identification, and more specifically, to a geological disaster automatic identification system and method based on multi-source remote sensing data. Background Art

[0002] With the rapid development of remote sensing technology and the demand for geological hazard monitoring, the need to use remote sensing data for geological hazard identification is becoming increasingly prominent. Traditional geological hazard identification methods rely primarily on manual interpretation of single-source remote sensing imagery, consisting of satellite remote sensing imagery, drone aerial data, and ground sensors. By manually interpreting the texture features of visible light imagery, geological hazard signs are identified and ultimately, a comprehensive risk assessment is conducted. However, these traditional methods rely heavily on the experience of experts, are inefficient, and are susceptible to subjective factors, making it difficult to achieve large-scale, rapid geological hazard identification. This is particularly true in areas with complex terrain and vegetation, where identification accuracy can be significantly reduced.

[0003] In order to solve the problems of insufficient efficiency and accuracy of traditional methods, existing technologies build end-to-end convolutional neural network models, use image processing algorithms and machine learning to realize the automatic extraction of geological disaster characteristics, and adopt parallel computing architecture to improve processing speed.

[0004] However, it still has some shortcomings in actual use. For example, the generalization ability of existing models is limited when applied across regions. When the geological structure and environmental background of the target area are significantly different from the training data set, the model recognition accuracy will drop significantly. At the same time, current multimodal data fusion strategies mostly remain at the feature level, failing to deeply explore the physical connections between different remote sensing data sources, and the feature extraction process often relies on manually designed rules, which makes it difficult to fully capture the disaster characteristics in complex geological environments. In particular, for precursory minor deformations, the recognition accuracy is not high and there is a risk of missed reports. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a geological disaster automatic identification system and method based on multi-source remote sensing data, which solves the problems raised in the above-mentioned background technology through the following scheme.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A geological hazard automatic identification system based on multi-source remote sensing data includes a system operation database, a system central processing module and a user information terminal, and also includes:

[0008] Multimodal remote sensing data acquisition module: used for receiving and performing spatial registration and radiation correction on the first remote sensing data to generate a first remote sensing feature;

[0009] A cross-domain physical fusion module is configured to combine the first remote sensing feature with a cross-domain physical fusion mechanism and output a second remote sensing feature. The cross-domain physical fusion mechanism includes a domain adversarial network mechanism and a physical constraint mechanism.

[0010] Spatiotemporal evolution decision module: used to obtain a third remote sensing feature by applying the second remote sensing feature to an automatic recognition model, wherein the automatic recognition model is a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical recognition layer;

[0011] Multi-level early warning decision module: based on the third remote sensing feature, constructing a geological disaster assessment matrix and generating a graded early warning signal;

[0012] An optimization control module is configured to receive the first remote sensing feature of the target area in real time and perform an optimization control operation, wherein the optimization control operation is configured to optimize and adjust the cross-domain physical fusion module and the spatiotemporal evolution decision module;

[0013] Visualization module: used to map the graded warning signals into a three-dimensional geological model map to display the explanation and analysis of the causes of geological disasters;

[0014] The system operation database includes all data texts of the system operation and collects the information text output by each module in real time. The system central processing module is used for the information text instructions output by each module in the central control system, and the user information terminal is the information output device of the receiving system.

[0015] Preferably, the multimodal remote sensing data acquisition module, the first remote sensing data includes satellite multispectral images, InSAR radar interferometry data and lidar point cloud data, wherein the satellite multispectral images include multispectral data of visible light band, near infrared band and shortwave infrared band, the InSAR radar interferometry data is composed of master-slave image pairs, and the lidar point cloud data includes three-dimensional coordinates, laser echo intensity and the time difference between the first and last echoes.

[0016] Preferably, the cross-domain physical fusion module processes the first remote sensing feature through the cross-domain physical fusion mechanism, specifically including:

[0017] A deep domain adversarial network composed of a feature extractor, a label classifier, and a domain classifier is used to generate a virtual feature corresponding to the first remote sensing feature;

[0018] Construct a physical correlation graph between the first remote sensing feature and the corresponding geological parameters of the target area, and mine the constitutive relationship through graph neural network;

[0019] The physical association graph is based on geological prior knowledge and uses a Bayesian network to model the conditional dependency between the first remote sensing features;

[0020] The second remote sensing feature includes the domain-adapted feature vector enhanced by the target region and the physical association graph structure.

[0021] Preferably, the spatiotemporal evolution decision module passes the second remote sensing feature through the automatic recognition model, and the three-layer processing chain of the automatic recognition model satisfies:

[0022] The size of the 3D dilated convolution kernel in the spatiotemporal feature extraction layer is positively correlated with the node degree of the physical association graph;

[0023] The node state vector dimension of the hazard body spatiotemporal graph corresponding to the dynamic graph evolution layer is equal to the hidden layer dimension of the LSTM unit in the critical recognition layer;

[0024] The output gradient of the critical recognition layer is back-propagated to the expansion rate adjustment coefficient of the spatiotemporal feature extraction layer.

[0025] Preferably, the spatiotemporal evolution decision module, the weight matrix WM corresponding to the three-dimensional dilated convolution kernel ck Dynamic calculation is performed based on the node degree D in the physical association graph, which is specifically expressed as:

[0026]

[0027] Among them, D i Expressed as the degree of node i in the physical association graph, ∑D j It is expressed as the sum of the degrees of all nodes in the physical association graph, It represents the normalization of node degrees through the Softmax function, and WM0 represents the weight matrix corresponding to the pre-set three-dimensional dilated convolution kernel.

[0028] Preferably, the spatiotemporal evolution decision module calculates the priority of the deformation propagation path through the structural attention head in the bimodal graph attention network, which is specifically expressed as:

[0029]

[0030] Among them, p (ti,tj) It is represented as the priority of the deformation propagation path from node ti to node tj, and p is represented as a learnable weight vector that measures the similarity of node attributes corresponding to the spatiotemporal graph structure of the disaster body. T Represented as the transpose operation on p, Wh ti and Wh tj Expressed as the embedding representation of node ti and node tj, [Wh i ‖Wh j ] is represented as the concatenated vector of the embedding representation of node ti and node tj, LeakyReLU (p T [Wh i ‖Wh j]) is expressed as p T [Wh i ‖Wh j ] Apply the LeakyReLU activation function, exp(LeakyReLU(p T [Wh ti ‖Wh tj ])) represents the LeakyReLU(p T [Wh i ‖Wh j ])The exponential function is applied to the activation result, and tk is represented as all neighbor nodes of node ti.

[0031] Preferably, the operation method of the critical identification layer of the spatiotemporal evolution decision module specifically includes:

[0032] The cumulative effect characteristics of the disaster body deformation rate are extracted through the time series of node states in the disaster body space-time graph;

[0033] The mutation synergy characteristics are captured through the flattened sequence of the adjacency matrix and node characteristics of the disaster body spatiotemporal graph;

[0034] The fusion weight is automatically adjusted according to the warning level at time t-1.

[0035] Preferably, in the multi-level cascade warning decision module, the graded warning signal is a disaster risk level instruction generated based on the quantitative analysis results of the probabilistic risk assessment matrix; the disaster risk level instruction includes four warning levels, including level I warning, level II warning, level III warning and level IV warning, among which level I warning indicates that a geological disaster is about to occur or initial destructive characteristics have appeared; level II warning indicates that the disaster body has entered a rapid deformation stage and there is a possibility of short-term instability; level III warning indicates that the geological disaster risk level has increased significantly and the local area has entered a deformation accumulation stage; level IV warning indicates that there is a potential risk in the target area.

[0036] To achieve the above-mentioned object, the present invention provides the following technical solution: a method for automatically identifying geological hazards based on multi-source remote sensing data, and a system for automatically identifying geological hazards based on multi-source remote sensing data, comprising:

[0037] S1: Receive and perform spatial registration and radiometric correction on first remote sensing data to generate a first remote sensing feature;

[0038] S2: The first remote sensing feature is subjected to a cross-domain physical fusion mechanism to output a second remote sensing feature, wherein the cross-domain physical fusion mechanism includes a domain adversarial network mechanism and a physical constraint mechanism;

[0039] S201: A deep domain adversarial network composed of a feature extractor, a label classifier, and a domain classifier generates a virtual feature corresponding to the first remote sensing feature and outputs a domain-adapted feature vector enhanced for the target region;

[0040] S202: constructing a physical association graph between the first remote sensing feature and the geological parameters corresponding to the target area, wherein the physical association graph is based on geological prior knowledge, uses a Bayesian network to model the conditional dependency relationship between multi-source data, and uses a graph neural network to mine the constitutive relationship between the modes, and outputs a physical association graph structure representing the modes;

[0041] S3: Obtain a third remote sensing feature by applying the second remote sensing feature to an automatic recognition model, wherein the automatic recognition model is a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical recognition layer;

[0042] S301: Using three-dimensional dilated convolution to capture the multi-scale spatiotemporal features in the second remote sensing feature, and fusing the multi-scale spatiotemporal features according to the dilation coefficient;

[0043] S302: Construct a spatiotemporal graph structure of the disaster body corresponding to the multi-scale spatiotemporal features, identify the deformation propagation path through the bimodal graph attention network, and output the spatiotemporal evolution characteristics of the disaster body;

[0044] S303: Construct the disaster body dynamic process based on the LSTM-Transformer hybrid framework to output the disaster body evolution label;

[0045] S4: constructing a geological disaster assessment matrix based on the third remote sensing feature and generating a graded warning signal;

[0046] S5: receiving the first remote sensing feature of the target area in real time, and performing an optimization control operation, wherein the optimization control operation is used to optimize and adjust the cross-domain physical fusion mechanism and the automatic recognition model;

[0047] S6: Mapping the graded warning signals into a three-dimensional geological model map to display an explanation and analysis of the causes of geological disasters.

[0048] Preferably, the step S5, performing an optimization control operation, specifically includes:

[0049] When the KL divergence between the first remote sensing feature of the target area and the historical data exceeds a preset threshold, updating the domain adaptation network parameters of the domain adversarial network mechanism through online knowledge distillation;

[0050] If the domain adaptation network parameters of the domain adversarial network mechanism are triggered, the weight distribution of the three-dimensional dilated convolution kernel is adjusted synchronously.

[0051] Technical effects and advantages of the present invention:

[0052] 1. This invention solves the problem of registration error accumulation caused by the non-uniform spatial benchmarks of multi-source data through the collaborative optimization technology of spatial registration and radiometric correction, significantly reducing the risk of generalization capability degradation caused by data heterogeneity in cross-regional applications.

[0053] 2. This invention uses a domain adversarial network to generate virtual features through a cross-domain physical fusion module, improving the model's generalization ability in cross-regional applications. It solves the problem of reduced recognition accuracy in existing models caused by differences in the geological environment between the target area and the training data. At the same time, it deeply explores the physical correlation between multi-source remote sensing data and gets rid of the reliance on manually designed rules.

[0054] 3. The present invention effectively improves the ability to capture disaster characteristics in complex geological environments through a three-layer processing chain technology consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical identification layer. In particular, it improves the accuracy of identifying precursory micro-deformations, reduces the risk of missed reports, and breaks through the limitations of traditional feature-level fusion strategies in physical mechanism mining and feature expression. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a module structure diagram of a geological disaster automatic identification system based on multi-source remote sensing data provided according to an embodiment of the present application.

[0056] Figure 2 This is a flow chart of a method for automatically identifying geological hazards based on multi-source remote sensing data provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] The terms used in the following examples of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification of the present application, the singular expressions "a", "an", "said", "above", "the", and "this" are intended to include plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used in the present application refers to and includes any or all possible combinations of one or more of the listed items.

[0059] In the following, the terms "first," "second," and "third" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of the technical features indicated. Therefore, a feature defined as "first," "second," and "third" may explicitly or implicitly include one or more of the features. In the description of the embodiments of this application, unless otherwise specified, "plurality" means two or more.

[0060] As attached Figure 1 The system is a geological hazard automatic identification system based on multi-source remote sensing data, comprising a system operation database, a system central processing module, and a user information terminal. The system operation database contains all data texts of system operation and collects information texts output by each module in real time. The system central processing module is used to control the information text instructions output by each module in the central control system. The user information terminal is an information output device for receiving the system.

[0061] It should be noted that the system central processor is the core operation and control unit of the entire geological disaster automatic identification system based on multi-source remote sensing data; it includes one or more processing cores, which connect the modules in the system through the PCIe bus and the Gigabit Ethernet interface; by running or executing the instructions, programs, code sets or instruction sets stored in the system operation database, and being able to call the data stored therein, it performs various functions of a geological disaster automatic identification system based on multi-source remote sensing data, including spatial alignment and radiation correction of multimodal remote sensing data, feature parameter extraction, disaster risk assessment and other operations to ensure the accuracy and efficiency of identification; the system operation The database is used to store a large amount of data related to the automatic identification of geological hazards, including a mathematical association rule library established based on geomechanics principles and remote sensing inversion models, and a large amount of historical monitoring data, including time-series change records of parameters such as deformation rate, rock permeability, rainfall threshold, as well as types of geological hazards that have occurred and corresponding remote sensing image features; when executing various functions, the system central processor will frequently call these data from the system operation database to perform data correction, model parameter adjustment, risk threshold setting and other operations, thereby realizing real-time monitoring and early warning of automatic identification of geological hazards based on multi-source remote sensing data; the user information terminal connects to the 4K geological three-dimensional display screen and multispectral imager through the HDMI 2.1 interface and USB4 protocol, providing users with an interface for interacting with the system.

[0062] This embodiment provides an automatic identification system for geological hazards based on multi-source remote sensing data, further comprising a multimodal remote sensing data acquisition module, a cross-domain physical fusion module, a spatiotemporal evolution decision module, a multi-cascade early warning decision module, an optimization control module, and a visualization module;

[0063] The multimodal remote sensing data acquisition module is used to receive and perform spatial registration and radiation correction on the first remote sensing data to generate a first remote sensing feature;

[0064] The cross-domain physical fusion module is used to combine the first remote sensing features through a cross-domain physical fusion mechanism and output a second remote sensing feature. The cross-domain physical fusion mechanism includes a domain adversarial network mechanism and a physical constraint mechanism.

[0065] The spatiotemporal evolution decision module is used to obtain a third remote sensing feature by applying the second remote sensing feature to an automatic recognition model, wherein the automatic recognition model is a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical recognition layer;

[0066] The multi-level early warning decision module constructs a geological disaster assessment matrix based on the third remote sensing feature and generates a graded early warning signal;

[0067] The optimization control module is used to receive the first remote sensing feature of the target area in real time and perform an optimization control operation, wherein the optimization control operation is used to optimize and adjust the cross-domain physical fusion module and the spatiotemporal evolution decision module;

[0068] The visualization module is used to map the graded warning signals into a three-dimensional geological model map to display the explanation and analysis of the causes of geological disasters.

[0069] Specifically, in the multimodal remote sensing data acquisition module, collaborative observation and collection of first remote sensing data are performed through multi-platform remote sensing sensors, wherein the first remote sensing data includes satellite multispectral images, InSAR radar interferometry data and lidar point cloud data, and the first remote sensing data is preprocessed. The preprocessing operation is used for spatial registration and radiation correction of the first remote sensing data to obtain first remote sensing features corresponding to the first remote sensing data, and the first remote sensing features include multispectral features, InSAR features and lidar features.

[0070] It should be noted that the satellite multispectral images in the first remote sensing data include multispectral data of visible light bands, near-infrared bands and short-wave infrared bands, the InSAR radar interferometry measurement data consists of master-slave image pairs, and the lidar point cloud data includes three-dimensional coordinates, laser echo intensity and the time difference between the first and last echoes; the multispectral features in the first remote sensing features include the apparent reflectance, normalized vegetation index and normalized water index of each band, the InSAR features include the average annual deformation rate, coherence coefficient matrix and elevation change, and the lidar features include digital elevation models and initial classification labels based on the number of echoes and laser echo intensity. The initial classification labels include but are not limited to ground points, vegetation points, building points, etc.

[0071] The method of using multi-platform remote sensing sensors to obtain the first observed remote sensing data in this embodiment includes but is not limited to using Landsat-8OLI and Sentinel-2MSI satellite multispectral images, covering the 450-680nm visible light band, the 700-900nm near-infrared band and the 1550-2300nm short-wave infrared band, with a spatial resolution of 10-30 meters and a temporal resolution of 5-10 days; at the same time, obtaining C-band radar interferometry data with a wavelength of 5.6cm from the Sentinel-1SAR satellite, generating differential interferograms through ascending and descending orbit observations, with a spatial resolution of 5×20 meters and a deformation measurement accuracy of ±3mm / year, so as to accurately monitor small surface deformations; obtaining high-density lidar point cloud data with a density of ≥20 points / m through the airborne RIEGL VQ-1560i system. 2 , wavelength is 1550nm, and elevation accuracy is ±5cm.

[0072] Furthermore, the format of the raw data transmitted by multi-platform remote sensing sensors is standardized, including: converting satellite multispectral images into radiometric brightness values and embedding them into the WGS84 geographic coordinate system; multi-viewing the InSAR radar interferometry data to reduce speckle noise and generate phase and coherence coefficient matrices; and using the statistical outlier removal algorithm to denoise the lidar point cloud data and classify ground points and non-ground points to generate a digital elevation model.

[0073] Furthermore, the first remote sensing data after format standardization is subjected to the spatial registration, including: using the digital elevation model generated by the lidar point cloud data as the spatial reference, and resampling the digital elevation model to the UTM projection coordinate system using the bilinear interpolation algorithm; extracting scale-invariant feature points on the differential interferogram in the satellite multispectral image and the InSAR radar interferometry data, with the parameters set to contrast > 0.03 and curvature ratio < 10, and using the Brute-Force matcher for two-way matching, screening matching points with a matching score > 0.7; applying the Levenberg-Marquardt algorithm to the initial matching points for position correction, and using the RANSAC algorithm to eliminate false matches; performing affine transformation based on the matching points, and solving the transformation parameters by the least squares method, with the residual controlled to within 0.3 pixels; and resampling the low-resolution data to the digital elevation model using the bicubic convolution interpolation method.

[0074] Furthermore, the radiation correction of the first remote sensing data after the format standardization includes: using the 6S radiation transfer model to perform atmospheric correction to output the apparent reflectance of the surface for satellite multispectral images; applying the SCS+C model based on the digital elevation model to eliminate the terrain shadow effect; using the minimum cost flow MCF algorithm to process the winding phase for InSAR radar interferometry data; and performing distance attenuation correction on the laser echo intensity for the lidar point cloud data, which is specifically expressed as follows:

[0075]

[0076] Wherein, I1 represents the laser echo intensity after distance attenuation correction, I0 represents the original laser echo intensity, R represents the laser ranging value, and R0 represents the preset distance.

[0077] Specifically, in the cross-domain physical fusion module, the first remote sensing feature is fused through a cross-domain physical fusion mechanism, which is a domain adversarial network mechanism and a physical constraint mechanism. The domain adversarial network mechanism adopts a deep domain adversarial network composed of a feature extractor, a label classifier and a domain classifier to generate a virtual feature corresponding to the first remote sensing feature, and outputs a domain adaptation feature vector enhanced in the target area; the physical constraint mechanism constructs a physical correlation graph between the first remote sensing feature and the geological parameters corresponding to the target area. The physical correlation graph is based on geological prior knowledge and uses a Bayesian network to model the conditional dependency relationship between multi-source data. The constitutive relationship between modalities is mined through a graph neural network, and the physical correlation graph structure representing the modalities is output; the output second remote sensing feature includes the domain adaptation feature vector enhanced in the target area and the physical correlation graph structure.

[0078] The feature extractor of this embodiment uses a ResNet-50 backbone network, removes the last fully connected layer, and outputs a 1024-dimensional feature vector; the label classifier uses a bilinear interpolation layer to output a probability distribution of disaster types; the domain classifier consists of a gradient reversal layer and a fully connected layer, and outputs binary classification results for the source domain and the target domain; a three-layer Bayesian network is used to construct a physical association map between the first remote sensing feature and the geological parameters corresponding to the target area. The observation layer nodes in the three-layer Bayesian network include the apparent reflectivity of each band, the InSAR deformation rate, and the initial classification labels based on the number of echoes and the laser echo intensity; the hidden layer nodes include the geological parameters corresponding to the target area; and the decision layer nodes include the risk probability of geological disasters stored in the system operation database; the physical association map between the characterization modes is composed of nodes composed of multispectral features, InSAR features, and geological parameters. The edge types include physical mechanism edges encoding geomechanical equations, statistical correlation edges connecting feature pairs with mutual information greater than 0.3, and spatiotemporal co-occurrence edges connecting spatially adjacent and temporally synchronized nodes.

[0079] Specifically, in the spatiotemporal evolution decision module, the second remote sensing feature is passed through an automatic recognition model, which is composed of a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer and a critical recognition layer; the third remote sensing feature includes multi-scale spatiotemporal features, spatiotemporal evolution features of disaster bodies and disaster body evolution labels.

[0080] Furthermore, the automatic recognition model is a pre-constructed three-layer processing chain. By inputting the second remote sensing feature into the automatic recognition model, the automatic recognition model obtains the third remote sensing feature based on the second remote sensing feature. The three-layer processing chain of the automatic recognition model includes: the spatiotemporal feature extraction layer uses three-dimensional dilated convolution to capture the multi-scale spatiotemporal features in the second remote sensing feature, and fuses the multi-scale spatiotemporal features according to the dilation coefficient, wherein the size of the three-dimensional dilated convolution kernel is positively correlated with the node degree of the physical association graph; the dynamic graph evolution layer constructs a spatiotemporal graph structure of the disaster body corresponding to the multi-scale spatiotemporal features, identifies the deformation propagation path through the bimodal graph attention network, and outputs the spatiotemporal evolution characteristics of the disaster body, wherein the node state vector dimension of the spatiotemporal graph of the disaster body corresponding to the dynamic graph evolution layer is equal to the hidden layer dimension of the LSTM unit in the critical recognition layer; the critical recognition layer constructs the dynamic process of the disaster body based on the LSTM-Transformer hybrid framework to output the disaster body evolution label, wherein the output gradient of the critical recognition layer is back-propagated to the dilation rate adjustment coefficient of the spatiotemporal feature extraction layer.

[0081] Furthermore, the implementation steps of the spatiotemporal feature extraction layer include: adopting a three-dimensional dilated convolution architecture, wherein the weight matrix WM corresponding to the three-dimensional dilated convolution kernel is ck According to the node degree D in the physical association graph i Perform dynamic calculation, specifically expressed as:

[0082]

[0083] Among them, D i Expressed as the degree of node i in the physical association graph, ∑D j It is expressed as the sum of the degrees of all nodes in the physical association graph, It is represented by the normalization processing of node degrees through the Softmax function, and WM0 is represented by the weight matrix corresponding to the preset three-dimensional dilated convolution kernel; the multi-scale spatiotemporal features include spatial scale features and temporal scale features; in this embodiment, the spatial scale features capture surface microtopography changes through a three-dimensional convolution kernel with a small dilation rate; the convolution kernel with a medium dilation rate is used to capture medium-scale spatial associations such as terrain gradient and slope consistency; the convolution kernel with a large dilation rate extracts macro-topographic trends; the time window for capturing sudden events of the temporal scale features is less than 7 days, the time window for seasonal changes is set at 1-6 months, and the time window for the cumulative effect of human engineering is set at more than 1 year.

[0084] Furthermore, the construction process of the dynamic graph evolution layer includes: converting multi-scale spatiotemporal features into a spatiotemporal graph structure of the hazard body, where the node attributes corresponding to the spatiotemporal graph structure of the hazard body include multispectral features, InSAR features, and lidar features; and calculating the priority of the deformation propagation path through the structural attention head in the bimodal graph attention network, which is specifically expressed as:

[0085]

[0086] Among them, p (ti,tj) It is represented as the priority of the deformation propagation path from node ti to node tj, and p is represented as a learnable weight vector that measures the similarity of node attributes corresponding to the spatiotemporal graph structure of the disaster body. T Represented as the transpose operation on p, Wh ti and Wh tj Expressed as the embedding representation of node ti and node tj, [Wh i ‖Wh j ] is represented as the concatenated vector of the embedding representation of node ti and node tj, LeakyReLU (p T [Wh i ‖Wh j ]) is expressed as p T [Wh i ‖Wh j ] Apply the LeakyReLU activation function, exp(LeakyReLU(p T [Wh ti ‖Wh tj ])) represents the LeakyReLU(p T [Wh i ‖Wh j ])The exponential function is applied to the activation result, and tk is represented as all neighbor nodes of node ti.

[0087] In this embodiment, the nodes of the spatiotemporal graph structure of the disaster body correspond to 50×50 meter grid cells. Spatial adjacency edges are automatically connected within the 3×3 neighborhood grid. Deformation propagation edges establish directed edges based on the direction of the displacement gradient and are activated when the gradient is greater than 5 mm / 100 m. Physical constraint edges filter valid connections using the Mohr-Coulomb criterion. The graph structure is refreshed every 6 hours. Regions with sudden deformation rate changes greater than 10 mm / day are automatically divided into subgrids. Decoherent regions with a coherence coefficient less than 0.25 are suspended from monitoring. The propagation weight of relevant edges is strengthened by real-time rainfall data greater than 50 mm / h.

[0088] Furthermore, the operation method of the critical identification layer includes: the LSTM branch extracts the cumulative effect characteristics of the disaster body deformation rate through the time series of node states in the disaster body spatiotemporal graph; the Transformer branch captures the mutation synergy characteristics through the flattened sequence of the adjacency matrix and node features of the disaster body spatiotemporal graph; and the fusion weight of the LSTM and Transformer outputs is automatically adjusted according to the warning level at time t-1, where time t-1 is represented as the moment before the current time t.

[0089] Specifically, in the multi-cascade warning decision module, a probabilistic risk assessment matrix is constructed based on the spatiotemporal evolution characteristics of the disaster body and the disaster body evolution label in the third remote sensing feature. The spatial position dimension of the probabilistic risk assessment matrix in this embodiment divides the target area into 1m×1m grid units in the UTM coordinate system, and each unit corresponds to a unique coordinate code. The disaster type dimension is preset to include three types of geological disaster labels: landslide, mudslide and ground subsidence, and the probability value range is [0, 1]. The evolution stage dimension is divided into four stages including stable period, incubation period, acceleration period and critical instability period, and the stage label is output by the critical identification layer; the input features of the risk probability calculation include the spatiotemporal evolution characteristics of the disaster body from the node embedding vector of the spatiotemporal graph structure of the disaster body; a graded warning signal is generated according to the probabilistic risk assessment matrix, and the graded warning signal is back-propagated to the dynamic graph evolution layer to optimize the learnable weight vector in the bimodal graph attention network that measures the similarity of node attributes corresponding to the spatiotemporal graph structure of the disaster body.

[0090] It should be noted that the graded warning signal is a disaster risk level instruction generated based on the quantitative analysis results of the probability risk assessment matrix. The disaster risk level instruction includes four warning levels, namely, level I warning, level II warning, level III warning and level IV warning. The triggering condition of level I warning is that the disaster risk probability enters the extremely high probability interval or the evolution stage is determined to be a critical instability period, indicating that a geological disaster is about to occur or initial destructive characteristics have appeared, and it is necessary to activate the highest level emergency rescue plan, close the surrounding traffic routes and push warning information to public terminals such as SMS and APP in real time through multiple channels; the triggering condition of level II warning is that the disaster risk probability reaches the medium-high probability interval or the evolution stage is determined to be an accelerated deformation period, indicating that the disaster body enters a rapid deformation stage and there is a possibility of short-term instability. It is necessary to demarcate the dangerous area including the potential disaster impact range, initiate the evacuation procedure for residents and implement targeted engineering reinforcement measures; The triggering conditions for the Level Ⅳ warning are that the probability of disaster occurrence enters the medium probability interval or the evolution stage is identified as the incubation period, indicating that the geological disaster risk level has increased significantly and the local area has entered the deformation accumulation stage. It is necessary to activate the high-resolution drone inspection program, start the emergency plan rehearsal, and restrict non-essential personnel from entering the potential risk area; the triggering conditions for the Level Ⅳ warning are that the probability of disaster occurrence is in the low probability interval and the evolution stage is determined to be the stable period, indicating that there is a potential risk in the target area. It is necessary to activate the basic monitoring mechanism, increase the InSAR deformation data update frequency to 1.5-2 times the regular monitoring cycle, send periodic monitoring reports to relevant departments, and recommend regular manual inspections; in this embodiment, the extremely high probability interval of the disaster risk probability is set to [0.9, 1), the medium-high probability interval is set to [0.7, 0.9), the medium probability interval is set to [0.5, 0.7), and the low probability interval is set to [0.3, 0.5).

[0091] Specifically, in the optimization control module, when the KL divergence between the first remote sensing feature of the target area and the historical data exceeds a preset threshold, the domain adaptation network parameters of the domain adversarial network mechanism of the cross-domain physical fusion module are updated through online knowledge distillation to minimize the warning error of the target area and improve the cross-regional generalization ability of the domain adaptation feature vector; if the domain adaptation network parameter update of the domain adversarial network mechanism is triggered, the three-dimensional dilated convolution kernel weight distribution of the spatiotemporal evolution decision module is adjusted synchronously; and the connection edges of non-relevant modes in the physical association graph are automatically closed according to the geological structure type of the target area.

[0092] Specifically, in the visualization module, the graded warning signals are mapped into a geomechanical interpretation model with the physical correlation diagram output by the cross-domain physical fusion module and the spatiotemporal evolution characteristics of the disaster body output by the spatiotemporal evolution decision module, and an interpretable analysis of the causes of the disaster is provided.

[0093] As attached Figure 2The method for automatically identifying geological hazards based on multi-source remote sensing data includes:

[0094] S1: Receive and perform spatial registration and radiometric correction on first remote sensing data to generate a first remote sensing feature;

[0095] S2: The first remote sensing feature is subjected to a cross-domain physical fusion mechanism to output a second remote sensing feature, wherein the cross-domain physical fusion mechanism includes a domain adversarial network mechanism and a physical constraint mechanism;

[0096] S201: A deep domain adversarial network composed of a feature extractor, a label classifier, and a domain classifier generates a virtual feature corresponding to the first remote sensing feature and outputs a domain-adapted feature vector enhanced for the target region;

[0097] S202: constructing a physical association graph between the first remote sensing feature and the geological parameters corresponding to the target area, wherein the physical association graph is based on geological prior knowledge, uses a Bayesian network to model the conditional dependency relationship between multi-source data, and uses a graph neural network to mine the constitutive relationship between the modes, and outputs a physical association graph structure representing the modes;

[0098] S3: Obtain a third remote sensing feature by applying the second remote sensing feature to an automatic recognition model, wherein the automatic recognition model is a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical recognition layer;

[0099] S301: Using three-dimensional dilated convolution to capture the multi-scale spatiotemporal features in the second remote sensing feature, and fusing the multi-scale spatiotemporal features according to the dilation coefficient;

[0100] S302: Construct a spatiotemporal graph structure of the disaster body corresponding to the multi-scale spatiotemporal features, identify the deformation propagation path through the bimodal graph attention network, and output the spatiotemporal evolution characteristics of the disaster body;

[0101] S303: Construct the disaster body dynamic process based on the LSTM-Transformer hybrid framework to output the disaster body evolution label;

[0102] S4: constructing a geological disaster assessment matrix based on the third remote sensing feature and generating a graded warning signal;

[0103] S5: receiving the first remote sensing feature of the target area in real time, and performing an optimization control operation, wherein the optimization control operation is used to optimize and adjust the cross-domain physical fusion mechanism and the automatic recognition model;

[0104] S6: Mapping the graded warning signals into a three-dimensional geological model map to display an explanation and analysis of the causes of geological disasters.

[0105] In this embodiment, in S5, the optimization control operation includes: when the KL divergence between the first remote sensing feature of the target area and the historical data exceeds a preset threshold, updating the domain adaptation network parameters of the domain adversarial network mechanism through online knowledge distillation to minimize the warning error of the target area and improve the cross-regional generalization ability of the domain adaptation feature vector; if the domain adaptation network parameter update of the domain adversarial network mechanism is triggered, the weight distribution of the three-dimensional dilated convolution kernel is adjusted synchronously.

[0106] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict.

[0107] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A geological hazard automatic identification system based on multi-source remote sensing data, comprising a system operation database, a system central processing module and a user information terminal, characterized in that: Also includes: Multimodal remote sensing data acquisition module: used for receiving and performing spatial registration and radiation correction on the first remote sensing data to generate a first remote sensing feature; A cross-domain physical fusion module is configured to combine the first remote sensing feature with a cross-domain physical fusion mechanism and output a second remote sensing feature. The cross-domain physical fusion mechanism includes a domain adversarial network mechanism and a physical constraint mechanism. Spatiotemporal evolution decision module: used to obtain a third remote sensing feature by applying the second remote sensing feature to an automatic recognition model, wherein the automatic recognition model is a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical recognition layer; Multi-level early warning decision module: based on the third remote sensing feature, constructing a geological disaster assessment matrix and generating a graded early warning signal; An optimization control module is configured to receive the first remote sensing feature of the target area in real time and perform an optimization control operation, wherein the optimization control operation is configured to optimize and adjust the cross-domain physical fusion module and the spatiotemporal evolution decision module; Visualization module: used to map the graded warning signals into a three-dimensional geological model map to display the explanation and analysis of the causes of geological disasters; The system operation database includes all data texts of the system operation and collects the information text output by each module in real time. The system central processing module is used for the information text instructions output by each module in the central control system, and the user information terminal is the information output device of the receiving system.

2. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 1 is characterized by: The multimodal remote sensing data acquisition module, the first remote sensing data includes satellite multispectral images, InSAR radar interferometry data and lidar point cloud data, wherein the satellite multispectral images include multispectral data of visible light band, near infrared band and shortwave infrared band, the InSAR radar interferometry data is composed of master-slave image pairs, and the lidar point cloud data includes three-dimensional coordinates, laser echo intensity and the time difference between the first and last echoes.

3. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 1 is characterized in that: The cross-domain physical fusion module processes the first remote sensing feature through the cross-domain physical fusion mechanism, specifically including: A deep domain adversarial network composed of a feature extractor, a label classifier, and a domain classifier is used to generate a virtual feature corresponding to the first remote sensing feature; Construct a physical correlation graph between the first remote sensing feature and the corresponding geological parameters of the target area, and mine the constitutive relationship through graph neural network; The physical association graph is based on geological prior knowledge and uses a Bayesian network to model the conditional dependency between the first remote sensing features; The second remote sensing feature includes the domain-adapted feature vector enhanced by the target region and the physical association graph structure.

4. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 1 is characterized by: The spatiotemporal evolution decision module passes the second remote sensing feature through the automatic recognition model, and the three-layer processing chain of the automatic recognition model satisfies: The size of the 3D dilated convolution kernel in the spatiotemporal feature extraction layer is positively correlated with the node degree of the physical association graph; The node state vector dimension of the hazard body spatiotemporal graph corresponding to the dynamic graph evolution layer is equal to the hidden layer dimension of the LSTM unit in the critical recognition layer; The output gradient of the critical recognition layer is back-propagated to the expansion rate adjustment coefficient of the spatiotemporal feature extraction layer.

5. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 5 is characterized in that: The spatiotemporal evolution decision module, the weight matrix WM corresponding to the three-dimensional expansion convolution kernel ck Dynamic calculation is performed based on the node degree D in the physical association graph, which is specifically expressed as: Among them, D i Expressed as the degree of node i in the physical association graph, ∑D j It is expressed as the sum of the degrees of all nodes in the physical association graph, It represents the normalization of node degrees through the Softmax function, and WM0 represents the weight matrix corresponding to the pre-set three-dimensional dilated convolution kernel.

6. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 5 is characterized by: The spatiotemporal evolution decision module calculates the priority of the deformation propagation path through the structural attention head in the bimodal graph attention network, which is specifically expressed as: Among them, p (ti,tj) It is represented as the priority of the deformation propagation path from node ti to node tj, and p is represented as a learnable weight vector that measures the similarity of node attributes corresponding to the spatiotemporal graph structure of the disaster body. T Represented as the transpose operation on p, Wh ti and Wh tj Expressed as the embedding representation of node ti and node tj, [Wh i ‖Wh i ] is represented as the concatenated vector of the embedding representation of node ti and node tj, LeakyReLU (p T [Wh i ‖Wh j ]) is expressed as p T [Wh i ‖Wh j ] Apply the LeakyReLU activation function, exp(LeakyReLU(p T [Wh ti ‖Eh tj ])) represents the LeakyReLU(p T [Wh i ‖Wh j ])The exponential function is applied to the activation result, and tk is represented as all neighbor nodes of node ti.

7. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 5 is characterized by: The operation method of the spatiotemporal evolution decision module and the critical identification layer specifically includes: The cumulative effect characteristics of the disaster body deformation rate are extracted through the time series of node states in the disaster body space-time graph; The mutation synergy characteristics are captured through the flattened sequence of the adjacency matrix and node characteristics of the disaster body spatiotemporal graph; The fusion weight is automatically adjusted according to the warning level at time t-1.

8. The automatic geological disaster identification system based on multi-source remote sensing data according to claim 1 is characterized by: The multi-level early warning decision module, the graded early warning signal is a disaster risk level instruction generated based on the quantitative analysis results of the probabilistic risk assessment matrix; the disaster risk level instruction includes four levels of early warning levels, including level I warning, level II warning, level III warning and level IV warning, where level I warning indicates that a geological disaster is about to occur or that initial damage characteristics have appeared; A Level II warning indicates that the disaster body has entered a rapid deformation stage and there is a possibility of short-term instability; a Level III warning indicates that the geological disaster risk level has increased significantly and the local area has entered a deformation accumulation stage; a Level IV warning indicates that there are potential risks in the target area.

9. A method for automatically identifying geological hazards based on multi-source remote sensing data, using a system for automatically identifying geological hazards based on multi-source remote sensing data as claimed in any one of claims 1 to 8, characterized in that: include: S1: Receive and perform spatial registration and radiometric correction on first remote sensing data to generate a first remote sensing feature; S2: The first remote sensing feature is subjected to a cross-domain physical fusion mechanism to output a second remote sensing feature, wherein the cross-domain physical fusion mechanism includes a domain adversarial network mechanism and a physical constraint mechanism; S201: A deep domain adversarial network composed of a feature extractor, a label classifier, and a domain classifier generates a virtual feature corresponding to the first remote sensing feature and outputs a domain-adapted feature vector enhanced for the target region; S202: constructing a physical association graph between the first remote sensing feature and the geological parameters corresponding to the target area, wherein the physical association graph is based on geological prior knowledge, uses a Bayesian network to model the conditional dependency relationship between multi-source data, and uses a graph neural network to mine the constitutive relationship between the modes, and outputs a physical association graph structure representing the modes; S3: Obtain a third remote sensing feature by applying the second remote sensing feature to an automatic recognition model, wherein the automatic recognition model is a three-layer processing chain consisting of a spatiotemporal feature extraction layer, a dynamic graph evolution layer, and a critical recognition layer; S301: Using three-dimensional dilated convolution to capture the multi-scale spatiotemporal features in the second remote sensing feature, and fusing the multi-scale spatiotemporal features according to the dilation coefficient; S302: Construct a spatiotemporal graph structure of the disaster body corresponding to the multi-scale spatiotemporal features, identify the deformation propagation path through the bimodal graph attention network, and output the spatiotemporal evolution characteristics of the disaster body; S303: Construct the disaster body dynamic process based on the LSTM-Transformer hybrid framework to output the disaster body evolution label; S4: constructing a geological disaster assessment matrix based on the third remote sensing feature and generating a graded warning signal; S5: receiving the first remote sensing feature of the target area in real time, and performing an optimization control operation, wherein the optimization control operation is used to optimize and adjust the cross-domain physical fusion mechanism and the automatic recognition model; S6: Mapping the graded warning signals into a three-dimensional geological model map to display an explanation and analysis of the causes of geological disasters.

10. The method for automatic identification of geological hazards based on multi-source remote sensing data according to claim 9, characterized in that: The step S5, performing the optimization control operation, specifically includes: When the KL divergence between the first remote sensing feature of the target area and the historical data exceeds a preset threshold, updating the domain adaptation network parameters of the domain adversarial network mechanism through online knowledge distillation; If the domain adaptation network parameters of the domain adversarial network mechanism are triggered, the weight distribution of the three-dimensional dilated convolution kernel is adjusted synchronously.

Images