51 results about "Topographic gradient" patented technology

The invention relates to a multi-modal sparse fusion three-dimensional target detection method in an unstructured environment, and the method is based on a complexity perception candidate region dynamic sampling strategy, and through the cross-domain self-compensation of point cloud geometric features and image spectral features, and the fusion of multi-view sparse features, the detection of a multi-modal sparse fusion three-dimensional target is realized. The perception capability and the recognition precision of the obstacle area, and the remote dynamic target detection performance and the real-time performance are improved; an anti-deformation convolution group and a terrain gradient perception attention module are introduced into a ParScaleNet image backbone network, so that the extraction of edge features of a complex scene is enhanced, and the cross-scale characterization capability of the features is improved; according to the method, a ContraSpaceOpt module with semantic maintenance capability is deployed behind a Transform, excessive smoothness of obstacle features is inhibited through class comparison loss functions and feature space anisotropy constraints, the space separability of obstacles is maintained, and the problems of point cloud fracture and texture misalignment caused by the obstacles are effectively solved.

The invention discloses a metal ore body identification method based on remote sensing interpretation, relates to the technical field of mineral resource exploration, and aims to solve the problems of low identification precision, neglect of geological laws and poor generalization caused by dependence on single remote sensing data in the prior art. After standardization processing, spectrum and texture features are fused to generate a composite vector; extracting a multi-scale mineralization signal by using a convolutional neural network, and interpreting and constructing an ore control area in combination with a terrain gradient and a curvature; alteration zoning is identified through spectrum angle matching, and a favorable mineralization area is determined by combining lithology combination analysis; and dynamically weighting geological elements by adopting a deep learning fusion network, outputting an ore body position according to a comprehensive score, and finally optimizing a boundary through spatial clustering and morphological filtering. The method significantly improves the automation degree and reliability of ore body identification in a complex environment, and is suitable for metal mineral exploration target area delineation.

The invention discloses a land space optimization configuration method under terrain gradient constraint. The method comprises the following steps: S1, collecting multi-source geographic data to construct a spatial database; s2, factors such as gradient and elevation are extracted, and terrain gradient partitions are divided; s3, constructing a three-dimensional conflict matrix based on suitability evaluation and landscape pattern analysis; s4, constructing a driving index system, and analyzing a driving path of the spatial conflict; s5, setting a multi-objective optimization function, and constructing game rules and spatial conversion constraint conditions of the three types of subjects; s6, integrating the C-FLUS model, the MAOOAM model and the R2D3 model, and constructing a coupling system to execute dynamic optimization; and S7, when the convergence condition is met, outputting an optimal configuration scheme and generating a compensation and control strategy. According to the method, efficient optimization and intelligent decision support of multi-target land configuration under the terrain differentiation condition are realized.

The invention relates to the technical field of water conservancy projects, in particular to a water conservancy earthwork dynamic monitoring method and system based on multi-source point cloud data fusion, and the method comprises the steps: carrying out the identification and standardization processing of airborne LiDAR, ground scanning and oblique photography point clouds, and generating a standard point cloud data set with uniform spatial positions and balanced density; dividing an abrupt slope area, a gentle slope area and a flat area based on topographic gradients, and constructing a grid framework adaptive to topographic features; carrying out weighted fusion by combining the point cloud coverage rate and the normalized terrain gradient to generate high-precision grid elevation data; calculating an earthwork variation distribution matrix by comparing historical elevation models; dynamically analyzing the change trend of the excavation and filling area, and synchronously evaluating the construction progress deviation and the geological disaster risk. According to the method, the problems of insufficient multi-source point cloud fusion precision, poor terrain adaptability and dynamic evaluation splitting in the prior art can be solved, and the monitoring precision and the engineering safety early warning efficiency under the complex terrain condition are remarkably improved.

The invention discloses an induction detection system for production and manufacturing based on artificial intelligence and a regulation and control method, and belongs to the field of artificial intelligence. Comprising an edge data acquisition module, a depth image recognition module, a dynamic sensor integrated array, an artificial intelligence analysis module, an intelligent abnormity feedback module and a linkage execution control module. The edge data acquisition module is deployed at key nodes of a production line, models pressure, current, temperature, vibration and other signals, and performs cross sampling with the sensor array. The depth image recognition module adopts a structured light laser array and a multi-phase shift interference technology to extract a morphology gradient map in real time. And the artificial intelligence analysis module fuses the multi-scale convolutional network and the time sequence reasoning unit, and performs high-dimensional alignment and modeling on the double-source data to form a multi-resolution sensing field. And the intelligent abnormity feedback module generates a response grade score, and realizes millisecond-level control mapping with the linkage control module based on a state path. The method has the beneficial effect that the accuracy and response speed of abnormal recognition in the manufacturing process are improved.

The invention relates to the technical field of land utilization change evaluation, in particular to a land change degree evaluation method and system based on gradient division and two-dimensional analysis and a storage medium. Dividing the to-be-evaluated land into a plurality of target sub-regions with different target gradients by adopting topographic position indexes; wherein the terrain position index is obtained by calculating an elevation value and a slope value of the land; and analyzing the target sub-regions from two evaluation dimensions including a terrain gradient effect dimension and a land gradient utilization characteristic dimension, and determining a land change degree evaluation result of the to-be-evaluated land based on the obtained terrain gradient effect evaluation result and the land gradient utilization characteristic dimension evaluation result of the plurality of target sub-regions. The objective of the invention is to solve the problem of how to carry out land utilization change evaluation on areas with obvious vertical zone performance and altitude gradient performance.

The invention relates to the technical field of landslide geological recognition, in particular to a reservoir area landslide geological disaster recognition method and system, and the method comprises the steps: collecting SAR images and DEM data of a reservoir area at each time point; forming the SAR images of the adjacent time points into master and slave SAR images, obtaining an interferogram of the master and slave SAR images, and dividing the interferogram of the master and slave SAR images into an incoherent water area, a terrain transition area and a high-coherence stable area; determining a terrain phase gradient of each pixel point in the interferogram, obtaining a terrain gradient amplitude of each pixel point, correcting an objective function of a least square phase unwrapping algorithm, and obtaining an interferometric phase after phase unwrapping; elevation information corresponding to the interferogram is determined, the average deformation rate of each pixel point is calculated, and the potential landslide risk of the reservoir area corresponding to each pixel point is identified. Therefore, the accuracy of reservoir area landslide risk identification is improved.

The invention relates to the technical field of topographic survey, in particular to an InSAR earth surface deformation information detection method and system based on an unmanned aerial vehicle, and the method comprises the following steps: S1, collecting point cloud data of a mining area through a laser radar system carried by the unmanned aerial vehicle, constructing a continuous topographic curved surface, calculating a topographic gradient vector based on the curved surface, and extracting a topographic main gradient direction angle; and S2, based on the terrain gradient vector, the rock mass integrity coefficient of the mining area and the mining direction angle, constructing a physical constraint field for quantifying the terrain sudden change risk and the mining and terrain interaction effect. According to the method, a physical field capable of quantifying fault risks and types is generated by fusing unmanned aerial vehicle laser radar topographic data, a rock mass integrity coefficient and a manual mining direction; according to the method, geological information is changed into beforehand constraint from post-event interpretation, an accurate physical basis is provided for subsequent phase processing, and the situation that the deformation direction is misjudged, for example, settlement is misjudged as uplift, is prevented.

The invention discloses a PSO-improved three-dimensional sensor network optimization deployment method. The scheme comprises the following steps: constructing a three-dimensional space monitoring model fusing terrain elevation and obstacle distribution; designing a dynamic multi-target fitness function fusing the space coverage rate, the energy consumption balance degree and the obstacle penetration loss coefficient; dynamically adjusting the search weight of the particle swarm in the Z-axis direction according to the terrain gradient; dividing a three-dimensional space into sub-regions and initializing a particle swarm; and iteratively optimizing the node position by adopting an improved PSO algorithm until a termination condition is met. The method has the advantages that the method is suitable for scenes such as mountain monitoring, unmanned aerial vehicle cluster networking and ocean three-dimensional monitoring, the three-dimensional coverage rate can be increased by 15%-30%, the network service life can be prolonged by 20% or above, real-time node redeployment in a dynamic environment is supported, and through innovative three-dimensional space modeling, a multi-target optimization algorithm and a parallel computing architecture, the real-time node redeployment can be realized. And comprehensive improvement of the deployment quality and efficiency of the sensor network in a complex environment is realized.

The invention relates to the technical field of urban flood control and drainage, and discloses an urban inland inundation rainfall flood data assimilation method and system and a storage medium. The method comprises the steps of obtaining rainfall, flow and surface elevation data from multi-source hydrological observation data, and obtaining a multi-source data set after preprocessing; hydrological parameters are extracted based on the data set, and the influence weight of the hydrological parameters on the submerging range is analyzed and determined in combination with parameter sensitivity; updating the key hydrological parameters through a filtering algorithm to obtain a refined hydrological parameter set; a terrain gradient analysis and prediction algorithm is combined to establish a prediction model of a submerging boundary change rate, and an inversion time interval is dynamically adjusted; the surface elevation is corrected in real time through a finite element method, and the submerging prediction precision is improved; and further optimizing the hydrological parameters according to the real-time prediction result, and generating a drainage scheduling decision instruction. The method has high accuracy and real-time performance, scientific and effective decision support can be provided for an urban flood control and drainage system, and the emergency response capability is remarkably improved.

The invention relates to the technical field of machine vision and industrial automatic detection, in particular to a workpiece defect detection system and method based on image recognition. An image mapping module, a flow field initialization module, a dynamic evolution module, a divergence identification module and a feedback optimization module are included; the system maps the gray scale into a terrain gradient potential energy field, and establishes a virtual flow field conforming to the texture; the core of the method is that dynamic evolution is carried out based on a Navier-Stokes equation, defects are converted into physical obstacles inducing flow field disturbance, and a defect area is judged according to divergence values; dynamically adjusting an evolution parameter according to the verification signal; according to the method, conversion from pixel statistics to physical simulation is realized, and the problem that weak defects are difficult to extract under a complex background is solved.

The invention discloses a U-Net loess plateau terrace boundary refined extraction method fusing terrain gradient constraints. The method comprises the steps of obtaining and preprocessing a remote sensing image, a DEM and meteorological data; extracting terrain gradient and texture morphological characteristics, and generating environment state codes; calculating feature weights based on environment states, and fusing to form a feature set; generating a region priori image and a region division mask by using DEM and environment coding; inputting the fusion features into a U-Net network, and enhancing boundary features through an attention guidance mechanism; and carrying out boundary regularization by utilizing slope direction consistency constraint, and outputting a terrace boundary. The problem of terrace boundary confusion under the loess plateau complex terrain condition is effectively solved, and the boundary precision and the environmental adaptability are improved.

The invention discloses a land area measurement method and system for land planning and design, and the method comprises the steps: collecting a sequence aerial image of a target land parcel, and synchronously recording the positioning and attitude determination data of each frame of image; edge sharpening processing and multi-view three-dimensional reconstruction are carried out on the sequence aerial image, and a digital elevation model with texture features is generated; performing terrain gradient analysis and edge contour extraction based on a digital elevation model, and identifying a land block boundary through an adaptive threshold segmentation algorithm and generating an initial vector boundary diagram; performing topological relation verification and fragment pattern spot fusion on the initial vector boundary diagram, and outputting an optimized plot vector boundary; and carrying out coordinate registration on the optimized plot vector boundary and a geographic information system base map, calculating a projection area through integral operation, and generating a final measurement report containing a topographic relief coefficient. According to the embodiment of the invention, the operation efficiency, boundary recognition precision and result reliability of land area measurement under complex terrains can be improved.

The invention discloses an adaptive multi-loop network InSAR phase unwrapping method and device, and relates to the technical field of interferometric synthetic aperture radar measurement. The method comprises the following steps: acquiring an interferometric phase diagram and a topographic map of a complex topographic region; calculating coherence, a topographic gradient map, a phase gradient standard deviation and residual density based on the interferometric phase map and the topographic map; based on the coherence, the terrain gradient map, the phase gradient standard deviation and the residual density, dynamically selecting the size of a ring through a decision function, and constructing a multi-ring network; determining the connectivity of the multi-ring network according to a bridging strategy and a connectivity detection method, and outputting a global connected network; on the basis of the global connected network, layered residual optimization is carried out by adopting a local absorption and global planning double-layer strategy, and an optimized residual is output; and based on the optimized residual error, unwrapping all rings in the global connected network through an in-ring local unwrapping and global optimization strategy, and outputting an unwrapping result. According to the invention, the problem of network connectivity in a complex area can be solved.

The invention discloses a power transmission line landslide disaster susceptibility evaluation and reinforcement measure decision-making method, which comprises the steps of obtaining geological environment factor data of a plurality of evaluation units in a target power transmission channel, and calculating a susceptibility index of each evaluation unit based on an information amount model, the geological environment factor data comprises landform types, topographic gradients, structural stability, geological rock groups, vegetation coverage conditions and human activity influences; performing susceptibility level zoning on the target power transmission channel according to the susceptibility index to generate a geological disaster susceptibility zoning map; for each divided susceptible grade area, analyzing main geological environment factors contributing to susceptibility indexes, and identifying main disaster causing factors of the area; matching and generating one or more targeted reinforcement scheme sets from a pre-established reinforcement measure knowledge base on the basis of the susceptible grade area and the identified main disaster-causing factors; wherein the reinforcement measure knowledge base comprises a structural reinforcement measure, an ecological slope protection measure, a drainage measure and a combination scheme thereof.

This application relates to the technical field of land use change evaluation, and particularly to a land change degree evaluation method, system and storage medium based on gradient division and two-dimensional analysis. The land to be evaluated is divided into target sub-regions with multiple different target gradients by using the topographic position index; wherein, the topographic position index is calculated from the elevation value and slope value of the land; the target sub-regions are analyzed from two evaluation dimensions, namely the topographic gradient effect dimension and the land gradient utilization characteristic dimension, and based on the evaluation results of the topographic gradient effect and the evaluation results of the land gradient utilization characteristic dimension of the obtained multiple target sub-regions, the evaluation result of the land change degree of the land to be evaluated is determined. It aims to solve the problem of how to evaluate land use change in areas with obvious vertical zonality and altitude gradient.

The invention discloses a three-dimensional power grid training simulation method based on a virtual reality technology, and relates to the technical field of power grid training simulation, and the method comprises the following steps: initializing a virtual reality environment, loading a mountainous area power grid three-dimensional model, and generating a power transmission line node; setting training environment parameters; calculating an effective wind speed at each line node according to the terrain gradient; calculating the line galloping amplitude in real time; dynamically rendering a line galloping effect in a VR environment, and visually displaying a dangerous section through color coding; students can deploy anti-galloping devices interactively, and the system updates the damping effect in real time and recalculates the galloping amplitude. The system continuously monitors the safety state, provides a real-time alarm and generates a training evaluation report. According to the invention, a terrain-adaptive galloping amplitude real-time calculation algorithm is introduced, and a virtual reality training system deeply fused with a real geographical environment is constructed, so that simulation and visualization of the galloping behavior of the high-voltage transmission line in a mountainous complex terrain can be realized.

The invention discloses a shallow low-temperature hydrothermal deposit modeling method based on short-wave infrared, and relates to the technical field of artificial intelligence mineral product prediction, and the method comprises the steps: carrying out the weighted normalization processing of multi-source heterogeneous geological data, mapping the multi-source heterogeneous geological data to a unified feature space through a nonlinear fusion coding algorithm with a weight coefficient, and carrying out the modeling of the multi-source heterogeneous geological data. Generating a fusion feature vector; constructing a self-attention mechanism perceived by a geological structure, introducing a geological prior matrix reflecting terrain gradient and geochemical abnormal distribution on the basis of a key, query and value matrix, and dynamically adjusting local geological coupling weight parameters in an attention calculation process; and establishing a geological consistency loss mechanism, performing spatial constraint on the attention response and the alteration label, forming a composite loss structure through a nonlinear amplification term, a logarithmic penalty term and a smooth boundary constraint term, and enabling the geological consistency loss mechanism, classification loss and a regular term to participate in optimization together.

The invention belongs to the technical field of ecological restoration, and particularly relates to landform gradient compliance mine landscape ecological restoration equipment which comprises a frame, connecting rods, butt joint blocks and connecting holes. The butt joint assembly is arranged at one end of the frame; the fixing mechanisms are arranged at one end of the frame in an array distribution manner; the water inlet pipe is connected to the frame, and the frame is further connected with a water outlet pipe; wherein the fixing mechanism comprises an anchor rod, the anchor rod is connected to the frame in an inserted mode, a guide seat is fixedly connected to the interior of the anchor rod, a vertical rod is connected to the guide seat in a sliding and inserted mode, through holes are formed in the anchor rod in an array distribution mode, reinforcing parts are arranged in the through holes, and abutting parts are connected to the vertical rod in an array distribution mode. And a detection part is arranged at the lower end of the anchor rod. The system can flexibly meet the repair requirements under different topographic conditions, integrates multiple functions, and ensures the stability, safety and high efficiency in the repair process.

A ground vehicle navigation system includes an inertial navigation system arranged to output an orientation estimate of the ground vehicle and a first position estimate of the ground vehicle, a terrain map comprising terrain data, a terrain gradient based navigation unit arranged to output a second position estimate of the ground vehicle based on a comparison between the orientation estimate and terrain gradient data extracted from the terrain map and an iterative algorithm unit arranged to determine a system error state in each iteration. In each iteration the iterative algorithm unit is arranged to receive the first position estimate and the second position estimate; and update the system error state for the next iteration based on the system error state, the first position estimate and the second position estimate. The iterative algorithm unit may then apply the updated system error state to the INS measurements.

The present application relates to the technical field of virtual simulation and fluid modeling, in particular to a sea current prediction and visualization method based on digital twinning, comprising the following steps: S1, setting a boundary water level and a terrain elevation to calculate an initial flow velocity field, and generating a projection velocity field satisfying an incompressible condition through projection correction; S2, generating a disturbance mapping by combining a difference value of the projection field and the initial field, a terrain gradient and a boundary change rate, and marking a potential visual discontinuous area; S3, constructing a terrain-sensitive disturbance factor to modulate a velocity input, driving a wave and particle system, and realizing smooth transition and consistent evolution of a physical field and a visual rendering system. The present application realizes the fusion effect of high consistency between physical precision and visual performance of sea current simulation under complex terrain by constructing a disturbance-modulated physically consistent velocity field and driving the wave and particle system to evolve cooperatively.

The invention discloses a karst complex terrain-oriented unmanned aerial vehicle precipitation enhancement data fusion quality control method, and relates to the technical field of electrical digital data processing, and the method comprises the steps: preprocessing multi-source data to form a standardized data set, extracting landform feature points to construct a terrain constraint space-time reference, and dividing terrain gradient partitions to carry out space-time registration. Weighting the fusion data by using a terrain-weather coupling weight algorithm, carrying out point-by-point quality control and exception processing on the fusion data by using a three-dimensional quality control threshold adaptive algorithm, and finally outputting the fusion data; according to the method, precise space-time alignment is realized by constructing the terrain constraint reference and gradient partition, and the landform feature recognition and the adaptive quality control threshold are integrated, so that the dependence on satellite signals is eliminated, the space consistency and the time continuity of the fused data are improved, the problems of space-time dislocation and fusion deviation caused by the terrain are effectively solved, and the accuracy and the accuracy of data fusion are improved. High-quality data support is provided for the rainfall enhancement operation of the unmanned aerial vehicle, and the scientificity and effectiveness of the rainfall enhancement operation are enhanced.

The invention discloses a farmland nitrogen and phosphorus pesticide dynamic prediction method and system based on an event trigger graph neural network. The method belongs to the technical field of pesticide intelligent prediction, and comprises the following steps: firstly, acquiring a farmland management record, weather and DEM data of a target area, and constructing an initial space diagram structure which takes a field parcel as a node, arranges edges according to geographical adjacency and determines the initial weight of the edges by topographic gradient and soil permeability; building a fertilization-rainfall-runoff ternary event detector, and triggering high-frequency prediction if effective rainfall is monitored in the preset window after fertilization or pesticide spraying; then updating the edge weight of the graph and reconnecting the topology; and inputting the two-channel graph neural network model, and finally outputting a nitrogen and phosphorus pesticide leaching loss spatial distribution thermodynamic diagram in a future preset time period to complete dynamic prediction.

The present application relates to the technical field of topographic survey, in particular to an InSAR ground deformation information detection method and system based on a UAV, comprising the following steps: S1: collecting point cloud data of a mining area by a laser radar system carried by a UAV, constructing a continuous terrain surface, and calculating a terrain gradient vector based on the surface to extract a main terrain gradient direction angle; S2: constructing a physical constraint field quantifying the terrain mutation risk and the interaction effect of mining and terrain based on the terrain gradient vector, the rock mass integrity coefficient of the mining area, and the mining direction angle. The present application generates a physical field quantifying the fault risk and type by fusing the UAV laser radar terrain data, the rock mass integrity coefficient, and the artificial mining direction; geological information is changed from post-explanation to prior constraint, providing accurate physical basis for subsequent phase processing and preventing misjudgment of the deformation direction, such as misjudging subsidence as uplift.

The invention relates to the field of territorial space planning, and particularly discloses a territorial space planning method, device, equipment and medium based on a GIS (Geographic Information System), comprising the following steps: S1, constructing a standardized planning data acquiring and preprocessing multi-source territorial space data, and storing the multi-source territorial space data into a space database according to a preset data standard and a field structure, the data at least comprises current land, topographic gradient, elevation, geological disasters, basic farmland, ecological protection red lines and urban development boundaries; according to the method, a standardized multi-source spatial database and a computer executable rule base are constructed, and complex territorial space planning laws and regulations are converted into quantifiable and computable spatial logic rules, so that the planning analysis process is changed from traditional experience-dependent judgment to data-based automatic processing; through automatic checking of a rigid constraint base line and automatic generation of planning partitions, the system can efficiently identify pattern spot level rule conflicts.

The invention provides a forest coverage evaluation detection method and system based on remote sensing image detection, and relates to the technical field of forest coverage evaluation detection, and the method comprises the steps: collecting the landform gradient, elevation and remote sensing image data of a target region, and carrying out the grading of the landform gradient, thereby obtaining gradient interval data; performing radiometric calibration and topographic correction on the remote sensing image to obtain an image data set; segmenting the image based on a preset reflectivity threshold to obtain a forest candidate region; fusing the data with a ground actual measurement data set to obtain coincidence degree data; and finally, performing precision correction on the candidate region in combination with gradient interval data to obtain forest coverage and coverage data, and completing evaluation, so that high-precision automatic evaluation of the coverage of the contiguous forests in the plain region can be realized by fusing remote sensing images, topographic gradients and ground measured data.

The invention discloses a comprehensive meteorological calculation method, device and equipment considering a complex terrain in regional power load prediction and a storage medium, and relates to the technical field of computers, and the method comprises the steps: determining an altitude average value, an altitude standard deviation value and terrain gradient information based on the elevation grid data of a target region; determining a gradient mean value and a topographic relief mean value based on the topographic gradient information; determining a feature matrix based on the altitude average value, the altitude standard deviation value, the gradient average value and the topographic relief average value, and processing the feature matrix by using a preset clustering algorithm and a preset evaluation index to obtain a clustering result; and determining regional meteorological information and regional weight based on the population density and power consumption of each region in the target region and each terrain in the clustering result, determining target meteorological information based on the regional weight and the regional meteorological information, and performing power consumption load prediction on the target meteorological information by using a preset power consumption load prediction model, and obtaining an electrical load prediction result. Therefore, the accuracy of predicting the electrical load can be improved.

This invention discloses a quality control method for UAV rain enhancement data fusion in complex karst terrain, belonging to the field of electronic digital data processing technology. This method preprocesses multi-source data to form a standardized dataset, extracts landform feature points to construct a terrain-constrained spatiotemporal benchmark, divides the terrain gradient into zones for spatiotemporal registration, applies a terrain-meteorological coupled weighting algorithm to weight and fuse the data, and utilizes a three-dimensional quality control threshold adaptive algorithm to perform point-by-point quality control and anomaly processing on the fused data, finally outputting the fused data. This invention achieves precise spatiotemporal alignment by constructing a terrain-constrained benchmark and gradient partitioning, integrates landform feature recognition and adaptive quality control thresholds, eliminates dependence on satellite signals, improves the spatial consistency and temporal continuity of the fused data, effectively solves the problems of spatiotemporal misalignment and fusion deviation caused by terrain, provides high-quality data support for UAV rain enhancement operations, and enhances the scientific rigor and effectiveness of rain enhancement operations.

The invention discloses a satellite landslide data generation method and device based on a multi-scale adversarial neural network, equipment and a medium, and relates to the technical field of satellite landslide disasters. Constructing an adaptive attention fusion mechanism of a terrain gradient field, a curvature field, a catastrophe intensity field and a thermal anomaly field, and generating multi-source disaster sensing features; a multi-scale generator is combined with DEM multi-scale features to generate images step by step, combined discrimination of optical authenticity and terrain rationality is carried out through a multi-task discriminator, and finally target satellite landslide data is obtained. Through accurate alignment of the multi-source disaster sensing data and the optical features, the satellite landslide data with the rationality of the terrain structure and the authenticity of the optical texture is generated, and the precision and the reliability of the generated satellite landslide data of the complex landform are improved.

This application discloses a method, apparatus, device, and medium for generating satellite landslide data based on a multi-scale adversarial neural network, relating to the field of satellite landslide disaster technology. The method includes: fusing optical remote sensing images, DEM and auxiliary disaster perception data such as InSAR deformation and thermal infrared data to construct an adaptive attention fusion mechanism for topographic gradient field, curvature field, disaster intensity field, and thermal anomaly field, generating multi-source disaster perception features; using a multi-scale generator combined with DEM multi-scale features to generate images step-by-step, and using a multi-task discriminator to jointly judge optical authenticity and topographic plausibility, ultimately obtaining target satellite landslide data. Through precise alignment of multi-source disaster perception data and optical features, satellite landslide data with both topographic structural plausibility and optical texture authenticity is generated, improving the accuracy and reliability of the generated complex terrain satellite landslide data.