Highway Disaster Detection Method and System Based on Cooperative Unmanned Inspection of Air and Ground

Abstract

The present invention provides a highway disaster detection method and system based on air-ground collaborative unmanned inspection. The method and system obtain a multi-source highway monitoring data set, perform spatiotemporal alignment processing on the multi-source highway monitoring data set, generate a highway monitoring data set with integrated spatiotemporal labels, and then perform disaster feature extraction processing on the spatiotemporal label-integrated highway monitoring data set based on a preset multimodal disaster recognition model to obtain a highway disaster feature set. A disaster detection result set is generated based on the highway disaster feature set, the disaster detection result set comprising multiple disaster detection units, and a patrol optimization strategy set is generated based on the disaster detection result set. The present invention comprehensively improves the automation level and response efficiency of highway disaster detection without relying on human experience, and solves the problems of high misjudgment rates, large monitoring blind spots, and rigid resource allocation caused by data fragmentation and spatiotemporal asynchrony in traditional methods.

Description

Technical Field

[0001] The present invention relates to the fields of data processing and highway detection, and in particular to a highway disaster detection method and system based on air-ground collaborative unmanned inspection. Background Art

[0002] With the continuous expansion of highway transportation infrastructure, highway disaster detection technology has become a core requirement in road operations and maintenance. Highway disaster detection refers to the use of monitoring methods to identify disaster risks such as landslides, pavement cracks, and bridge structural damage along highways. Current mainstream methods primarily utilize aerial remote sensing image analysis or ground-based sensor data acquisition. These methods independently process surface deformation characteristics in optical images or abnormal signals in structural vibration sensor data to assess environmental disaster or structural damage risks, respectively. These methods, due to the disconnected spatiotemporal relationship between air-ground data and the lack of correlation between environmental and structural characteristics, struggle to capture the coupled evolutionary mechanisms of disasters. For example, the synergistic effects of slow surface deformation and bridge stress anomalies cannot be effectively identified, leading to misjudgment of disaster type and underestimation of impact scope. Furthermore, static inspection routes and fixed parameter configurations struggle to adapt to the dynamic spread of disasters, resulting in insufficient coverage of monitoring blind spots and delayed response in high-risk areas. This limits the accuracy and timeliness of highway disaster detection. Summary of the Invention

[0003] The present invention provides a highway disaster detection method and system based on air-ground collaborative unmanned inspection.

[0004] In a first aspect, an embodiment of the present invention provides a highway disaster detection method based on air-ground collaborative unmanned inspection, comprising: obtaining a multi-source highway monitoring data set, the multi-source highway monitoring data set comprising a multi-dimensional remote sensing image data set collected by aerial inspection equipment and a multi-dimensional structural sensing data set collected by ground inspection equipment; the multi-dimensional remote sensing image data set comprises image area coverage information at different timestamps, and the multi-dimensional structural sensing data set comprises structural vibration response information at different spatial locations; performing spatiotemporal alignment processing on the multi-source highway monitoring data set to generate a highway monitoring data set with fused spatiotemporal labels; performing disaster feature extraction processing on the highway monitoring data set with fused spatiotemporal labels based on a preset multimodal disaster recognition model to obtain a highway disaster feature set; the highway disaster feature set comprises environmental correlation features and structural damage features; generating a disaster detection result set based on the highway disaster feature set, the disaster detection result set comprising multiple disaster detection units, each disaster detection unit corresponding to a disaster type identifier and a disaster impact range identifier for a highway area; and generating an inspection optimization strategy set based on the disaster detection result set. In a second aspect, an embodiment of the present invention provides a highway disaster detection system, comprising: a memory storing a computer program; and a processor for loading the computer program to implement the above-mentioned highway disaster detection method based on air-ground collaborative unmanned inspection.

[0005] The highway disaster detection method based on air-ground collaborative unmanned inspection provided by the present invention realizes the deep fusion of multi-dimensional remote sensing image data and structural sensing data through the collaborative collection of multi-source data of aerial inspection equipment and ground inspection equipment. It combines the spatiotemporal alignment processing to eliminate the spatiotemporal benchmark differences of the data of air-ground equipment, and ensures the accurate mapping of environmental correlation characteristics and structural damage characteristics in a unified spatiotemporal framework. It further performs cross-modal feature coupling analysis on the fused data based on the multimodal disaster identification model, effectively correlating macro-environmental changes such as surface deformation and vegetation anomalies with micro-structural responses such as vibration frequency offset and stress distribution anomalies, breaking through the traditional The data source or single-modal analysis method has limitations in characterizing the complex disaster coupling mechanism, significantly improving the early identification accuracy and type differentiation ability of disasters such as landslides, pavement cracks and bridge structure failures; at the same time, through the closed-loop inspection optimization strategy driven by disaster detection results, the collaborative path and monitoring parameters of air-ground equipment are dynamically adjusted to achieve refined monitoring of high-risk areas and real-time tracking of disaster spread trends. Without relying on human experience intervention, the automation level and response efficiency of highway disaster detection are comprehensively improved, solving the problems of high misjudgment rate, large monitoring blind spots and rigid resource allocation caused by data fragmentation and spatiotemporal asynchrony in traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1This is a flow chart of a highway disaster detection method based on air-ground collaborative unmanned inspection provided by an embodiment of the present invention.

[0007] Figure 2 The figure is a schematic diagram of the composition of a highway disaster detection system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0008] See also Figure 1 , which is a flowchart of a highway disaster detection method based on air-ground collaborative unmanned inspection provided by an embodiment of the present invention. The highway disaster detection method based on air-ground collaborative unmanned inspection can be executed by a highway disaster detection system and may include the following steps: Step S100: Acquire a multi-source highway monitoring data set, the multi-source highway monitoring data set including a multi-dimensional remote sensing image data set collected by aerial inspection equipment and a multi-dimensional structural sensing data set collected by ground inspection equipment; wherein the multi-dimensional remote sensing image data set includes image area coverage information of different time stamps, and the multi-dimensional structural sensing data set includes structural vibration response information of different spatial positioning.

[0009] In embodiments of the present invention, aerial inspection equipment, such as drones or satellites equipped with various remote sensing imaging devices, can conduct large-scale aerial monitoring of highways and their surrounding areas. Timestamps record the specific time of image acquisition, and images captured at different times can reflect changes in the highway and its surrounding environment over time. Image coverage information specifies the geographic scope of each image. By analyzing images of different regions, a comprehensive understanding of the topography and surface conditions along the highway can be obtained. Ground inspection equipment, such as various sensors installed on or around highway structures, such as accelerometers and strain sensors, can be obtained. These sensors can monitor the physical condition of highway structures in real time. The multi-dimensional structural sensing data collected contains structural vibration response information at different spatial locations. Different spatial locations refer to sensors installed at different locations on the highway, such as bridge piers and beams, or different sections of the road surface. Structural vibration response information describes the vibrations generated by highway structures when subjected to various external forces (such as vehicle traffic and wind), including parameters such as frequency, amplitude, and phase.

[0010] Step S200: Perform spatiotemporal alignment on the multi-source highway monitoring data set to generate a spatiotemporal label-fused highway monitoring data set. The spatiotemporal alignment process involves performing temporal synchronization and spatial coordinate conversion on the multi-dimensional remote sensing image data set and the multi-dimensional structural sensing data set. After acquiring the multi-source highway monitoring data set, due to potential temporal and spatial inconsistencies between the data from different sources, spatiotemporal alignment is required to generate a spatiotemporal label-fused highway monitoring data set for more effective comprehensive analysis. The spatiotemporal alignment process primarily involves two key steps: temporal synchronization and spatial coordinate conversion. Temporal synchronization ensures temporal consistency between the multi-dimensional remote sensing image data set and the multi-dimensional structural sensing data set. This is because the acquisition times of aerial inspection equipment and ground inspection equipment may be out of sync. For example, aerial inspection equipment may concentrate on collecting image data during a certain period of time, while ground inspection equipment continuously collects structural vibration response information. Through temporal synchronization, the overlapping intervals between the two data sets can be identified, allowing for effective correlation and analysis of the data within these overlapping intervals. Spatial coordinate transformation is used to resolve discrepancies between the spatial coordinate systems of two sets of data. Multidimensional remote sensing image datasets typically use geographic coordinate systems (such as longitude and latitude) to represent image locations, while multidimensional structural sensing datasets may use relative coordinate systems (such as a coordinate system based on a fixed point on the highway). To spatially unify these two sets of data, their coordinates need to be transformed into the same spatial coordinate system. Through spatiotemporal alignment, the resulting highway monitoring dataset, fused with spatiotemporal labels, will have a unified temporal and spatial reference. This allows for comprehensive analysis of data from different sources within the same spatiotemporal framework, improving the accuracy and reliability of highway disaster detection.

[0011] As an embodiment, step S200 may specifically include the following steps: Step S210: Extracting first spatiotemporal reference information from the multi-dimensional remote sensing image data set. The first spatiotemporal reference information includes an image acquisition time series and an image geographic coordinate range. The image acquisition time series records the acquisition time of each remote sensing image. This time series can be used to understand the acquisition sequence and time intervals of different images. For example, if changes in surface deformation are observed in images acquired at different times, the time period during which these changes occurred can be inferred by combining the image acquisition time series. The image geographic coordinate range specifies the geographic area covered by each image and can be expressed in latitude and longitude. The geographic coordinate range can accurately locate the actual geographic location corresponding to the image, facilitating accurate association between remote sensing images and actual ground conditions. For example, when analyzing vegetation cover around a highway, the image geographic coordinate range can be used to determine which sections of the road have been affected by vegetation. When extracting the first spatiotemporal reference information, for example, for the image acquisition time series, the metadata of the remote sensing image data includes information about the image acquisition time. The acquisition time can be extracted from the image data file and sorted in chronological order to generate an image acquisition time series. The image geographic coordinate range can be obtained through remote sensing image processing software (such as ENVI, ERDAS, etc.).

[0012] Step S220: Extracting second spatiotemporal reference information from the multi-dimensional structural sensor data set. The second spatiotemporal reference information includes a sensor timestamp sequence and a sensor location coordinate sequence. The sensor timestamp sequence records the specific time when each sensor on the ground inspection equipment collected data. Similar to an image acquisition time series, it is accurate to a specific moment. The sensor timestamp sequence can be used to determine the acquisition time corresponding to each structural vibration response data point. The sensor location coordinate sequence specifies the specific spatial location of each sensor and can be represented using either relative or geographic coordinates. If relative coordinates are used, a coordinate system can be established with a fixed point on the highway as the origin; if geographic coordinates are used, the sensor location can be represented using longitude and latitude. The sensor location coordinate sequence helps associate structural vibration response data with a specific geographic location. For example, if abnormal vibration data collected by a sensor is detected, the sensor location can be quickly located using the sensor location coordinate sequence, allowing further analysis of whether there are any problems with the highway structure at that location. When extracting the second spatiotemporal reference information, the ground inspection equipment adds timestamp information to each data point during data collection. This timestamp information can be read out by a data logger and arranged in chronological order to generate a sensor timestamp sequence. For the sensor location coordinate sequence, if a GPS positioning device is used during sensor installation, the sensor's geographic coordinates can be directly obtained from the GPS device. If relative coordinates are used, the relative coordinates can be calculated by measuring the sensor's distance and direction relative to the origin. The location coordinates of all sensors are then arranged in order to generate a sensor location coordinate sequence.

[0013] Step S230: Time window matching is performed on the first spatiotemporal reference information and the second spatiotemporal reference information to determine the temporal overlap between the multidimensional remote sensing image data set and the multidimensional structural sensing data set. Time window matching ensures that the remote sensing image data and structural sensing data used in subsequent analysis were acquired within the same time period, thereby improving the accuracy of data association analysis. During the time window matching process, an appropriate time window size is first determined based on the image acquisition time series in the first spatiotemporal reference information and the sensor timestamp series in the second spatiotemporal reference information. The time window size can be determined based on actual needs and data characteristics. For example, if the data has a high temporal resolution, a smaller time window can be selected; if the data has a large temporal span, a larger time window can be selected. The time window is then slid across the image acquisition time series and the sensor timestamp series, and the intersection of the two time series within each time window is compared. During the comparison process, an intersection algorithm can be used to calculate the overlap between the two time series within each time window. Specifically, if the time length of the overlap exceeds a preset threshold, a valid temporal overlap interval is considered to exist within the time window. By continuously sliding the time window and comparing them, all the time overlapping intervals between the multi-dimensional remote sensing image data set and the multi-dimensional structure sensing data set can be determined.

[0014] Step S240: Data slicing is performed on the multidimensional remote sensing image dataset and the multidimensional structural sensing data dataset based on the temporal overlap intervals to generate time-synchronized sub-image datasets and sub-sensor datasets. The purpose of data slicing is to segment the original multidimensional data according to the temporal overlap intervals, ensuring temporal consistency of the segmented sub-data sets, facilitating subsequent analysis and processing. Specifically, for the multidimensional remote sensing image dataset, image data acquired within the temporal overlap interval is filtered out. The image acquisition time sequence is used to determine whether each image falls within the temporal overlap interval. If so, the image data is included in the sub-image dataset. Similarly, for the multidimensional structural sensing data, structural vibration response data acquired within the temporal overlap interval is filtered out. The sensor timestamp sequence is used to determine whether each data point falls within the temporal overlap interval. If so, the data point is included in the sub-sensor dataset. The time-synchronized sub-image datasets and sub-sensor datasets generated through data slicing ensure temporal consistency between the remote sensing image data and the structural sensing data.

[0015] Step S250: Perform spatial coordinate conversion on the sub-image dataset and the sub-sensor dataset, mapping the image geographic coordinate range to the spatial coordinate system corresponding to the sensor positioning coordinate sequence, thereby generating a highway monitoring dataset with a unified spatial reference and fused spatiotemporal labels. After completing the data slicing process and obtaining the time-synchronized sub-image dataset and sub-sensor dataset, perform spatial coordinate conversion on the sub-image dataset and sub-sensor dataset, mapping the image geographic coordinate range to the spatial coordinate system corresponding to the sensor positioning coordinate sequence, thereby generating a highway monitoring dataset with a unified spatial reference and fused spatiotemporal labels. Because multi-dimensional remote sensing image datasets use a geographic coordinate system (e.g., longitude and latitude) to represent image locations, while multi-dimensional structural sensor datasets may use a relative coordinate system (e.g., a coordinate system established with a fixed point on the highway as the origin), these two coordinate systems differ and require conversion.

[0016] The specific implementation of spatial coordinate transformation can be achieved using a coordinate transformation model. For example, a seven-parameter transformation model includes three translation parameters (translations in the X, Y, and Z directions), three rotation parameters (rotation angles around the X, Y, and Z axes), and a scale parameter. First, a set number of control points are selected, each with accurate coordinate values in both the geographic coordinate system and the relative coordinate system. Then, an error equation is established based on the coordinate values of these control points. The error equation is solved using the least squares method to obtain seven transformation parameters. Once the transformation parameters are obtained, the geographic coordinates of each image point in the sub-image dataset are converted to the coordinates in the spatial coordinate system corresponding to the sensor positioning coordinate sequence using the transformation parameters. Similarly, the coordinates of each data point in the sub-sensor dataset are already in the target coordinate system, eliminating the need for further transformation. Through spatial coordinate transformation, the image geographic coordinate range is mapped to the spatial coordinate system corresponding to the sensor positioning coordinate sequence, ensuring a unified spatial reference between the sub-image dataset and the sub-sensor dataset.

[0017] Step S300: Based on a pre-set multimodal disaster recognition model, disaster feature extraction is performed on the spatiotemporally labeled highway monitoring data set to obtain a highway disaster feature set. This feature set includes environmentally relevant features and structural damage features. The pre-set multimodal disaster recognition model combines the characteristics of remote sensing imagery and structural sensor data to extract features relevant to highway disasters from the spatiotemporally labeled highway monitoring data set. This model can utilize a deep learning model architecture, such as a combination of a convolutional neural network (CNN) and a recurrent neural network (RNN). The CNN processes the remote sensing imagery data and extracts spatial features, while the RNN processes temporal features from the structural sensor data. During the disaster feature extraction process, the spatiotemporally labeled highway monitoring data set is input into the pre-set multimodal disaster recognition model. The model performs hierarchical processing on the input data to extract environmentally relevant features and structural damage features. Environmentally relevant features reflect the impact of environmental factors surrounding the highway on highway disasters, including surface deformation, abnormal vegetation cover, and changes in water distribution. Structural damage characteristics reflect the health status of the highway structure itself, including vibration frequency deviation characteristics, stress distribution abnormality characteristics and material fatigue accumulation characteristics.

[0018] As an embodiment, step S300 may specifically include the following steps: Step S310: Performing environmental feature extraction processing on the sub-image datasets within the spatiotemporal-labeled highway monitoring dataset to generate an environmentally relevant feature set; the environmentally relevant feature set includes surface deformation features, vegetation cover anomaly features, and water body distribution change features. The environmentally relevant feature set includes information such as surface deformation features, vegetation cover anomaly features, and water body distribution change features. The sub-image datasets are time-synchronized subsets of remote sensing image data obtained through spatiotemporal alignment, and include image data from different bands, such as visible light, infrared, and radar. Performing environmental feature extraction processing on these image data allows information about the highway's surrounding environment to be obtained from different perspectives. Surface deformation feature extraction can be achieved by analyzing elevation information and crack distribution within the sub-image datasets. For example, digital elevation model (DEM) technology can be used to process remote sensing image data from different times to generate high-precision DEM data. By comparing DEM data from different times, the surface elevation change rate can be calculated, thereby reflecting the vertical deformation of the surface. Image processing algorithms, such as edge detection and morphological processing, can be used to identify surface cracks in images and calculate their distribution characteristics, such as length, width, and density. Extracting abnormal vegetation cover features can be done with infrared imagery data. Infrared imagery can reflect the thermal radiation characteristics of vegetation. Healthy vegetation and vegetation affected by pests, diseases, fires, and other factors exhibit significant differences in thermal radiation intensity. By analyzing thermal radiation intensity on subsets of infrared data, such as calculating statistical parameters like mean thermal radiation intensity and standard deviation, areas of surface temperature anomalies and heat source distribution can be identified, thereby determining whether vegetation cover is abnormal. Extracting features of water distribution changes primarily relies on radar imagery data. Radar images have the ability to penetrate clouds and fog, clearly reflecting the distribution of surface water bodies. Using interferometric synthetic aperture radar (InSAR) technology, by processing subsets of radar data collected at different times, vertical and horizontal displacement vectors of the surface can be measured, thereby analyzing changes in water distribution.

[0019] As an embodiment, step S310 may specifically include the following steps: Step S311: Performing multispectral band decomposition on the sub-image data sets to generate visible light band data subsets, infrared band data subsets, and radar band data subsets. Multispectral band decomposition, based on the different physical properties and information expression capabilities of image data in different bands, separates the sub-image data sets by band, facilitating more targeted analysis and feature extraction of data from different bands. The sub-image data sets are obtained by temporal and spatial alignment of multi-dimensional remote sensing image data collected by aerial inspection equipment and contain image information from multiple bands. Multispectral band decomposition can be performed using, for example, principal component analysis (PCA) or independent component analysis (ICA). For example, PCA performs a linear transformation on the original multispectral image data, converting it to a new coordinate system. This transforms the data components in the new coordinate system into independent components and sorts them according to their variance. Specifically, the original multispectral image data is normalized so that the data in each band has the same mean and variance. Next, the covariance matrix of the standardized data is calculated. Next, the covariance matrix is subjected to eigenvalue decomposition to obtain eigenvalues and eigenvectors. The eigenvectors are sorted according to their eigenvalues, and the first N eigenvectors with the largest eigenvalues are selected as principal components, with the value of N being unrestricted. Finally, the original multispectral image data is projected onto these principal components to obtain the principal component image data. Based on the distribution of different bands within the principal component image data, it is separated into visible light band data subsets, infrared band data subsets, and radar band data subsets.

[0020] Step S312: Surface deformation analysis is performed on the visible light band data subset to extract the surface elevation change rate and surface crack distribution characteristics. After completing the multispectral band decomposition process and obtaining the visible light band data subset, surface deformation analysis is performed on this data subset to extract the surface elevation change rate and surface crack distribution characteristics. Surface deformation is an indicator of environmental changes surrounding the highway and can be caused by natural factors (such as earthquakes and landslides) or human factors (such as construction projects and groundwater extraction). By analyzing the surface elevation change rate and surface crack distribution characteristics, potential highway disaster risks can be promptly identified.

[0021] For example, the process of analyzing surface deformation using a subset of visible light band data involves using a stereo matching algorithm to process visible light band images taken at different times to generate a digital elevation model (DEM). The principle of the stereo matching algorithm is to find corresponding points of the same feature in different images, calculate the disparity of these corresponding points, and then calculate the feature's three-dimensional coordinates based on the disparity and the camera's intrinsic and extrinsic parameters. Possible stereo matching algorithms include region-based matching, feature-based matching, and energy-optimization-based matching. After obtaining DEMs taken at different times, the surface elevation change rate can be calculated by comparing these DEM data. Specifically, for each pixel, the elevation difference at different times is calculated and then divided by the time interval to obtain the pixel's elevation change rate. The elevation change rates of all pixels are statistically analyzed, such as by calculating the mean and standard deviation, to obtain the surface elevation change rate for the entire area. Image processing algorithms, such as edge detection and morphological processing, can be used to extract the distribution characteristics of surface cracks. Edge detection algorithms can identify the edges of features in images, thereby detecting the locations of surface cracks. Possible edge detection algorithms include the Sobel operator and the Canny operator. Taking the Canny operator as an example, it involves Gaussian smoothing, gradient calculation, non-maximum suppression, and double-thresholding. First, Gaussian smoothing is performed on the visible light band image to reduce the impact of noise. Then, the image's gradient magnitude and direction are calculated. Next, non-maximum suppression is performed to retain only pixels with local maximum gradient magnitudes. Finally, double-thresholding is used to classify pixels into strong edge points, weak edge points, and non-edge points. Only strong edge points and weak edge points connected to strong edge points are retained to obtain edge information for surface cracks. After obtaining the edge information for surface cracks, morphological processing algorithms such as dilation, erosion, opening, and closing are used to refine, connect, and remove noise from the cracks, thereby obtaining a complete representation of the surface crack distribution. For example, dilation can connect small cracks, while erosion can remove noise points surrounding the cracks.

[0022] Step S313: Thermal radiation intensity analysis is performed on the infrared band data subset to extract surface temperature anomaly areas and heat source distribution characteristics. For example, radiometric calibration is performed on the infrared band image to convert the image's grayscale values into actual thermal radiation intensity values. Radiometric calibration converts the image's digital quantization value (DN value) into radiometric brightness or temperature values based on the infrared sensor's calibration coefficient. Possible radiometric calibration methods include absolute and relative calibration. Absolute calibration measures the radiation intensity of a known radiation source to establish a relationship between the image's grayscale value and the actual radiation intensity. Relative calibration compares different areas within the same image to establish a relative radiation intensity relationship. After radiometric calibration is completed, the thermal radiation intensity values are statistically analyzed to calculate the image's average thermal radiation intensity and standard deviation. A threshold range is then set based on the average thermal radiation intensity and standard deviation. Areas with thermal radiation intensity values outside this threshold range are defined as surface temperature anomaly areas. For example, if the thermal radiation intensity value in a region is higher than the average thermal radiation intensity plus a preset multiple of the standard deviation, the region is considered to be abnormally high temperature. If it is lower than the average thermal radiation intensity minus a preset multiple of the standard deviation, the region is considered to be abnormally low temperature. Cluster analysis algorithms, such as the K-means clustering algorithm, can be used to extract heat source distribution characteristics. Specifically, K data points are randomly selected as initial cluster centers. Then, the distance from each data point to each cluster center is calculated, and the data point is assigned to the cluster with the closest cluster center. Next, the center position of each cluster is recalculated. These steps are repeated until the cluster center no longer changes or the preset number of iterations is reached. K-means clustering analysis is performed on the thermal radiation intensity values in the infrared band data subset, resulting in different clustering results. Each cluster represents a heat source area. By analyzing the location, size, and thermal radiation intensity characteristics of the cluster, the heat source distribution characteristics can be obtained. For example, parameters such as the area, average thermal radiation intensity, and maximum thermal radiation intensity of each heat source area can be calculated to understand the distribution and intensity of the heat source.

[0023] Step S314: Interferometric synthetic aperture radar (ISAR) processing is performed on the radar band data subset to extract the vertical and horizontal displacement vectors of the ground surface. The InSAR processing process for the radar band data subset, for example, involves first selecting two or more radar band images taken at different times. These images must cover the same area and have a certain degree of overlap. These images are then registered to precisely align them spatially. This registration method can employ feature-based registration algorithms, such as the Scale-Invariant Feature Transform (SIFT) algorithm or the Speeded Up Robust Features (SURF) algorithm. These algorithms extract feature points from the images and achieve image registration by matching these feature points. After image registration is complete, interferometric processing is performed to generate an interferogram. The interferogram is generated by complex multiplying the two registered radar images and contains phase information about the ground surface. This phase information is related to factors such as surface deformation, topography, and atmospheric delay. To isolate surface deformation information, the interferogram requires phase unwrapping. Phase unwrapping is to restore the phase value in the interferogram from the range of [-π,π] to the true phase value. Possible phase unwrapping algorithms include the minimum cost flow algorithm, the branch cutting method, etc. After completing the phase unwrapping, atmospheric correction and terrain correction are performed. The purpose of atmospheric correction is to eliminate the influence of atmospheric delay on the phase. Possible atmospheric correction methods include correction methods based on meteorological data and statistical analysis methods based on interferograms. The purpose of terrain correction is to eliminate the influence of terrain factors on the phase. Digital elevation model (DEM) is usually used for terrain correction. After atmospheric correction and terrain correction, a phase map containing only surface deformation information is obtained. According to the relationship between phase and surface displacement, the phase map can be converted into a surface vertical displacement map. The specific conversion formula can be: ,in is the vertical displacement of the ground surface, is the radar wavelength, is the phase change. To extract the horizontal surface displacement vector, multi-orbit InSAR data can be used or combined with other measurement methods. For example, radar images from different orbits can be processed using InSAR to obtain surface displacement information in different directions. The horizontal surface displacement vector can then be derived through vector synthesis.

[0024] Step S315: Integrate the surface elevation change rate, surface crack distribution characteristics, surface temperature anomaly areas, heat source distribution characteristics, surface vertical displacement, and horizontal displacement vectors to generate a set of environmentally relevant features. The process of integrating these features, for example, involves normalizing numerical features such as the surface elevation change rate, surface vertical displacement, and horizontal displacement vectors so that they have the same scale range. Normalization can employ a minimum-maximum normalization method, where the minimum value of each feature is subtracted from the feature's value, and then divided by the difference between the maximum and minimum values of the feature. For non-numerical features such as surface crack distribution characteristics, surface temperature anomaly areas, and heat source distribution characteristics, encoding can be used to convert them into numerical features. For example, for surface crack distribution characteristics, information such as crack length, width, and density can be quantized, and then these quantized values can be combined into a feature vector. For surface temperature anomaly areas and heat source distribution characteristics, information such as the area, location, and intensity of the anomaly area can be encoded to generate the corresponding feature vector. After normalization and encoding of the features, these feature vectors are concatenated to generate a comprehensive feature vector. This comprehensive feature vector contains all environmental feature information, including the surface elevation change rate, surface crack distribution characteristics, surface temperature anomaly areas, heat source distribution characteristics, and surface vertical and horizontal displacement vectors. Finally, this comprehensive feature vector is used as an element of the environmental correlation feature set. The above steps are repeated to comprehensively process all extracted environmental features and generate a complete environmental correlation feature set.

[0025] Step S320: Structural feature extraction is performed on the sub-sensor data set within the spatiotemporal label-fused highway monitoring data set to generate a structural damage feature set. This structural damage feature set includes vibration frequency offset features, stress distribution anomaly features, and material fatigue accumulation features. The sub-sensor data set is a time-synchronized subset of structural sensor data obtained through spatiotemporal alignment, containing information such as the vibration response and stress distribution of the highway structure. Structural feature extraction from this data provides insights into the mechanical properties and damage status of highway structures from various perspectives.

[0026] Extracting vibration frequency shift features is primarily accomplished by performing frequency domain conversion on the structural vibration response information. The Fast Fourier Transform (FFT) algorithm can be used to convert the time-domain vibration response signal into a frequency-domain vibration spectrum feature set. The FFT algorithm can convert a time-domain signal of length N into a frequency-domain signal of length N, with a computational complexity of O(NlogN). After obtaining the vibration spectrum feature set, the fundamental frequency component, harmonic components, and noise components are extracted, and the fundamental frequency offset and harmonic energy fraction are calculated. The fundamental frequency offset is the difference between the current fundamental frequency and the initial fundamental frequency and reflects the stiffness change of the highway structure. The harmonic energy fraction is the proportion of the harmonic component energy in the total energy and reflects the nonlinear vibration characteristics of the highway structure. Extracting stress distribution anomaly features can be accomplished by performing spatial interpolation on the stress distribution information in the sub-sensor data sets. Spatial interpolation is a method of estimating the values of unknown points based on the values of known points. Possible spatial interpolation methods include inverse distance weighted interpolation and kriging interpolation. Taking the inverse distance weighted interpolation method as an example, the value of the unknown point is determined based on the distance between the unknown point and the known point. The closer the distance, the greater the influence of the known point on the unknown point. Through spatial interpolation processing, a stress distribution cloud map can be generated, from which the maximum stress value and stress gradient change rate can be extracted. The maximum stress value reflects the maximum stress level borne by the highway structure, and the stress gradient change rate reflects the spatial variation of stress. The determination of the fatigue accumulation characteristics of the material requires comprehensive consideration of parameters such as the fundamental frequency offset, the proportion of harmonic energy, the maximum stress value and the stress gradient change rate. A material fatigue accumulation model can be established to evaluate the fatigue accumulation degree of the highway structure material based on the changes in these parameters. For example, the Miner linear cumulative damage criterion can be used. This criterion believes that the fatigue damage of the material is the accumulation of damage caused by each stress cycle. When the damage accumulation reaches a certain level, the material will suffer fatigue failure.

[0027] As an implementation method, step S320 may specifically include the following steps: Step S321: Perform frequency domain conversion processing on the structural vibration response information in the sub-sensor data set to generate a vibration spectrum feature set. The structural vibration response information is the vibration signal generated by the highway structure when it is subjected to external force and collected by the ground inspection equipment, and exists in the form of a time domain signal. The time domain signal reflects the change of the vibration signal over time, but it is difficult to directly obtain the frequency information of the vibration from it. The frequency domain conversion processing can convert the time domain signal into a frequency domain signal, thereby clearly showing the frequency component and energy distribution of the vibration signal, and providing a basis for the subsequent extraction of structural damage characteristics such as vibration frequency offset characteristics. The frequency domain conversion processing can use the fast Fourier transform (FFT) algorithm to reduce the computational complexity of DFT from O(N) to O(N). 2) is reduced to O(NlogN), where N is the length of the signal. Specifically, the structural vibration response information is discretized according to the sampling frequency to obtain a discrete time domain signal sequence. Then, the time domain signal sequence is zero-padded so that its length is an integer power of 2 to improve the computational efficiency of the FFT algorithm. Next, the zero-padded time domain signal sequence is input into the FFT algorithm for calculation to obtain a frequency domain signal sequence. Finally, the amplitude spectrum of the frequency domain signal sequence is calculated to obtain a vibration spectrum feature set. The vibration spectrum feature set contains the various frequency components of the structural vibration response signal and their corresponding energy values. By analyzing the vibration spectrum feature set, the vibration characteristics of the highway structure, such as the fundamental frequency and harmonic frequency, can be understood. The fundamental frequency is the main frequency component of the structural vibration, which is related to the stiffness and mass of the structure. When the highway structure is damaged, its stiffness will change, causing the fundamental frequency to shift. Harmonic frequencies are integer multiples of the fundamental frequency, and they reflect the nonlinear vibration characteristics of the structure. By monitoring the changes in the fundamental frequency and harmonic frequencies, potential damage to the highway structure can be discovered in a timely manner.

[0028] Step S322: Extract the fundamental frequency component, harmonic components, and noise components from the vibration spectrum feature set, and calculate the fundamental frequency offset and harmonic energy ratio. The fundamental frequency component is the frequency component with the highest energy in the vibration spectrum and represents the main vibration frequency of the highway structure. Harmonic components are frequency components that are integer multiples of the fundamental frequency and reflect the nonlinear vibration characteristics of the structure. Noise components are frequency components in the vibration spectrum other than the fundamental frequency component and harmonic components. They are caused by factors such as environmental interference and measurement errors.

[0029] The process of extracting the fundamental frequency component, harmonic components, and noise components, for example, involves performing peak detection on the vibration spectrum feature set to identify peak points in the spectrum. The frequencies corresponding to the peak points are likely fundamental frequencies or harmonic frequencies. Then, based on the frequencies and energy values of the peak points, it is determined which peak points are fundamental frequency components and which are harmonic components. The frequency corresponding to the peak point with the highest energy is the fundamental frequency, while the frequencies corresponding to other peak points that are integer multiples of the fundamental frequency are harmonic components. Finally, the frequency components in the spectrum other than the fundamental frequency and harmonic components are considered noise components. The fundamental frequency offset is the difference between the current fundamental frequency and the initial fundamental frequency. The initial fundamental frequency is the fundamental frequency of the highway structure in its healthy state and can be measured within a period of time after construction or repair. The fundamental frequency offset is calculated by subtracting the initial fundamental frequency from the current fundamental frequency. The magnitude of the fundamental frequency offset reflects changes in the stiffness of the highway structure. A large fundamental frequency offset indicates possible damage to the highway structure. The harmonic energy fraction is the proportion of the energy of the harmonic component to the total energy. The harmonic energy percentage is calculated by first calculating the total energy of the harmonic components by adding the energy values of all harmonic components. Then, the total energy is calculated by adding the energy values of the fundamental frequency component, harmonic components, and noise components. Finally, the total energy of the harmonic components is divided by the total energy to obtain the harmonic energy percentage. Changes in the harmonic energy percentage reflect changes in the nonlinear vibration characteristics of the highway structure. When the harmonic energy percentage increases, it indicates that the highway structure may have nonlinear damage.

[0030] Step S323: Spatial interpolation is performed on the stress distribution information in the sub-sensor data set to generate a stress distribution cloud map and extract the maximum stress value and stress gradient change rate. Spatial interpolation can be performed using a variety of methods, such as inverse distance weighted interpolation and kriging interpolation. Taking inverse distance weighted interpolation as an example, the principle of this method is to determine the stress value of an unknown point based on the distance between the unknown point and the known point. The closer the distance, the greater the influence of the known point on the unknown point. Specifically, the position of the unknown point to be interpolated is determined. Then, several known points closest to the unknown point are found. Next, the weight of each known point is calculated based on the distance between the unknown point and the known point. The closer the distance, the greater the weight. Finally, the stress values of the known points are weighted averaged according to the weights to obtain the stress value of the unknown point. By performing the above interpolation process on all unknown points, continuous stress distribution data is obtained.

[0031] After obtaining continuous stress distribution data, it is visualized in the form of a cloud map, generating a stress distribution cloud map. The stress distribution cloud map intuitively displays the stress distribution within the highway structure. Different colors represent different stress levels, with darker colors indicating greater stress. The maximum stress value is extracted from the stress distribution cloud map by traversing all pixels in the cloud map and finding the pixel with the maximum stress value. The stress value corresponding to this pixel is the maximum stress value. The maximum stress value reflects the maximum stress level experienced by the highway structure. When the maximum stress value exceeds the design load-bearing capacity of the highway structure, it may cause structural damage. The stress gradient change rate is the spatial rate of stress change. To calculate the stress gradient change rate, select several adjacent pixels in the stress distribution cloud map, calculate the stress difference and distance difference between them, and then divide the stress difference by the distance difference to obtain the stress gradient change rate. Statistical analysis of the stress gradient change rate at different locations can provide an understanding of the spatial variation of stress. A large stress gradient change rate indicates rapid stress variation in that area, potentially indicating stress concentration, which increases the risk of highway structure damage.

[0032] Step S324: Based on the fundamental frequency offset, harmonic energy ratio, maximum stress value and stress gradient change rate, the vibration frequency offset feature, stress distribution abnormality feature and material fatigue accumulation feature in the structural damage feature set are determined. The vibration frequency offset feature is mainly related to the fundamental frequency offset. The fundamental frequency offset reflects the change in the stiffness of the highway structure. When the fundamental frequency offset exceeds the threshold, it indicates that the highway structure may have been damaged, resulting in a change in its stiffness. Therefore, the fundamental frequency offset can be used as an important indicator of the vibration frequency offset feature. At the same time, the harmonic energy ratio can also be used as a supplementary indicator of the vibration frequency offset feature. When the harmonic energy ratio increases, it indicates that the nonlinear vibration characteristics of the highway structure have changed, and may also indicate that the structure is damaged. By comprehensively considering the fundamental frequency offset and the harmonic energy ratio, the vibration frequency offset feature can be determined more accurately.

[0033] Abnormal stress distribution characteristics are primarily related to the maximum stress value and the stress gradient change rate. The maximum stress value reflects the maximum stress level experienced by the highway structure. When the maximum stress value exceeds the design load-bearing capacity of the highway structure, it indicates possible stress distribution anomaly. The stress gradient change rate reflects the spatial variation of stress. A large stress gradient change rate indicates rapid stress variation in that region, potentially indicating stress concentration, which is also a manifestation of abnormal stress distribution. Therefore, the maximum stress value and the stress gradient change rate can be used as important indicators of stress distribution anomaly. By setting appropriate thresholds, it can be used to determine whether the highway structure exhibits abnormal stress distribution. Determining the fatigue accumulation characteristics of a material requires comprehensive consideration of multiple parameters, including fundamental frequency offset, harmonic energy proportion, maximum stress value, and stress gradient change rate. A material fatigue accumulation model can be developed to assess the degree of fatigue accumulation of highway structure materials based on the changes in these parameters. For example, the Miner linear cumulative damage criterion can be used. This criterion states that fatigue damage to a material is the accumulation of damage caused by each stress cycle. When the damage accumulation reaches a certain level, fatigue failure will occur. When establishing a material fatigue accumulation model, parameters such as fundamental frequency offset, harmonic energy ratio, maximum stress value and stress gradient change rate can be used as input. The fatigue accumulation degree of the material can be calculated through the model and used as an indicator of the material fatigue accumulation characteristics.

[0034] Step S330: Input the environmental-related feature set and the structural damage feature set into the multimodal disaster identification model for feature fusion processing to generate a fused feature vector set. The multimodal disaster identification model comprehensively considers environmental factors surrounding the highway and the damage to the highway structure itself. Through feature fusion processing, it can organically combine environmental-related features and structural damage features to extract more representative and discriminative features, thereby improving the accuracy and reliability of highway disaster identification. The multimodal disaster identification model can adopt a deep learning model architecture, such as a combination of a convolutional neural network (CNN) and a recurrent neural network (RNN). CNNs can be used to process spatial features in the environmental-related feature set, such as surface deformation features and abnormal vegetation cover features; RNNs can be used to process temporal features in the structural damage feature set, such as vibration frequency offset features and abnormal stress distribution features.

[0035] As an embodiment, step S330 may specifically include the following steps: Step S331: Normalize each environment-related feature in the environment-related feature set to generate a standardized environment feature vector. The environment-related feature set includes various types of features, such as surface deformation features, vegetation cover anomaly features, and water body distribution change features. These features may have different value ranges and dimensions. Directly fusion of these features may result in some features having an excessively large influence while others have an insufficient influence, thereby affecting model performance. Therefore, it is necessary to normalize these features so that they have the same scale range to facilitate subsequent feature fusion. Normalization can be performed using a variety of methods, such as minimum-maximum normalization and Z-score normalization, without specific limitation. It will be understood that in other aspects of the present invention involving data calculations of different dimensions or dimensions, those skilled in the art can perform preprocessing such as data dimension elimination and dimensionality unification based on general knowledge before performing the corresponding data processing. By performing minimum-maximum normalization on each environment-related feature in the environment-related feature set, each feature value is mapped to the range [0, 1], eliminating scale differences between features. Combining all normalized environmental features into a single vector yields a standardized environmental feature vector. Each element in the standardized environmental feature vector has the same scale range, ensuring that different environmental features are treated equally in subsequent feature fusion processing, improving the model's ability to learn environmental features and the effectiveness of feature fusion.

[0036] Step S332: Normalize each structural damage feature in the structural damage feature set to generate a standardized structural feature vector. Similarly, the structural damage feature set includes various features, such as vibration frequency offset features, stress distribution anomaly features, and material fatigue accumulation features. These features may have different value ranges and dimensions, and thus require normalization to eliminate scale differences and improve feature fusion. Normalization can also employ the min-max normalization method.

[0037] Step S333: Align and concatenate the standardized environmental feature vectors and the standardized structural feature vectors based on their timestamps and spatial coordinates to generate a spatiotemporally aligned multimodal feature matrix. Timestamps and spatial coordinates are key information in a spatiotemporally labeled highway monitoring dataset. They clearly identify the specific time and geographic location corresponding to environmental and structural damage features. Aligning and concatenating by timestamps and spatial coordinates ensures consistency between environmental and structural damage features in both the temporal and spatial dimensions, enabling the multimodal disaster identification model to more accurately capture the correlation between environmental factors and structural damage. Specifically, each feature element in the standardized environmental and structural feature vectors is first matched based on their corresponding timestamps and spatial coordinates. During the matching process, it is ensured that environmental and structural damage features with the same timestamps and spatial coordinates are matched. The matched standardized environmental and structural feature vectors are then concatenated. Finally, the concatenated vectors corresponding to all different timestamp and spatial coordinate combinations are sequentially arranged to form a matrix, which is the spatiotemporally aligned multimodal feature matrix. Each row of the matrix represents a combination of environmental correlation features and structural damage features at a specific time and spatial location, and each column represents a feature dimension. By generating a spatiotemporally aligned multimodal feature matrix, a structured data input is provided for the multimodal disaster recognition model, enabling the model to more effectively perform cross-modal feature analysis and learning.

[0038] Step S334: The feature fusion layer in the multimodal disaster identification model performs cross-modal correlation analysis on the multimodal feature matrix to generate a set of fused feature vectors. Cross-modal correlation analysis includes calculating correlation weights between environmental and structural features and reducing feature dimensionality. An attention mechanism can be used to calculate correlation weights between environmental and structural features. This mechanism automatically learns the correlations between different features and assigns different weights to them, allowing the model to focus more on important features. Specifically, the spatiotemporally aligned multimodal feature matrix is first input into the attention layer. The attention layer calculates the correlation score between each feature and the other features. A higher correlation score indicates a closer correlation with the other features. Then, a weight is assigned to each feature based on the correlation score. A larger weight indicates a greater importance of the feature in the cross-modal correlation analysis. Finally, each feature is multiplied by its corresponding weight to obtain the weighted features. Feature dimensionality reduction can be performed using the principal component analysis (PCA) or linear discriminant analysis (LDA) algorithm. The principles of this algorithm have been described above and will not be elaborated here. The feature fusion layer performs cross-modal correlation analysis on the multimodal feature matrix, organically combining environmental and structural features to generate a set of fused feature vectors. Each vector in this set of fused feature vectors combines information from both environmental correlation features and structural damage characteristics, enabling a more comprehensive and accurate reflection of highway disaster characteristics.

[0039] Step S340: Disaster pattern matching is performed on the fused feature vector set to determine a highway disaster feature set that matches a disaster pattern in a preset disaster pattern library. Specifically, during disaster pattern matching, multiple historical disaster pattern feature vectors are obtained from the preset disaster pattern library. Each historical disaster pattern feature vector corresponds to a standard feature template for a disaster type. These standard feature templates are derived by analyzing and summarizing a large amount of historical disaster data and are representative and typical. A similarity score is then calculated between each fused feature vector in the fused feature vector set and each historical disaster pattern feature vector. The similarity score can be calculated using a variety of methods, such as Euclidean distance and cosine similarity, though these methods are not specifically limited. Based on the similarity score, the optimal matching disaster type corresponding to each fused feature vector is determined, and a preliminary disaster feature set is generated, including a disaster type identifier and a matching confidence score. The matching confidence score represents the degree of match between the fused feature vector and the corresponding disaster type and can be expressed as a similarity score. The higher the similarity score, the higher the matching confidence. Finally, a spatial continuity check is performed on the preliminary disaster feature set to eliminate disaster features with matching confidences below a preset threshold and spatially isolated distributions, thereby generating a highway disaster feature set. Spatial continuity verification is used to eliminate false disaster signatures caused by data noise or accidental factors. A preset threshold can be set based on actual conditions. Generally speaking, when the matching confidence level falls below this threshold, the matching result for that disaster signature is considered unreliable. Spatially isolated disaster signatures are those that are not spatially adjacent to or are distant from other disaster signatures. These signatures may be caused by local anomalies and do not necessarily represent actual highway disasters. Spatial continuity verification can improve the accuracy and reliability of the highway disaster signature set.

[0040] As an implementation method, step S340 may specifically include the following steps: Step S341: Obtain multiple historical disaster pattern feature vectors from a preset disaster pattern library, each of which corresponds to a standard feature template for a disaster type. The preset disaster pattern library is established by collecting, organizing, and analyzing a large amount of historical highway disaster data. This historical disaster data includes information on various aspects, such as the environmental characteristics and structural damage characteristics of different types of highway disasters. To construct the preset disaster pattern library, various highway disaster cases occurring in different regions and at different times are first collected, including landslides, pavement cracks, and bridge structural failures. For each disaster case, relevant environmental data (such as surface deformation, vegetation cover, and water distribution) and structural data (such as vibration frequency and stress distribution) are collected. The collected data is then preprocessed, including data cleaning and feature extraction, to remove noise and redundant information and extract key information that represents the disaster characteristics. Next, for each disaster type, the corresponding feature data is analyzed and summarized to extract representative feature patterns. These feature patterns serve as the standard feature template for that disaster type. For example, for landslide disasters, standard feature templates may include surface deformation characteristics (such as a large rate of surface elevation change and dense distribution of surface cracks), vegetation cover anomaly characteristics (such as large areas of vegetation collapse), and structural damage characteristics (such as large vibration frequency deviations of adjacent highway structures). Finally, the standard feature templates for each disaster type are converted into historical disaster pattern feature vectors and stored in a preset disaster pattern library. The historical disaster pattern feature vector is a multidimensional vector, with each dimension corresponding to a characteristic indicator. By obtaining historical disaster pattern feature vectors from the preset disaster pattern library, a reference standard can be provided for subsequent disaster pattern matching.

[0041] Step S342: Calculate the similarity score between each fused feature vector in the fused feature vector set and each historical disaster pattern feature vector. There are many methods that can be used to calculate the similarity score, such as Euclidean distance, cosine similarity, Manhattan distance, etc. For each fused feature vector in the fused feature vector set, it is necessary to perform similarity calculations with all historical disaster pattern feature vectors in the preset disaster pattern library. For example, assuming that there are m fused feature vectors in the fused feature vector set and k historical disaster pattern feature vectors in the preset disaster pattern library, a total of m×k similarity calculations are required. Through these similarity scores, the degree of matching between each fused feature vector and different disaster types can be preliminarily judged.

[0042] Step S343: Determine the optimal matching disaster type corresponding to each fused feature vector based on the similarity score, and generate a preliminary disaster feature set containing a disaster type identifier and a matching confidence. For each fused feature vector, find the maximum value from all similarity scores. The disaster type represented by the historical disaster pattern feature vector corresponding to the maximum value is the optimal matching disaster type for the fused feature vector. The matching confidence can be expressed by the similarity score, that is, the similarity score corresponding to the maximum value. The matching confidence reflects the degree of matching between the fused feature vector and the corresponding disaster type. The higher the confidence, the more reliable the matching result. The disaster type identifier and matching confidence corresponding to each fused feature vector are combined to generate a preliminary disaster feature set. Each element in the preliminary disaster feature set contains the disaster type information corresponding to a fused feature vector and the matching reliability, providing basic data for subsequent spatial continuity verification processing.

[0043] Step S344: Spatial continuity verification is performed on the preliminary disaster feature set. Disaster features with matching confidence levels below a preset threshold and spatially isolated distributions are eliminated to generate a highway disaster feature set. The preset threshold is a similarity scoring standard set based on actual conditions. When the matching confidence level is below this threshold, the matching result for that disaster feature is considered unreliable. Spatially isolated disaster features are those that are not spatially adjacent to or are located far from other disaster features. These features may be caused by local anomalies and may not necessarily represent actual highway disasters. The spatial continuity verification process, for example, involves determining the spatial location of each disaster feature based on the spatial coordinate information in the highway monitoring data set fused with spatiotemporal labels. Then, for each disaster feature, its matching confidence level is checked to see if it is below a preset threshold. If so, its spatial distribution is further checked. Neighborhood analysis can be used to determine whether the spatial distribution of a disaster feature is isolated. For example, a neighborhood radius is set, and for each disaster feature, the presence of other disaster features within this neighborhood radius with matching confidence levels above the preset threshold is checked. If it does not exist, the disaster feature is considered to be spatially isolated and is removed from the preliminary disaster feature set.

[0044] Step S400: Generate a disaster detection result set based on the highway disaster feature set. The disaster detection result set contains multiple disaster detection units, each corresponding to a disaster type identifier and a disaster impact range identifier for a highway region. This generation of the disaster detection result set clarifies the potential disaster types and impact ranges in different highway regions, providing important information for highway disaster warning and response.

[0045] As an embodiment, step S400 may specifically include the following steps: Step S410: Classifying each highway disaster feature in the highway disaster feature set into a disaster type and determining a corresponding disaster type identifier for each highway disaster feature; the disaster type identifier may be a landslide risk identifier, a pavement crack identifier, or a bridge structure failure identifier. In generating a disaster detection result set based on the highway disaster feature set, each highway disaster feature in the highway disaster feature set must first be classified into a disaster type to determine a corresponding disaster type identifier. Disaster type identifiers may be landslide risk identifiers, pavement crack identifiers, or bridge structure failure identifiers. These identifiers clearly indicate the type of disaster a highway may face, providing an important basis for subsequent disaster response and management. Disaster type classification can utilize machine learning classification algorithms, such as decision tree classification algorithms and support vector machine classification algorithms. Taking the decision tree classification algorithm as an example, a decision tree model is constructed to perform classification decisions based on the different attribute values of the highway disaster features. Specifically, a portion of highway disaster features of known disaster types is selected as training data and input into the decision tree classification algorithm for training, thereby constructing a decision tree model. The training data contains various attribute information about highway hazard characteristics (such as surface deformation and vibration frequency offset) as well as corresponding hazard type identifiers. During the training process, the decision tree classification algorithm continuously partitions nodes and learns decision rules based on the attribute values and hazard type identifiers of the training data. For example, if the value of the surface elevation change rate attribute exceeds a certain threshold, the highway hazard characteristic may be classified as a landslide risk indicator; if the value is below the threshold, the classification decision is continued based on other attributes. After training is complete, each highway hazard characteristic in the highway hazard characteristic set is input into the trained decision tree model, where it is classified according to the decision tree's decision rules to determine the hazard type identifier corresponding to each highway hazard characteristic.

[0046] Step S420: Determine the disaster impact range identifier corresponding to each highway disaster feature based on the spatial distribution information of the environmental-related features and structural damage features in the highway disaster feature set. The process of determining the disaster impact range identifier, for example, involves obtaining the spatial coordinate sequences of each surface deformation feature, vegetation cover anomaly feature, and water body distribution change feature in the environmental-related feature set to generate first spatial distribution information. The spatial coordinate sequences of these environmental-related features can be obtained using geographic coordinate information from a highway monitoring data set fused with spatiotemporal tags. For example, for surface deformation features, the latitude and longitude coordinates of their occurrence location can be obtained; for vegetation cover anomaly features, the boundary coordinates of the anomaly region can be obtained. Then, the spatial location sequences of each vibration frequency offset feature, stress distribution anomaly feature, and material fatigue accumulation feature in the structural damage feature set are obtained to generate second spatial distribution information. The spatial location sequences of structural damage features can also be obtained from a highway monitoring data set fused with spatiotemporal tags; this information can reflect the specific location of highway structural damage.

[0047] Next, the first and second spatial distribution information are spatially overlaid to generate spatially distributed overlay data containing the density distribution of environmental and structural features. This overlay process can utilize Geographic Information System (GIS) technology to overlay the spatial distribution information of environmental association features and structural damage features within the same geographic coordinate system, visually displaying the comprehensive spatial distribution of environmental factors and structural damage. The spatially distributed overlay data is then spatially gridded to generate multiple spatial grid cells, each containing both environmental and structural feature density values. The spatial grid division can be customized based on actual needs and the size of the geographic area. For example, a highway area can be divided into several equal-sized square grids. The number or intensity of environmental and structural features within each grid cell is then counted to calculate the environmental and structural feature density values. Feature density fusion is then performed on each spatial grid cell, calculating the weighted overlay of the environmental and structural feature density values for each grid cell to generate a spatial grid density distribution map. The weighted overlay values can be calculated based on the degree of impact of environmental and structural characteristics on highway disasters. For example, if structural damage characteristics are considered to have a greater impact on highway disasters, a greater weight can be assigned to the structural characteristic density value. Based on the weighted overlay values of each spatial grid cell in the spatial grid density distribution map, contiguous spatial grid areas exceeding a preset density threshold are extracted to generate a preliminary disaster impact area. The preset density threshold can be set based on historical disaster data and practical experience. When the weighted overlay value of a spatial grid cell exceeds this threshold, the area is considered potentially affected by the disaster. Spatial morphological analysis is performed on the preliminary disaster impact area to extract the boundary coordinates and spatial diffusion direction vector of the preliminary disaster impact area. Spatial morphological analysis can utilize image processing algorithms, such as edge detection and morphological processing, to identify the boundary of the preliminary disaster impact area and calculate its diffusion direction. Finally, the core area of the disaster impact range is determined based on the boundary coordinates, and the diffusion buffer zone of the disaster impact range is generated in combination with the spatial diffusion direction vector. The core area and the diffusion buffer zone are spatially merged to generate a disaster impact range identifier containing the boundary coordinates and diffusion trend parameters of the disaster impact range.

[0048] As an embodiment, step S420 may specifically include the following steps: Step S421: Obtaining spatial coordinate sequences for each surface deformation feature, vegetation cover anomaly feature, and water body distribution change feature in the set of environmentally relevant features to generate first spatial distribution information. Surface deformation features, vegetation cover anomaly features, and water body distribution change features are key features in the set of environmentally relevant features. Their spatial coordinate sequences can accurately reflect the geographic distribution of these environmental factors. A specific method for obtaining the spatial coordinate sequences of these features is, for example, to generate a digital elevation model (DEM) using a stereo matching algorithm during surface deformation analysis and calculation of surface elevation change rates and surface crack distribution characteristics for surface deformation features. During this process, the geographic coordinates of the locations where surface deformation occurs can be recorded to form a spatial coordinate sequence for the surface deformation features. For vegetation cover anomaly features, when performing thermal radiation intensity analysis on the infrared band data subset, abnormal surface temperature areas and heat source distribution characteristics can be identified. The boundary coordinates or center point coordinates of these abnormal areas can be obtained to form a spatial coordinate sequence for the vegetation cover anomaly features. To characterize water distribution changes, the interferometric synthetic aperture radar (InSAR) processing of radar band data subsets extracts vertical and horizontal displacement vectors from the ground surface, thereby analyzing changes in water distribution. The boundary coordinates of the water distribution change area can be recorded to generate a spatial coordinate sequence of water distribution change characteristics. By integrating the spatial coordinate sequences of surface deformation characteristics, vegetation cover anomalies, and water distribution change characteristics, the first spatial distribution information is generated.

[0049] Step S422: Obtain the spatial positioning sequence of each vibration frequency offset feature, stress distribution abnormality feature, and material fatigue accumulation feature in the structural damage feature set to generate second spatial distribution information. The vibration frequency offset feature, stress distribution abnormality feature, and material fatigue accumulation feature are key features in the structural damage feature set, and their spatial positioning sequence can accurately reflect the distribution of highway structural damage in geographical space. The process of obtaining the spatial positioning sequence of these features is, for example, that for the vibration frequency offset feature, after performing frequency domain conversion processing on the structural vibration response information in the sub-sensor data set, the fundamental frequency component, harmonic component, and noise component are extracted, and the fundamental frequency offset and harmonic energy ratio are calculated. In this process, the position coordinates of each sensor can be recorded. Since parameters such as the fundamental frequency offset correspond to the sensor position, these sensor position coordinates constitute the spatial positioning sequence of the vibration frequency offset feature.

[0050] For stress distribution anomaly characteristics, after spatial interpolation of the stress distribution information in the sub-sensor data set, a stress distribution cloud map is generated, and the maximum stress value and stress gradient change rate are extracted. The center point coordinates or boundary coordinates of the stress anomaly area in the stress distribution cloud map can be obtained to form a spatial positioning sequence of the stress distribution anomaly characteristics. For material fatigue accumulation characteristics, their values are determined based on parameters such as fundamental frequency offset, harmonic energy ratio, maximum stress value, and stress gradient change rate. Similarly, the sensor position coordinates or the coordinates of the stress anomaly area corresponding to these parameters can be used as the spatial positioning sequence of the material fatigue accumulation characteristics. By integrating the spatial positioning sequences of the vibration frequency offset characteristics, stress distribution anomaly characteristics, and material fatigue accumulation characteristics, a second spatial distribution information is generated.

[0051] Step S423: The first spatial distribution information and the second spatial distribution information are spatially overlaid to generate spatial distribution overlay data containing the density distribution of environmental features and the density distribution of structural features. Spatial overlay processing can utilize Geographic Information System (GIS) technology, enabling the overlay and analysis of spatial data from different sources. The specific process for implementing spatial overlay processing is as follows: First, the first and second spatial distribution information are imported into GIS software, ensuring that they are in the same geographic coordinate system. For example, a common geographic coordinate system such as WGS84 can be selected. Then, within the GIS software, spatial overlay analysis tools are used to overlay the first and second spatial distribution information. This overlay operation merges the geographic features in the two spatial distribution information to generate a new spatial dataset. In this new dataset, each geographic unit (e.g., grid, polygon, etc.) contains information related to environmental and structural features. Next, the number of environmental and structural features within each geographic unit is counted to calculate the density of the environmental and structural features. Environmental feature density can be defined as the number of environmental features (such as surface deformation and vegetation cover anomalies) within each geographic unit divided by the unit's area. Structural feature density can be defined as the number of structural features (such as vibration frequency shift and stress distribution anomalies) within each geographic unit divided by the unit's area. Finally, the environmental and structural feature density information is added to a new spatial dataset to generate spatially distributed overlay data containing both the environmental and structural feature density distributions.

[0052] Step S424: The spatially distributed overlay data is spatially gridded to generate multiple spatial grid cells, each containing environmental feature density values and structural feature density values. The spatial gridding process, for example, involves determining the size and shape of the spatial grid based on the scope of the highway area and actual requirements. The size of the spatial grid can be set based on the scale of the highway and the required data accuracy, and the shape can be square, rectangular, or hexagonal. Then, in the Geographic Information System (GIS) software, a grid generation tool is used to divide the spatially distributed overlay data based on the determined grid size and shape. The grid generation tool divides the highway area into multiple non-overlapping grid cells and assigns a unique identifier to each grid cell. Next, for each spatial grid cell, the number of environmental features and structural features within it is counted. The number of environmental features can be determined by counting the number of surface deformation features, vegetation cover anomaly features, and water distribution change features within the grid cell. The number of structural features can be determined by counting the number of vibration frequency shift features, stress distribution anomaly features, and material fatigue accumulation features within the grid cell. Finally, based on the grid cell area and the statistically calculated number of environmental and structural features, the environmental and structural feature density values for each spatial grid cell are calculated. The environmental feature density value can be expressed as the number of environmental features divided by the grid cell area, while the structural feature density value can be expressed as the number of structural features divided by the grid cell area. By adding the environmental and structural feature density values to the attribute information of each spatial grid cell, a spatial grid cell containing these values is generated.

[0053] Step S425: Perform feature density fusion processing on each spatial grid cell, calculating the weighted superposition of the environmental feature density value and the structural feature density value for each spatial grid cell, and generating a spatial grid density distribution map. The feature density fusion process, for example, involves determining weights for the environmental feature density value and the structural feature density value. The weights are determined based on the impact of environmental factors and structural damage on highway hazards. For example, if the geological conditions surrounding a highway in a certain area are complex and environmental factors such as surface deformation have a significant impact on highway hazards, a higher weight can be assigned to the environmental feature density value. If the highway's structural design is critical and the impact of structural damage on highway hazards is more pronounced, a higher weight can be assigned to the structural feature density value. The weight ranges from [0 to 1], and the sum of the weights for the environmental feature density value and the structural feature density value is 1. Then, for each spatial grid cell, a weighted superposition of the environmental feature density value and the structural feature density value is calculated based on the determined weights. Finally, the weighted superposition value for each spatial grid cell is added to its attribute information and visualized in Geographic Information System (GIS) software to generate a spatial grid density distribution map. The spatial grid density distribution map can use different colors or grayscale to represent different weighted overlay values. The darker the color or the higher the grayscale, the greater the comprehensive density value of the grid unit, that is, the higher the potential risk of the area being affected by highway disasters.

[0054] Step S426: Based on the weighted superposition values of each spatial grid cell in the spatial grid density distribution map, continuous spatial grid areas exceeding a preset density threshold are extracted to generate a preliminary disaster-affected area. The preset density threshold is a comprehensive density value standard pre-set based on historical disaster data and actual experience. When the weighted superposition value of a spatial grid cell exceeds this threshold, the area is considered to be potentially affected by the disaster. The process of extracting the preliminary disaster-affected area includes, for example, determining a preset density threshold. The preset density threshold can be determined by analyzing a large amount of historical disaster data and statistically analyzing the distribution of comprehensive density values under different disaster levels to determine an appropriate preset density threshold. For example, if historical data shows that when the comprehensive density value exceeds 0.8, the probability of a highway disaster occurring in the area is high, then the preset density threshold can be set to 0.8. Then, in the spatial grid density distribution map, each spatial grid cell is traversed to check whether its weighted superposition value exceeds the preset density threshold. Spatial grid cells exceeding the preset density threshold are marked as potentially affected by the disaster. Next, a connected region analysis algorithm, such as a flood fill algorithm or a four-neighborhood or eight-neighborhood search algorithm, is used to identify all contiguous spatial grid regions marked as potentially affected by the disaster. A contiguous spatial grid region is defined as a region consisting of adjacent marked cells. Adjacency can be defined as either four-neighborhood (up, down, left, and right) or eight-neighborhood (up, down, left, and right, as well as the four diagonal adjacent cells) based on actual needs. Finally, these contiguous spatial grid regions are merged to generate a preliminary disaster impact area. This preliminary disaster impact area is a polygonal area composed of multiple contiguous spatial grid cells that reflects the geographic scope of the potential impact of a highway disaster.

[0055] Step S427: Perform spatial morphological analysis on the preliminary disaster-affected area to extract the boundary coordinates and spatial diffusion direction vector of the preliminary disaster-affected area. For example, the spatial morphological analysis process involves extracting the boundary of the preliminary disaster-affected area. Edge detection algorithms used in image processing, such as the Canny edge detection algorithm or the Sobel edge detection algorithm, can be used to separate the boundary of the preliminary disaster-affected area from the background. These algorithms calculate the gradient values of pixels in the image, identify pixels with larger gradient values, and connect them to form a boundary. After extracting the boundary of the preliminary disaster-affected area, the coordinates of each point on the boundary are obtained. These coordinates are the boundary coordinates of the preliminary disaster-affected area. The boundary coordinates accurately describe the geometric shape and location of the preliminary disaster-affected area. The spatial diffusion direction vector is then extracted. The principal component analysis (PCA) algorithm can be used to analyze the spatial grid cells within the preliminary disaster-affected area. Specifically, the coordinates of the spatial grid cells within the preliminary disaster-affected area are used as input data to calculate the covariance matrix. Eigenvalue decomposition is then performed on the covariance matrix to obtain eigenvalues and eigenvectors. The eigenvectors are sorted according to their eigenvalues, and the eigenvector with the largest eigenvalue is selected as the principal component direction. The principal component direction represents the main diffusion direction of the initial disaster impact area, and the unit vector of this direction is used as the spatial diffusion direction vector.

[0056] Step S428: Determine the core area of the disaster impact range based on the boundary contour coordinates, and generate a diffusion buffer zone for the disaster impact range in combination with the spatial diffusion direction vector. The core area of the disaster impact range is the area most severely affected by the disaster, typically located within the initial disaster impact area. The diffusion buffer zone for the disaster impact range is the area where the disaster is likely to spread further, surrounding the core area. Determining the core area of the disaster impact range involves, for example, calculating the geometric center of the initial disaster impact area based on the boundary contour coordinates. For example, using a centroid calculation method, the coordinates of each point on the boundary contour are weighted averaged to obtain the centroid coordinates of the initial disaster impact area. Then, with the centroid as the center, the core area is determined according to a predefined rule. For example, the core area can be determined based on the weighted superposition values of each spatial grid cell in the spatial grid density distribution map, with the area with the highest weighted superposition value selected as the core area. Alternatively, a suitable radius can be set with the centroid as the center, and the area within this radius is designated as the core area. Generating the diffusion buffer zone for the disaster impact range involves, for example, determining the direction of potential disaster spread in combination with the spatial diffusion direction vector. The spatial diffusion direction vector represents the primary diffusion direction of the initial disaster impact area. Based on the direction and magnitude of this vector, the potential range of the disaster spread can be predicted. Then, based on the boundary of the core area, the diffusion buffer zone is expanded along the direction of the spatial diffusion direction vector by a preset distance. The expansion distance can be set based on historical disaster data and actual experience.

[0057] Step S429: Spatially merge the core area and the diffusion buffer zone to generate a disaster impact zone identifier containing the boundary coordinates and diffusion trend parameters. Specifically, first import the geographic data (e.g., polygon boundary coordinates) of the core area and diffusion buffer zone into GIS software, ensuring they are in the same geographic coordinate system. Then, use the spatial merge tool in the GIS software to merge the core area and diffusion buffer zone. This merge operation integrates the boundaries of the two zones, generating a new polygonal zone, which serves as the final boundary of the disaster impact zone. The boundary coordinates of the merged polygonal zone are obtained; these coordinates are the boundary coordinates of the disaster impact zone. The boundary coordinates of the disaster impact zone accurately describe the geometry and location of the disaster impact zone. The diffusion trend parameters can be determined based on the previously extracted spatial diffusion direction vector and the expansion distance of the diffusion buffer zone. For example, the direction and magnitude of the spatial diffusion direction vector can be used as the diffusion direction parameters, and the expansion distance of the diffusion buffer zone can be used as the diffusion distance parameter. These diffusion trend parameters, combined with the boundary coordinates of the disaster impact zone, form the disaster impact zone identifier.

[0058] Step S430: The disaster type identifier and the disaster impact range identifier are associated and combined to generate a disaster detection result set comprising multiple disaster detection units. Each disaster detection unit includes the coordinates of the disaster location, a disaster type priority score, and a predicted disaster impact spread trend. For example, the association process involves associating the corresponding disaster type identifier and disaster impact range identifier for each highway disaster feature. The disaster type identifier and disaster impact range identifier can be bound to the highway disaster feature by adding corresponding fields to the data record. Next, the coordinates of the disaster location are determined. The coordinates of the disaster location can be determined based on the geometric center coordinates of the core area in the disaster impact range identifier. For example, the centroid coordinates of the core area's boundary contour coordinates can be calculated and used as the coordinates of the disaster location. Next, a disaster type priority score is assigned. This priority score comprehensively considers the severity of the disaster type and the size of the disaster impact range. Different initial priority scores can be set for different disaster types (such as landslide risk, pavement cracks, and bridge structural failure) based on historical disaster data and expert experience. For example, a landslide risk may cause significant damage to a highway, so its initial priority score can be set to a higher value. Pavement cracks, however, pose a relatively minor threat, so their initial priority score can be set to a lower value. The initial priority score is then adjusted based on the size of the disaster's impact area. A larger impact area indicates a more severe disaster, and the priority score should be increased accordingly. This adjustment can be performed using a linear weighting approach. For example, a range adjustment coefficient can be set, and the area of the disaster's impact area multiplied by this coefficient. This coefficient is then added to the initial priority score to obtain the final disaster type priority score. Finally, the disaster impact diffusion trend prediction results are determined. These prediction results can be determined based on the diffusion trend parameters in the disaster impact area identifier. These diffusion trend parameters include information such as diffusion direction and diffusion distance, and can be used to predict the likely range and direction of the disaster's spread over a period of time. The disaster location coordinates, disaster type priority score, and disaster impact diffusion trend prediction results are associated with the disaster type identifier and the disaster impact area identifier to form a disaster detection unit. Repeat the above steps for all highway disaster features to generate multiple disaster detection units. These disaster detection units are then integrated to generate a disaster detection result set containing multiple disaster detection units.

[0059] As an embodiment, step S430 may specifically include the following steps: Step S431: Obtaining the disaster type classification result corresponding to the disaster type identifier and the disaster impact range boundary coordinates corresponding to the disaster impact range identifier. The process of obtaining the disaster type classification result corresponding to the disaster type identifier may be, for example, that the disaster type identifier corresponding to each highway disaster feature has been determined during the previous disaster type classification process. The specific disaster type corresponding to each disaster type identifier can be obtained by querying a mapping table between disaster type identifiers and disaster type classification results. For example, the mapping table may specify that disaster type identifier "1" corresponds to landslide risk, and disaster type identifier "2" corresponds to pavement cracks. By querying this mapping table, the disaster type classification result corresponding to each disaster type identifier can be obtained. The process of obtaining the disaster impact range boundary coordinates corresponding to the disaster impact range identifier may be, for example, that the disaster impact range identifier containing the disaster impact range boundary coordinates has been obtained through spatial merging during the previous disaster impact range identifier generation process. The disaster impact range boundary coordinates can be directly extracted from the data record of the disaster impact range identifier. The disaster impact range boundary coordinates are typically represented in the form of a polygon, with each vertex coordinate represented by latitude and longitude or other coordinate values in a geographic coordinate system.

[0060] Step S432: Extract the surface deformation feature change rate from the environmentally relevant feature set and the stress gradient change rate from the structural damage feature set within the boundary coordinates of the disaster impact area to generate the coordinates of the disaster location. The process of extracting the surface deformation feature change rate and the stress gradient change rate involves, for example, filtering feature data from the environmentally relevant feature set and the structural damage feature set within the boundary coordinates of the disaster impact area. The spatial query function of geographic information system (GIS) software can be used to compare the geographic coordinates of the environmentally relevant feature set and the structural damage feature set with the boundary coordinates of the disaster impact area to filter out feature data within the boundary coordinates. For the surface deformation features in the filtered environmentally relevant feature set, their change rate is calculated. The surface deformation feature change rate can be calculated by comparing the surface elevation change rate at different times. For example, in the previous surface deformation analysis process, digital elevation models (DEMs) at different times were obtained. The surface deformation feature change rate can be obtained by calculating the elevation difference between two adjacent DEMs and dividing it by the time interval. For the stress distribution information in the filtered structural damage feature set, the stress gradient change rate is calculated. The stress gradient change rate can be obtained by selecting several adjacent pixel points in the stress distribution cloud map, calculating the stress difference and distance difference between them, and then dividing the stress difference by the distance difference. After obtaining the surface deformation characteristic change rate and the stress gradient change rate, they are used as elements of the feature vector. Cluster analysis algorithms, such as the K-means clustering algorithm, can be used to perform cluster analysis on these feature vectors. The principle of the K-means clustering algorithm is to divide the data points into K clusters so that the distance between the data points in each cluster is minimized, while the distance between the data points in different clusters is maximized. Through cluster analysis, the geographic coordinates corresponding to the cluster center are found, and the geographic coordinates are used as the coordinates of the disaster location.

[0061] Step S433: Based on the diffusion rate parameters and impact range growth rate parameters in the historical disaster pattern library for each type of disaster in the disaster type classification result, a diffusion trend prediction process is performed on the boundary coordinates of the current disaster impact range to generate a disaster impact diffusion trend prediction result. The historical disaster pattern library is pre-established and contains standard feature templates and related parameter information for various disaster types, such as diffusion rate parameters and impact range growth rate parameters. The diffusion rate parameter represents the distance the disaster spreads per unit time, and the impact range growth rate parameter represents the growth rate of the disaster impact range per unit time. The diffusion trend prediction process is, for example, to obtain the diffusion rate parameters and impact range growth rate parameters of the corresponding disaster type from the historical disaster pattern library based on the disaster type classification result. For example, if the disaster type classification result is landslide risk, the diffusion rate parameters and impact range growth rate parameters of the landslide disaster are searched from the historical disaster pattern library.

[0062] Then, based on the diffusion rate parameters and the impact range growth rate parameters, the diffusion simulation of the current disaster impact range boundary coordinates is performed. A vector-based diffusion model, such as a buffer analysis model, can be used. The principle of the buffer analysis model is to expand the current disaster impact range boundary according to the diffusion rate parameters and the impact range growth rate parameters. Specifically, the current disaster impact range boundary coordinates are converted into vector data, such as polygons. Then, based on the diffusion rate parameters and the impact range growth rate parameters, the expansion distance and direction are determined. The expansion direction can be determined based on the spatial diffusion direction vector extracted previously. Finally, based on the current disaster impact range boundary, expansion is performed according to the determined distance and direction to generate the disaster impact range boundary at a certain point in the future.

[0063] Repeat the above steps to simulate the boundaries of the disaster impact area at different time points to determine how the disaster impact area changes over time. These simulation results are compiled into a prediction of the disaster impact diffusion trend, including information such as the coordinates of the disaster impact area boundary at different time points, the diffusion direction, and the diffusion speed.

[0064] Step S434: Based on the preset disaster hazard level mapping table corresponding to the disaster type classification results, an initial disaster type priority score is assigned to each disaster type identifier. The process for establishing the preset disaster hazard level mapping table, for example, involves assessing the hazard level of various highway disaster types (such as landslide risk, pavement cracks, and bridge structural failure). This hazard level assessment considers multiple factors, such as the impact of the disaster on highway traffic, the extent of damage to the surrounding environment, and the resulting economic losses. An assessment team composed of experts in highway engineering, geology, and environmental science can be invited to assign a hazard level score for each disaster type based on historical disaster data and practical experience. Then, based on the hazard level scores, the disaster types are classified into different hazard levels, such as low, medium, and high. A corresponding priority score range is set for each hazard level. For example, the priority score range for the low hazard level can be set between 1 and 3 points, the priority score range for the medium hazard level can be set between 4 and 6 points, and the priority score range for the high hazard level can be set between 7 and 10 points. Finally, each disaster type is associated with the corresponding hazard level and priority score to form a preset disaster hazard level mapping table.

[0065] Step S435: Based on the diffusion rate and the growth rate of the impact range in the disaster impact diffusion trend prediction results, the initial disaster type priority score is dynamically corrected to generate a disaster type priority score. The process of dynamic correction processing is, for example, to determine the influence coefficient of the diffusion rate and the growth rate of the impact range on the priority score. The influence coefficient can be set according to historical disaster data and actual experience. For example, the influence coefficient of the diffusion rate can be set to k1, and the influence coefficient of the growth rate of the impact range can be set to k2. The larger the value of the influence coefficient, the greater the influence of the factor on the priority score. Then, based on the diffusion rate and the growth rate of the impact range in the disaster impact diffusion trend prediction results, the correction value is calculated. For example, the correction value = k1×diffusion rate+k2×influence range growth rate. Finally, the correction value is added to the initial disaster type priority score to obtain the disaster type priority score. For example, assuming that the initial disaster type priority score is P0, the correction value is , then the disaster type priority score P=P0+ .

[0066] Step S436: The coordinates of the disaster occurrence location, the disaster type priority score and the disaster impact diffusion trend prediction result are associated and mapped according to the spatial coordinates to generate a disaster detection unit corresponding to each disaster impact range identifier.

[0067] For example, the association mapping process involves ensuring that the disaster location coordinates, disaster type priority score, and disaster impact spread trend prediction results all have clear spatial coordinate information. The disaster location coordinates themselves are based on geographic spatial positioning, and the boundary coordinates of the disaster impact area at different time points in the disaster impact spread trend prediction results also contain spatial location information. The disaster type priority score can be associated with the corresponding disaster location coordinates, giving them spatial attributes. The disaster location coordinates are then matched with the spatial coordinates of the disaster impact area identifier. If the disaster location coordinates fall within the geographic area defined by the disaster impact area identifier, the disaster type priority score and disaster impact spread trend prediction results corresponding to the disaster location coordinates are associated with the disaster impact area identifier. During the matching process, the spatial query function of the Geographic Information System (GIS) can be used to determine whether the point (disaster location coordinates) is within a polygon (the area defined by the boundary coordinates of the disaster impact area identifier). For each successfully matched combination, the disaster location coordinates, disaster type priority score, and disaster impact spread trend prediction results are integrated to form a complete disaster detection unit. This unit clearly displays the location, severity, and potential future spread of a disaster within its impact area. For example, a disaster detection unit might include specific latitude and longitude coordinates representing the disaster location, a specific numerical value representing the disaster type priority score, and a series of coordinates representing the boundary of the disaster impact area and a diffusion direction vector at different time points to represent the predicted spread trend of the disaster impact.

[0068] Step S437: Aggregate the spatial distribution information of all disaster detection units to generate a disaster detection result set including a disaster occurrence location coordinate set, a disaster type priority score sequence, and a disaster impact diffusion trend prediction result set. The process of aggregating the spatial distribution information of all disaster detection units is, for example, to establish three different data structures, each used to store the disaster occurrence location coordinate set, the disaster type priority score sequence, and the disaster impact diffusion trend prediction result set. For the disaster occurrence location coordinate set, a list or array can be used to store the disaster occurrence location coordinates in each disaster detection unit. For example, if the disaster occurrence location coordinates are expressed in longitude and longitude, then each element in the list can be a two-tuple containing longitude and latitude values. For the disaster type priority score sequence, a list or array can also be used to store the disaster type priority score in each disaster detection unit. The elements in the list are arranged in the order of the disaster detection units, so that each priority score corresponds to the corresponding disaster occurrence location coordinates and disaster impact diffusion trend prediction result.

[0069] A nested data structure can be used to store the disaster impact diffusion trend prediction result set. Since each disaster impact diffusion trend prediction result contains information such as the boundary coordinates of the disaster impact range, diffusion direction, and diffusion speed at different time points, the disaster impact diffusion trend prediction result of each disaster detection unit can be stored as a sublist or sub-dictionary, and then all sublists or sub-dictionaries can be stored in a main list. After establishing the data structure, all disaster detection units are traversed, and the disaster location coordinates in each disaster detection unit are added to the disaster location coordinate set, the disaster type priority score is added to the disaster type priority score sequence, and the disaster impact diffusion trend prediction result is added to the disaster impact diffusion trend prediction result set. By aggregating the spatial distribution information of all disaster detection units, the generated disaster detection result set contains comprehensive information on highway disasters, including the location, severity, and future diffusion trend of the disaster.

[0070] Step S500: Generate an inspection optimization strategy set based on the disaster detection result set. The inspection optimization strategy set is used to adjust the coordinated inspection paths and monitoring parameter configurations of aerial and ground inspection equipment. For example, the process of generating the inspection optimization strategy set includes determining resource allocation weights for aerial and ground inspection equipment based on the disaster type priority scores in the disaster detection result set.

[0071] Then, based on the disaster impact spread trend prediction results, dynamic path adjustment instructions for aerial inspection equipment and monitoring parameter update instructions for ground inspection equipment are generated. The disaster impact spread trend prediction results provide the likely range and direction of the disaster's spread over a period of time. Based on this information, the aerial inspection equipment's inspection path can be dynamically adjusted to cover the areas where the disaster is likely to spread. Simultaneously, the ground inspection equipment's monitoring parameters, such as sensor sampling frequency and monitoring range, are updated to better monitor the disaster's development. Next, the coordinated inspection frequency and monitoring data acquisition resolution of aerial and ground inspection equipment are adjusted based on the spatial distribution density of the disaster location coordinates. If the spatial distribution density of the disaster location coordinates is high, indicating a high probability of disaster in that area, the coordinated inspection frequency and monitoring data acquisition resolution should be increased to more promptly and accurately detect potential disaster risks. Finally, the resource allocation weights, dynamic path adjustment instructions, monitoring parameter update instructions, and coordinated inspection frequency are integrated into a set of inspection optimization strategies and sent to the corresponding inspection equipment control terminals. After receiving the inspection optimization strategy set, the inspection equipment control terminal adjusts the working mode of the aerial inspection equipment and the ground inspection equipment according to the instructions and parameters therein to achieve the optimized configuration of the collaborative inspection path and monitoring parameters.

[0072] As an implementation method, step S500 may specifically include the following steps: Step S510: Determine resource allocation weights for aerial inspection equipment and ground inspection equipment based on the disaster type priority scores in the disaster detection result set. The disaster type priority scores are derived by comprehensively considering factors such as the degree of harm and impact of the disaster, reflecting the severity and urgency of different disasters. Reasonable resource allocation weights ensure that limited inspection resources are prioritized to areas with higher disaster risks, improving the efficiency and effectiveness of highway disaster detection.

[0073] As an implementation method, step S510 may specifically include the following steps: Step S511: assigning an initial priority score to each disaster detection unit according to the disaster hazard level corresponding to the disaster type identifier. In order to establish a corresponding relationship between the disaster type identifier and the disaster hazard level, a disaster hazard level mapping table may be pre-constructed. This mapping table is formulated by analyzing and summarizing a large amount of historical highway disaster data, combined with the opinions of highway engineering experts, geological experts and environmental experts. For example, for disaster types such as landslide risk, since it may cause large-scale damage to the highway, traffic interruption, and even endanger the lives and property of surrounding residents, its corresponding disaster hazard level is usually set to a high level in the mapping table; and for disaster types such as road cracks, although it will also affect the normal use of the highway, the degree of harm is relatively small, and its corresponding disaster hazard level may be set to a medium level.

[0074] When assigning an initial priority score to each disaster detection unit, the corresponding disaster hazard level is searched from the disaster hazard level mapping table based on the disaster type identifier. By assigning an initial priority score to each disaster detection unit based on the disaster hazard level corresponding to the disaster type identifier, we can preliminarily quantify the severity of different disasters, providing a basis for subsequent dynamic weighting adjustments based on the prediction results of the disaster impact diffusion trend.

[0075] Step S512: Dynamically weight the initial priority score based on the diffusion rate and impact range growth rate from the disaster impact diffusion trend prediction results to generate a disaster type priority score. The diffusion rate represents the distance the disaster spreads per unit time, while the impact range growth rate represents the percentage increase in the disaster's impact range per unit time. Faster diffusion rates and greater impact range growth rates indicate more rapid disaster development and greater impact on the highway and surrounding environment, necessitating a corresponding increase in the disaster's priority score.

[0076] For example, the dynamic weighted adjustment process involves determining the influence coefficients of the diffusion rate and impact range growth rate on the priority score. Determining the influence coefficients requires considering multiple factors, such as the type of disaster, the importance of the highway, and the sensitivity of the surrounding environment. By analyzing a large amount of historical disaster data and incorporating expert experience, the influence coefficients for the diffusion rate are set as k1 and the influence coefficient for the impact range growth rate are set as k2. Then, based on the diffusion rate and impact range growth rate from the disaster impact diffusion trend prediction results, a correction value is calculated. Finally, the correction value is added to the initial priority score to obtain the disaster type priority score. For example, assuming a disaster detection unit has an initial priority score of P0, a diffusion rate of v, and an impact range growth rate of r, the disaster type priority score P = P0 + k1 × v + k2 × r. By dynamically weighting the initial priority score based on the diffusion rate and impact range growth rate from the disaster impact diffusion trend prediction results, the resulting disaster type priority score more accurately reflects the actual severity of the disaster and provides a more accurate basis for subsequently calculating resource allocation weights for aerial and ground inspection equipment based on the spatial distribution density of the disaster type priority score.

[0077] Step S513: Based on the spatial distribution density of the disaster type priority scores, calculate the image acquisition frequency weight required for aerial inspection equipment and the sensor deployment density weight required for ground inspection equipment. The process of calculating the image acquisition frequency weight and the sensor deployment density weight is, for example, to divide the highway area into a number of grid cells, each of which has an area. Geographic Information System (GIS) technology can be used for grid division, and the size of the grid cells can be determined based on the actual scope and accuracy requirements of the highway. For example, the highway area can be divided into square grid cells with a side length of 100 meters. Then, the sum of the disaster type priority scores within each grid cell is calculated. For each disaster detection unit, the grid cell to which it belongs is determined based on the coordinates of the disaster occurrence location, and the disaster type priority score of the disaster detection unit is added to the sum of the scores of the grid cells to which it belongs.

[0078] Next, the spatial distribution density of the disaster type priority scores for each grid cell is calculated. The spatial distribution density can be obtained by dividing the sum of the disaster type priority scores within the grid cell by the area of the grid cell. Based on the spatial distribution density of the disaster type priority scores, the image acquisition frequency weight required for aerial inspection equipment and the sensor deployment density weight required for ground inspection equipment are calculated. A scaling factor can be set to calculate the image acquisition frequency weight and the sensor deployment density weight, respectively. For example, the image acquisition frequency weight scaling factor can be set to m1, and the sensor deployment density weight scaling factor can be set to m2. For each grid cell, its image acquisition frequency weight can be obtained by multiplying the spatial distribution density of the disaster type priority scores for that grid cell by m1, and its sensor deployment density weight can be obtained by multiplying the spatial distribution density of the disaster type priority scores for that grid cell by m2. Finally, the image acquisition frequency weight and sensor deployment density weight for all grid cells are normalized to ensure that their sum is 1. Normalization can be achieved by dividing the weight of each grid cell by the sum of the weights of all grid cells.

[0079] Step S514: Resource allocation weights are generated based on the image acquisition frequency weights and sensor deployment density weights, and the task scheduling queues for aerial and ground inspection equipment are dynamically adjusted. The resource allocation weights clarify the resource allocation ratio between aerial and ground inspection equipment, while the dynamic adjustment of the task scheduling queue ensures that inspection resources are allocated to areas with high disaster risk in a timely and accurate manner.

[0080] The process of generating resource allocation weights, for example, involves integrating the image acquisition frequency weight and the sensor deployment density weight. These weights can be added together according to a preset ratio to obtain the resource allocation weight for each grid cell. For example, assuming the image acquisition frequency weight is w1 and the sensor deployment density weight is w2, an integration coefficient α (0 ≤ α ≤ 1) can be set, resulting in the resource allocation weight w = α × w1 + (1 - α) × w2. The value of the integration coefficient α can be adjusted based on actual circumstances. If the image acquisition function of aerial inspection equipment is prioritized, a larger value can be set; if the sensor monitoring function of ground inspection equipment is prioritized, a smaller value can be set. The resource allocation weights for all grid cells are then normalized to ensure that their sum is 1. Normalization can be achieved by dividing the resource allocation weight of each grid cell by the sum of the resource allocation weights for all grid cells.

[0081] After generating resource allocation weights, the task scheduling queues for aerial and ground inspection equipment are dynamically adjusted. The task scheduling queue records the inspection task schedules for aerial and ground inspection equipment, including information such as inspection areas, times, and frequency. Based on resource allocation weights, inspection tasks are increased in areas with higher disaster risk, while those in areas with lower risk are reduced. For example, for grid cells with high resource allocation weights, image acquisition tasks for aerial inspection equipment and sensor monitoring tasks for ground inspection equipment can be increased to increase inspection frequency. For grid cells with low resource allocation weights, inspection tasks can be appropriately reduced. Furthermore, the operating capacity and resource constraints of aerial and ground inspection equipment must be considered. If a region has a high resource allocation weight but limited capacity, aerial and ground inspection equipment cannot meet the inspection needs, reasonable task allocation and coordination are necessary. Inspection equipment efficiency can be improved by optimizing inspection routes and adjusting inspection times, ensuring coverage of high-risk areas as much as possible within resource constraints.

[0082] Step S520: Based on the disaster impact spread trend prediction results, dynamic route adjustment instructions for aerial inspection equipment and monitoring parameter update instructions for ground inspection equipment are generated. The process of generating dynamic route adjustment instructions for aerial inspection equipment, for example, involves determining the areas where the disaster is likely to spread based on the boundary coordinates and spread direction of the disaster impact area at different time points in the disaster impact spread trend prediction results. The boundary coordinates of the disaster impact area at different time points can be visualized in a geographic information system (GIS). By analyzing the boundary changes and spread direction, the main areas where the disaster is likely to spread can be determined.

[0083] Then, a new inspection route is planned based on the current location and operating capacity of the aerial inspection equipment. The new inspection route should cover as many areas as possible where the disaster may spread, while also taking into account factors such as the flight time and endurance of the aerial inspection equipment. Path planning algorithms, such as the A* algorithm and the Dijkstra algorithm, can be used to calculate the optimal inspection route based on the scope of the disaster and the constraints of the inspection equipment. Finally, the new inspection route is converted into dynamic path adjustment instructions, including parameters such as flight direction, flight altitude, and flight speed. These instructions are sent to the control terminal of the aerial inspection equipment, which adjusts the flight path according to the instructions and inspects areas where the disaster may spread.

[0084] For example, the process of generating monitoring parameter update instructions for ground inspection equipment involves analyzing the impact of the spread of a disaster on the highway structure and surrounding environment based on the predicted results of the spread of the disaster's impact. For example, if the spread of a disaster is likely to lead to increased ground vibration or stress changes, the monitoring parameters of the ground inspection equipment need to be adjusted accordingly. The sampling frequency and monitoring range of the ground inspection equipment's sensors are adjusted. If the risk in the disaster spread area increases, the sensor sampling frequency is increased to more promptly capture information such as structural vibration and stress changes; the monitoring range is expanded to ensure coverage of areas where the disaster may spread. The adjusted monitoring parameters are converted into monitoring parameter update instructions and sent to the control terminal of the ground inspection equipment. The ground inspection equipment updates the monitoring parameters according to the instructions, providing more effective monitoring of the disaster spread area. By generating dynamic route adjustment instructions for aerial inspection equipment and monitoring parameter update instructions for ground inspection equipment based on the predicted results of the spread of the disaster's impact, the inspection equipment can better adapt to the development and changes of disasters, improving the timeliness and accuracy of highway disaster detection.

[0085] Step S530: Adjust the coordinated inspection frequency and monitoring data acquisition resolution of aerial and ground inspection equipment based on the spatial distribution density of disaster location coordinates. The spatial distribution density of disaster location coordinates reflects the concentration of disasters in different areas of the highway. By properly adjusting the coordinated inspection frequency and monitoring data acquisition resolution, inspection efficiency can be improved and highway disasters can be detected more accurately.

[0086] For example, the process of adjusting the coordinated inspection frequency involves dividing the highway area into several grid cells, each with an area. Grid division can be performed using Geographic Information System (GIS) technology, with the grid cell size determined based on the actual highway area and accuracy requirements. The number of disaster location coordinates within each grid cell is then counted. For each disaster detection unit, the grid cell to which it belongs is determined based on its disaster location coordinates, and the disaster location coordinates are added to the coordinate count of the grid cell to which it belongs. Next, the spatial distribution density of the disaster location coordinates for each grid cell is calculated. The spatial distribution density can be calculated by dividing the number of disaster location coordinates within a grid cell by the grid cell's area. The coordinated inspection frequency of aerial and ground inspection equipment is adjusted based on the spatial distribution density of the disaster location coordinates. Grid cells with a higher spatial distribution density indicate a higher likelihood of disasters in the area, necessitating an increase in the coordinated inspection frequency and the number of inspections. Grid cells with a lower spatial distribution density can be appropriately reduced, resulting in a smaller number of inspections. A frequency adjustment coefficient can be set to adjust the coordinated inspection frequency based on the product of the spatial distribution density and the frequency adjustment coefficient.

[0087] The process of adjusting the monitoring data acquisition resolution is, for example, to increase the image acquisition resolution of the aerial inspection equipment and the sensor data acquisition resolution of the ground inspection equipment for grid cells with higher spatial distribution density, also based on the spatial distribution density of the coordinates of the disaster location. High-resolution images and data can more clearly reflect the structural and environmental information of the highway, which helps to discover potential disaster hazards. For grid cells with lower spatial distribution density, the monitoring data acquisition resolution can be appropriately lowered to reduce the pressure on data processing and storage. By adjusting the collaborative inspection frequency and monitoring data acquisition resolution of the aerial inspection equipment and the ground inspection equipment according to the spatial distribution density of the coordinates of the disaster location, it is possible to reasonably allocate inspection resources and improve the efficiency and accuracy of highway disaster detection.

[0088] Step S540: Integrate the resource allocation weight, dynamic path adjustment instruction, monitoring parameter update instruction and collaborative inspection frequency into an inspection optimization strategy set, and send it to the corresponding inspection equipment control terminal. The inspection optimization strategy set is an instruction set for comprehensively optimizing the working modes of aerial inspection equipment and ground inspection equipment. By sending it to the inspection equipment control terminal, the inspection equipment can work according to the optimized strategy, thereby improving the efficiency and accuracy of highway disaster detection. The process of integrating the inspection optimization strategy set is, for example, to establish a data structure to store the inspection optimization strategy set. A dictionary or list can be used for storage, where each element corresponds to an optimization strategy information. For example, a dictionary can be used for storage, and the dictionary keys can be "resource allocation weight", "dynamic path adjustment instruction", "monitoring parameter update instruction" and "collaborative inspection frequency", etc., and the corresponding values are the corresponding optimization strategy information.

[0089] Resource allocation weight information is added to the inspection optimization policy set. This resource allocation weight information can be stored in a list or matrix format, with each element corresponding to the resource allocation weight for a grid unit. The resource allocation weight information is used as the value of the "resource allocation weight" key in the dictionary. Dynamic path adjustment instructions are added to the inspection optimization policy set. These instructions can include parameters such as flight direction, altitude, and speed, and are stored in a list or dictionary format. These instructions are used as the value of the "dynamic path adjustment instruction" key in the dictionary. Monitoring parameter update instructions are added to the inspection optimization policy set. These instructions can include parameters such as the sensor sampling frequency and monitoring range, and are also stored in a list or dictionary format. These instructions are used as the value of the "monitoring parameter update instruction" key in the dictionary. Collaborative inspection frequency information is added to the inspection optimization policy set. This information can be stored in a list format, with each element corresponding to the collaborative inspection frequency for a grid unit. These instructions are used as the value of the "collaborative inspection frequency" key in the dictionary. After the inspection optimization policy set is integrated, it is sent to the corresponding inspection device control terminal. After receiving the inspection optimization strategy set, the inspection device control terminal parses the instructions and parameters therein and adjusts the working mode of the inspection device according to this information to achieve optimized configuration of collaborative inspection paths and monitoring parameters.

[0090] See also Figure 2 , Figure 2This is a schematic diagram of the structure of a highway disaster detection system provided by an embodiment of the present invention. The highway disaster detection system, for example, is a server or computer device in a monitoring and command center and includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 may be connected via a bus or other means. The processor 101 (also known as the Central Processing Unit (CPU)) is the computing and control core of the highway disaster detection system, capable of parsing various instructions within the highway disaster detection system and processing various data within the system. The communication interface 102 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi or a mobile communication interface), and may be used to send and receive data under the control of the processor 101. The communication interface 102 may also be used for data transmission and interaction within the highway disaster detection system. The memory 103 is a storage device within the highway disaster detection system, used to store programs and data. It is understood that the memory 103 herein may include both the built-in memory of the highway disaster detection system and the extended memory supported by the highway disaster detection system. Memory 103 provides storage space for storing the operating system of the highway disaster detection system, which is not limited by the present invention. In one embodiment, processor 101 executes the highway disaster detection method based on air-ground collaborative unmanned inspection provided in the above embodiments of the present invention by running the computer program in memory 103.

Claims

1. A highway disaster detection method based on air-ground collaborative unmanned inspection, characterized in that: include: Acquire a multi-source highway monitoring data set, the multi-source highway monitoring data set comprising a multi-dimensional remote sensing image data set collected by aerial inspection equipment and a multi-dimensional structural sensing data set collected by ground inspection equipment; the multi-dimensional remote sensing image data set comprising image area coverage information at different timestamps, and the multi-dimensional structural sensing data set comprising structural vibration response information at different spatial locations; Performing spatiotemporal alignment processing on the multi-source highway monitoring data set to generate a highway monitoring data set fused with spatiotemporal labels; Based on a preset multimodal disaster recognition model, disaster feature extraction processing is performed on the highway monitoring data set fused with spatiotemporal labels to obtain a highway disaster feature set, including: environmental feature extraction processing is performed on the sub-image data set in the highway monitoring data set fused with spatiotemporal labels to generate an environmental correlation feature set; the environmental correlation feature set includes surface deformation features, vegetation cover anomaly features and water body distribution change features; structural feature extraction processing is performed on the sub-sensor data set in the highway monitoring data set fused with spatiotemporal labels to generate a structural damage feature set; the structural damage feature set includes vibration frequency offset features, stress distribution anomaly features and material fatigue accumulation features; the environmental correlation feature set and the structural damage feature set are input into the multimodal disaster recognition model for feature fusion processing to generate a fused feature vector set; disaster pattern matching is performed on the fused feature vector set to determine a highway disaster feature set that matches the disaster pattern in a preset disaster pattern library; The highway disaster feature set includes environmental correlation features and structural damage features; Generating a disaster detection result set according to the highway disaster feature set includes: extracting first spatiotemporal reference information from the multidimensional remote sensing image data set, the first spatiotemporal reference information including an image acquisition time series and an image geographic coordinate range; extracting second spatiotemporal reference information from the multidimensional structural sensor data set, the second spatiotemporal reference information including a sensor timestamp series and a sensor positioning coordinate series; performing time window matching on the first spatiotemporal reference information and the second spatiotemporal reference information to determine a time overlap interval between the multidimensional remote sensing image data set and the multidimensional structural sensor data set; performing data slicing processing on the multidimensional remote sensing image data set and the multidimensional structural sensor data set according to the time overlap interval to generate a time-synchronized sub-image data set and a sub-sensor data set; performing spatial coordinate conversion processing on the sub-image data set and the sub-sensor data set to map the image geographic coordinate range to a spatial coordinate system corresponding to the sensor positioning coordinate series, thereby generating a highway monitoring data set with a unified spatial reference and fused spatiotemporal labels; The disaster detection result set includes multiple disaster detection units, each disaster detection unit corresponds to a disaster type identifier and a disaster impact range identifier of a highway area; A set of inspection optimization strategies is generated based on the set of disaster detection results.

2. The method according to claim 1, wherein Generating a disaster detection result set according to the highway disaster feature set includes: Performing disaster type classification processing on each highway disaster feature in the highway disaster feature set to determine a disaster type identifier corresponding to each highway disaster feature; the disaster type identifier is a landslide risk identifier, a pavement crack identifier, or a bridge structure failure identifier; Determining a disaster impact range identifier corresponding to each highway disaster feature based on spatial distribution information of environmental association features and structural damage features in the highway disaster feature set; Associating and combining the disaster type identifier and the disaster impact range identifier to generate a disaster detection result set including multiple disaster detection units; Each disaster detection unit includes the coordinates of the disaster location, the disaster type priority score, and the disaster impact diffusion trend prediction results; Generating a set of inspection optimization strategies based on the set of disaster detection results includes: Determining resource allocation weights for the aerial inspection equipment and the ground inspection equipment based on the disaster type priority scores in the disaster detection result set; Based on the prediction result of the disaster impact diffusion trend, generating a dynamic path adjustment instruction for the aerial inspection equipment and a monitoring parameter update instruction for the ground inspection equipment; Adjusting the coordinated inspection frequency and monitoring data acquisition resolution of the aerial inspection equipment and the ground inspection equipment according to the spatial distribution density of the disaster location coordinates; The resource allocation weights, dynamic path adjustment instructions, monitoring parameter update instructions and collaborative inspection frequency are integrated into an inspection optimization strategy set and sent to the corresponding inspection equipment control terminal.

3. The method according to claim 2, wherein The step of performing environmental feature extraction processing on the sub-image data set in the highway monitoring data set fused with spatiotemporal labels to generate an environmental correlation feature set includes: Performing multispectral band decomposition processing on the sub-image data set to generate a visible light band data subset, an infrared band data subset, and a radar band data subset; Performing surface deformation analysis on the visible light band data subset to extract surface elevation change rate and surface crack distribution characteristics; Performing thermal radiation intensity analysis on the infrared band data subset to extract surface temperature anomaly areas and heat source distribution characteristics; performing interferometric synthetic aperture radar processing on the radar band data subset to extract the vertical displacement and horizontal displacement vector of the ground surface; The surface elevation change rate, surface crack distribution characteristics, surface temperature anomaly areas, heat source distribution characteristics, surface vertical displacement and horizontal displacement vector are comprehensively considered to generate an environment-related feature set.

4. The method according to claim 3, wherein The performing structural feature extraction processing on the sub-sensor data set in the highway monitoring data set fused with spatiotemporal labels to generate a structural damage feature set includes: Performing frequency domain conversion processing on the structural vibration response information in the sub-sensor data set to generate a vibration spectrum feature set; Extracting fundamental frequency components, harmonic components and noise components from the vibration spectrum feature set, and calculating fundamental frequency offset and harmonic energy ratio; Performing spatial interpolation processing on the stress distribution information in the sub-sensor data set to generate a stress distribution cloud map, and extracting the maximum stress value and the stress gradient change rate; According to the fundamental frequency offset, harmonic energy ratio, maximum stress value and stress gradient change rate, the vibration frequency offset feature, stress distribution abnormality feature and material fatigue accumulation feature in the structural damage feature set are determined.

5. The method according to claim 4, wherein The step of inputting the environment-related feature set and the structural damage feature set into the multimodal disaster identification model for feature fusion processing to generate a fused feature vector set includes: Normalizing each environment-related feature in the environment-related feature set to generate a standardized environment feature vector; Normalizing each structural damage feature in the structural damage feature set to generate a standardized structural feature vector; Aligning and splicing the standardized environmental feature vector and the standardized structural feature vector according to timestamps and spatial coordinates to generate a spatiotemporally aligned multimodal feature matrix; Performing cross-modal correlation analysis on the multimodal feature matrix through a feature fusion layer in the multimodal disaster identification model to generate a fusion feature vector set; The cross-modal correlation analysis includes the calculation of correlation weights between environmental features and structural features and the dimensionality reduction processing of feature dimensions.

6. The method according to claim 5, wherein The performing disaster pattern matching on the fused feature vector set to determine a highway disaster feature set that matches a disaster pattern in a preset disaster pattern library includes: Acquire multiple historical disaster pattern feature vectors from the preset disaster pattern library, each historical disaster pattern feature vector corresponding to a standard feature template of a disaster type; Calculating a similarity score between each fused feature vector in the fused feature vector set and each historical disaster pattern feature vector; Determining the optimal matching disaster type corresponding to each fused feature vector based on the similarity score, and generating a preliminary disaster feature set including a disaster type identifier and a matching confidence level; The preliminary disaster feature set is subjected to spatial continuity verification processing, disaster features with matching confidence lower than a preset threshold and isolated spatial distribution are eliminated, and a highway disaster feature set is generated.

7. The method according to claim 6, wherein The determining of resource allocation weights of the aerial inspection equipment and the ground inspection equipment according to the disaster type priority score in the disaster detection result set includes: assigning an initial priority score to each disaster detection unit according to the disaster hazard level corresponding to the disaster type identifier; Combined with the diffusion rate and the growth rate of the impact range in the disaster impact diffusion trend prediction result, the initial priority score is dynamically weighted and adjusted to generate a disaster type priority score; Calculating the image acquisition frequency weight required for the aerial inspection equipment and the sensor deployment density weight required for the ground inspection equipment based on the spatial distribution density of the disaster type priority score; A resource allocation weight is generated based on the image acquisition frequency weight and the sensor deployment density weight, and the task scheduling queues of the aerial inspection equipment and the ground inspection equipment are dynamically adjusted.

8. A highway disaster detection system, characterized in that: include: a memory storing a computer program; A processor is used to load the computer program to implement the highway disaster detection method based on air-ground collaborative unmanned inspection as described in any one of claims 1 to 7.

Images