44 results about "Spatial model" patented technology

The invention provides a traffic flow prediction method combining a dynamic graph neural network and a Mama mechanism. The future short-term traffic flow is predicted by using historical traffic data. According to the method, firstly, normalization preprocessing is carried out on traffic state data collected by multiple sensors, a training sample is generated by adopting a sliding window, and a traffic flow value in the next one hour is predicted according to data in the past 24 hours. On the basis of the model structure, a time modeling module composed of multiple layers of MambaBlocks is constructed and used for capturing historical time sequence dependence; constructing a spatial modeling module of dynamic graph convolution, and combining a static adjacency matrix of the road network with a learnable adaptive adjacency structure to extract spatial association; and finally, the outputs of the modules are fused, and a prediction result is obtained through a prediction output module. In the training process, a Huber loss function is used as an optimization target, and evaluation indexes such as a mean absolute error (MAE), a root mean square error (RMSE) and a mean absolute percentage error (MAPE) are used for evaluating the performance of the model. According to the method, the traffic space-time dynamic characteristics are effectively mined, and the long-range dependence modeling capability and the prediction precision are improved.

The invention discloses a foggy day real-time target detection method and system based on Fourier transform image defogging, and relates to the technical field of image target detection. The method comprises the following steps: preprocessing data, normalizing the size of an image, implementing a multi-dimensional image enhancement strategy, and designing a self-adaptive amplitude enhancement module based on fast Fourier transform to carry out defogging processing on the image; feature extraction: inputting the preprocessed image into a backbone network to generate feature maps with different scales, and enhancing the spatial modeling capability of the feature maps by a spatial modeling SimAM module by using a SimAM attention mechanism; feature fusion: adding a weight convolution pyramid module into the neck network to form a feature pyramid architecture, and performing multi-scale feature fusion operation on different-scale feature maps to obtain multi-scale semantic features; information post-processing decoding and prediction are carried out, after the multi-scale semantic features are decoupled through a detection head, two independent tensors of classification and positioning are generated, and tensor information is analyzed to obtain a detection result. The accuracy of the detection model is improved.

The invention relates to an urban green land high-resolution remote sensing monitoring method. The method comprises the following steps: data acquisition and preprocessing: synchronously acquiring point cloud data, hyperspectral data and multi-temporal remote sensing data; feature engineering: constructing a multi-dimensional feature space, and optimizing the multi-dimensional feature space through feature selection; model training: on the basis of the optimized multi-dimensional feature space, performing filling, differential normalization processing and regularization improvement on a U-net + model through pixel values of the edges of mirror images in sequence for urban green land features, and establishing the urban green land-oriented and multi-dimensional feature-based U-net + model, so that the urban green land-oriented and multi-dimensional feature-based U-net + model is obtained; adding the multi-dimensional feature data of the training sample into the U-net + model for training, predicting the spatial distribution of the urban green land after the training is completed, obtaining a prediction result of the U-net + model, and combining a mixed loss function with an adaptive training strategy; and post-processing and result generation: applying adaptive CRFs to optimize a U-net + model prediction result, generating a high-resolution urban green land distribution map, and outputting a precision evaluation report.

The application belongs to the technical field of height data analysis, and provides a DEM space model height point automatic extraction method, the application carries out clustering on the height data of sampling points by setting different clustering radii, judges the terrain integrity of each clustering range, and divides subspaces; analyzes the reference value of different subspaces on the height interpolation of the projection position point, obtains the reference height value of the projection position point; then analyzes the distance interval condition of each height data in the same region in different subspaces, adjusts the coordinate position of the projection position point, and obtains the corrected DEM space model; finally, the difference condition between the water flow trend and the actual terrain in the corrected DEM space model is analyzed, the reference height value of part of the projection position points is adjusted, and the twice corrected DEM space model is obtained. The accuracy of the interpolation height points of different positions is improved by the method, and the accuracy of the DEM space model construction is improved.

The invention relates to the technical field of data security management, and discloses a security management system for data access control. The system comprises an access request receiving unit, a hierarchical analysis unit, a space mapping unit, a permission decision unit and a permission execution unit. The access request receiving unit obtains a data access request of a user, and analyzes a request main body identifier and a request resource identifier. The layering analysis unit carries out layering processing on the resource identification and extracts resource level feature information. And the space mapping unit retrieves the historical access log according to the main body identifier, and constructs an access behavior space distribution model. And the authority decision-making unit generates a dynamic access authority strategy in combination with the resource level characteristics and the access behavior space model. And the permission execution unit controls the access operation of the user to the resources according to the strategy. The system can adapt to a complex data resource structure and realize fine-grained dynamic access control.

This invention belongs to the field of campus management and discloses a method for monitoring campus personnel behavior trajectories based on multi-source data fusion, including the following steps: acquiring multi-source data within the campus; preprocessing the multi-source data using a data weighted fusion algorithm; performing differentiated scene optimization processing on the collected facial images to output clear facial image frames and effective feature data; constructing real-time behavior trajectories of campus personnel based on a temporal modeling model using clear facial image frames, effective feature data, and preprocessed multi-source data; analyzing the real-time behavior trajectories through an anomaly detection mechanism to determine the trajectory anomaly status and classify the anomaly level; executing multi-channel alarm push according to the anomaly level; visually restoring the real-time behavior trajectories and historical behavior trajectories based on a campus spatial model; and absorbing new data and anomaly judgment results through an incremental learning algorithm to update the parameters of the temporal modeling model, anomaly detection mechanism, and differentiated scene optimization processing related models.

A method for generating a spatial model tile data organization structure, relating to the field of geographic information technology, is characterized by the following steps: generating tile network leaf nodes; generating tile network intermediate nodes; selecting an initial aggregation point for the tile leaf nodes and performing multiple aggregation operations with each leaf-level tile until all leaf-level tiles are aggregated to obtain multiple tile aggregates; binding the tile network leaf nodes to their parent tile network intermediate nodes based on the correspondence between the tile network leaf nodes and the aggregated tile network intermediate nodes, ensuring that each tile network intermediate node has a logical relationship with its child nodes; and generating a tile network root node. Advantages: Effectively generates a multi-level detailed tile organization structure for model data in a given data block; the level of detail in each layer of tile data is reasonable, and the data organization is efficient; the data storage scale of a single tile is low and stable, saving network bandwidth and improving file transfer efficiency.

The invention provides a logging analysis model construction method and system based on multilevel regional feature association, and belongs to the technical field of oil-gas exploration artificial intelligence. In order to solve the problems of high cost, poor generalization and the like of a conventional method for performing data correction by using a standard well, the method comprises the following steps of: firstly, constructing multi-level geographic region embedding according to well location hierarchical codes, and capturing regional geologic features; then, constructing a regional relation graph based on well location space coordinates; and finally, aggregating features of adjacent regions through a graph neural network, and fusing multi-level region features and logging data through gating weight. According to the method, automatic learning of regional features is realized through a technology of performing gating weight fusion on the logging data by using the multistage regional nesting embedding module, the graph neural network spatial modeling module and the logging data, dependence and cost on standard well correction are removed, and cross-regional prediction precision is remarkably improved by establishing the regional features.

The invention relates to the field of intelligent video monitoring and early warning, and discloses a water area video intelligent analysis and early warning method based on an AI intelligent agent, which comprises the following steps: acquiring unmanned aerial vehicle aerial survey data of a target river channel to construct a river channel space model, identifying a monitoring area including a first type and / or a second type of risk areas, establishing a water area video monitoring system corresponding to the monitoring area, the edge computing device and the camera identifier; generating an unmanned aerial vehicle inspection strategy by combining the hydrological season state, and enabling the unmanned aerial vehicle to inspect according to the strategy; when abnormal event indication information is received, triggering an edge computing device to carry out AI analysis on camera videos and unmanned aerial vehicle real-time data to obtain target characteristic parameters, and dividing events into low, medium or high risks according to the target characteristic parameters; and performing selective storage and scheduling on the video data of the low-risk events and the medium-risk events, and generating early warning for the high-risk events and pushing the early warning to the management terminal. Therefore, graded monitoring and graded treatment of the key risky areas of the riverway are realized.

This invention relates to the field of machining and intelligent control technology, and discloses a multi-process collaborative control system for machining equipment based on multi-source data. The system includes a spatial modeling module, a multi-source sensing and feature mapping subsystem, a reference calibration module, and a collaborative execution module. The spatial modeling module constructs a hierarchical partitioned spatial mesh model enveloping the workpiece and drives dynamic node growth. The multi-source sensing and feature mapping subsystem synchronously acquires operating status data, converts it into spatial attribute values, and stores them in voxel units. After the workpiece is transferred to a new process, the reference calibration module constructs a cross-process coordinate transformation matrix and performs pose transformation on the spatial attribute values ​​stored in the model. The collaborative execution module generates machining compensation increments based on the transformed spatial attribute values ​​and injects them into the CNC system. This invention solves the problem of physical feature discontinuity in multi-process flow, realizing spatial mapping and adaptive compensation from previous physical states to the current process.

The application discloses a terrain self-adaptive rainstorm waterlogging spatio-temporal joint prediction method and system, and belongs to the field of alarm responding to disaster events. In order to overcome the defects of the existing method, such as insufficient terrain adaptability of fixed feature weight, and prediction fragmentation caused by independent operation of the spatio-temporal model, the technical scheme of the application is as follows: on the one hand, a terrain self-adaptive dynamic feature weight mechanism is constructed, an adjustment coefficient is generated by quantifying the terrain undulation, and the feature factor weight of the geographical spatio-temporal weighted regression model is dynamically adjusted to adapt to the complex underlying surface; on the other hand, a spatio-temporal joint weighted fusion framework is proposed, a random forest spatial model and a PredFormer time series model are combined, the spatial prediction result is taken as the initial condition of the time series, and the time series collaborative prediction of the flooded range and the water depth is realized. The application is suitable for rainstorm waterlogging prediction of different terrains in cities, and has the advantages of strong terrain adaptability, high prediction accuracy and stability, and unified spatio-temporal rapid response.

The invention provides a payment flow prediction method and device, electronic equipment and a storage medium, and is applied to the technical field of artificial intelligence. The method comprises the following steps: acquiring multi-dimensional time sequence data of life payment of a target area through a life payment system within a set historical duration; performing preprocessing of corresponding dimensions on each dimension of time series data in the multi-dimension time series data to obtain feature vectors of the corresponding dimension of time series data; according to a preset splicing sequence, sequentially splicing the feature vectors of the time sequence data of the corresponding dimensions to obtain multi-dimensional feature vectors; and the multi-dimensional feature vector is inputted to a time specialized model for fee-paying flow prediction, the predicted fee-paying flow of the target area in the future duration is obtained, and the time specialized model is obtained by removing a spatial modeling part irrelevant to fee-paying flow prediction in a space-time diagram convolutional network model based on geographical island characteristics. The method is used for achieving the effects of accurately predicting the payment flow and further accurately capturing the payment flow peak value.

An information processing system, method, and program are provided that evaluate the feasibility of a design system from the perspective of both spatial simulation and system specifications, and grasp the evaluation results in the context of their relevance to the specifications. [Solution] In an information processing system, information processing by an information processing device includes acquiring (S1) spatial model information and design information from a user terminal. The spatial model information describes a physical model in a virtual space of a design system to be designed and includes information on physical quantities that can be calculated in the virtual space, and the design information indicates specifications of the design system, the specifications being set by a user who designs the design system. The information processing also includes associating (S5) the spatial model information with the design information, calculating (S9) an evaluation index related to the feasibility of the specifications in the physical model based on the result of the association, and outputting (S10) the calculated evaluation index in association with the specifications included in the design information.

The application discloses a method for improving transient stability of a network-forming converter in an islanded alternating-current microgrid, and the method comprises the following steps: building a spatial model of the islanded alternating-current microgrid based on a model predictive control strategy, and adding an additional adjustment term in a Q-V loop of a virtual synchronous generator control mode; when a grid-connected converter generates a frequency deviation due to a fault of the microgrid, an electromagnetic torque target function minimization method is used to rollingly optimize the additional adjustment term in the Q-V loop according to angular frequency data at each moment, so that an equivalent input torque and an output torque of the network-forming converter are balanced, and thus the stability of the network-forming converter is restored. The application can effectively reduce the frequency deviation and system oscillation after the fault, significantly shorten the system recovery time, and improve the operation reliability and stability of the microgrid.

This application provides a fire lane occupancy identification system and method using laser and radar collaboration, relating to the field of laser radar technology. The system includes: a point cloud generation module, used to emit laser beams to the target area using a solid-state laser radar, receive reflected signals, extract time difference and phase difference, convert them into polar coordinate data, and generate a point cloud dataset; a 3D spatial model construction module, used to perform point cloud filtering, clustering, and 3D reconstruction based on the point cloud dataset to construct a 3D spatial model of the target area; and an occupancy decision module, used to calculate the space occupancy rate based on the 3D spatial model, and when the detected space occupancy rate exceeds a preset dynamic threshold, to perform adaptive occupancy decision and push alarm information. This application can solve the technical problem of low occupancy detection accuracy in existing technologies, achieving the technical effects of improving detection stability, reducing false alarm rate, and enhancing system adaptability and linkage response capabilities.

The invention relates to a multi-view combined tracking method for a complex target in an industrial production scene, and the method comprises the following steps: S1, carrying out the calibration of internal parameters and external parameters of all cameras, constructing a physical space 3D model of a target production region, and forming a factory coordinate system; s2, collecting a multi-view image frame through the calibrated camera, and carrying out the preprocessing of the multi-view image frame; s3, detecting the position of each frame of target by using a deep learning detection model Detectron2, performing monocular inter-frame association on the detected target to obtain a local trajectory, and extracting a depth feature vector for each target; s4, when the same target appears at a plurality of visual angles, fusing the space coordinates to obtain accurate 3D positioning, and obtaining a physical 3D space coordinate candidate set of each actual target; and S5, according to the physical 3D space coordinate candidate set, constructing a matching candidate set, and adopting a joint matching algorithm to realize trajectory merging to obtain a complete 3D trajectory. According to the invention, multi-target efficient and multi-view collaborative accurate tracking in a complex industrial scene is realized.

This invention discloses a method for managing the three-dimensional spatial layout of a data center based on digital twins, comprising the following steps: collecting data center equipment operation data, environmental monitoring data, and maintenance behavior data to generate a multi-source time-series dataset of the data center; constructing a three-dimensional spatial model of the data center to generate a three-dimensional spatial digital twin of the data center; constructing a behavior trajectory sequence and statistically analyzing the access frequency and path congestion of spatial units to generate a set of spatial behavior distribution characteristics; calculating the behavior friction coefficient and path efficiency loss to generate spatial layout evaluation results; allocating space occupancy rights to data center equipment and establishing occupancy conflict relationships; determining the evolutionary stability state of the occupancy right allocation process to determine the optimal layout scheme; and generating data center spatial layout adjustment instructions to achieve dynamic optimization management of the three-dimensional spatial layout of the data center. This invention utilizes digital twins and behavior modeling methods to achieve dynamic optimization of the data center spatial layout, possessing the advantages of controllable conflicts and high resource utilization.

This invention discloses a multi-objective dynamic evacuation path planning method for floods at multi-line transfer stations in subways. The method includes: establishing a three-dimensional spatial model by acquiring subway station structural information and abstracting it into a network representation; acquiring real-time hydrological information using hydrological sensors and video surveillance, preprocessing and fusing it to generate hydrological perception data under a unified spatiotemporal reference; acquiring real-time crowd information within the station through video target detection, and calculating crowd density and passage capacity to form crowd flow situation perception data; combining hydrological conditions, crowd flow situation, and passability constraints, using a real-time path search algorithm to generate multiple candidate evacuation paths for different types of passengers; classifying evacuation paths according to passenger type, and issuing evacuation guidance through multi-terminal channels to achieve differentiated dynamic guidance for exiting passengers, transferring passengers, and passengers with mobility impairments. This method can respond in real-time to flood disasters and changes in crowd flow, improving the evacuation efficiency and safety of subway transfer stations.

This invention discloses a 3D modeling and intelligent scheduling system for urea warehouses, belonging to the field of intelligent modeling and scheduling technology for urea warehouses. Specifically, it includes: real-time acquisition of internal images of the urea warehouse, preprocessing the acquired images, using computer vision technology to generate 3D point cloud data of the urea warehouse, constructing a 3D model of the urea warehouse, building a spatiotemporal heat map, and intelligently generating a scheduling plan based on real-time inventory status, warehouse space usage, task status, and transportation needs; real-time scanning and analysis of the spatial layout, item location, and type of the urea warehouse, enabling dynamic generation of a 3D spatial model of the urea warehouse based on real-time acquired data. The constructed spatiotemporal heat map can detect obstacles in advance and intelligently plan the operating path of the robotic arm, eliminating the need to adjust or replan the robotic arm's path based on obstacles, saving handling time and greatly improving the operating efficiency of the robotic arm.

The present application belongs to the technical field of smart city, and particularly relates to a smart city environment management method and system based on digital twinning. The method not only realizes high-precision collection of species distribution, vegetation coverage, ground temperature and wind speed data in the urban area, but also successfully simulates the dynamic changes of ground thermal radiation absorption and release caused by the increase and decrease of different species by constructing a dynamic spatial model containing species-environment interaction. Further, by calculating the hourly influence coefficient and iteratively deriving the microclimate evolution trend in a ten-year period, a priority ranking scheme for species protection or regulation is finally generated. This method greatly improves the understanding depth and response speed of urban managers on ecological environment changes, makes real-time calibration based on the latest monitoring data possible, and ensures the scientificity and foresight of urban management decisions.

This invention discloses an automatic identification method and system for mural protection and restoration based on image analysis. The method includes: generating the degradation state of the mural at different time points using an image simulation algorithm; converting two-dimensional damage images into a three-dimensional spatial model using 3D mapping technology to accurately pinpoint high-risk damage points; further integrating scattered damage points using a clustering analysis algorithm to extract the properties and spatial characteristics of damage groups; and automatically identifying damage boundaries by analyzing texture continuity and line direction using a curve feature detection algorithm. Dynamic sequence analysis tracks inter-frame differential changes to reveal the degradation propagation trajectory; encoding technology efficiently compresses and stores damage information; and historical comparison analysis uncovers matching degradation patterns and updates potential damage areas. This invention, with 3D spatial modeling as its core, integrates multi-dimensional analysis and efficient storage to achieve precise location, dynamic monitoring, and preventative protection of mural damage, significantly improving the scientific rigor and efficiency of cultural relic protection.

This invention discloses a signal positioning method, device, system, program product, and storage medium, relating to the field of signal testing technology. During the signal positioning process, the server acquires first signal data of the site and images of different areas within the site. The server compares the visual features of the first image among multiple images with a feature library of the site, and determines the first network signal value of a first location point in the three-dimensional spatial model of the site based on the first signal data. A first label is then rendered at the first location point in the three-dimensional spatial model to display the first network signal value. The server does not use satellite signals during signal positioning; instead, it utilizes the visual features of the compared images and a feature library to align the signal value with the location point. This improves the accuracy of signal testing and signal positioning. Furthermore, the server's rendering of the network signal value of the location point in the three-dimensional spatial model provides a visual representation of the site's network signal.

Methods for object- and scene-related storage of image, sensor and / or sound sequences, where objects are found in the image, sensor and / or sound sequences and information about the objects is generated, a scene is formed from these objects, which contains information about the spatial relationships of the objects and / or the recording sensors to each other, where image, sensor and / or sound data are stored in such a way that the image, sensor and / or sound data are checked for a possible full or partial assignment to the objects found, where, in the case of a possible match, only the changed information is stored in addition to this match, Finally, the found objects are not always stored as a whole, but after the initial storage as reference objects, only the reference to the found object along with the deviations of the The found object is transferred from or stored to the reference object, thereby creating a spatial model of the objects, and where 3-dimensional polygon meshes or splines are used to model the surfaces.

Disclosed in the present application are an anti-collision detection method and apparatus for a boom, and a device and a storage medium. The method comprises: acquiring first point cloud data of a boom (101); performing curve fitting on the first point cloud data, so as to obtain a boom curve of the boom (102); on the basis of the boom curve, determining a target curvature for the boom (103); on the basis of the target curvature, determining a target modeling parameter value (104); on the basis of the target modeling parameter value, performing spatial modeling on the first point cloud data, so as to obtain a boom-curvature boom spatial model (105); and on the basis of the boom spatial model, performing anti-collision detection to generate an anti-collision detection result of the boom (106), wherein the anti-collision detection result is used for indicating an obstacle that poses a risk of collision with the boom. In this way, the accuracy of anti-collision detection of booms can be improved.

The invention relates to the technical field of mental health assessment, in particular to a mental electrophysiology assessment method based on a video follow-up visit system, which comprises the following steps of: acquiring doctor-patient double-channel audio and video data through the video follow-up visit system; preprocessing the collected data and extracting multi-modal features including facial behavior features, voice acoustic features and motion dynamics features; and constructing a spatial model of the mental pathological state of the patient based on the extracted multi-modal features, wherein the spatial model describes a dynamic evolution process of a multi-dimensional state variable through a stochastic differential equation. According to the method, a multi-modal feature extraction system of doctor-patient dual-channel audio and video data is constructed, and then a psychiatric pathology state space model based on a stochastic differential equation is established, so that the problems that subjective scale and static observation are mostly adopted in a traditional mental assessment method, and due to lack of objective quantitative indexes and a dynamic tracking mechanism, the accuracy of mental assessment is poor are solved. Therefore, the evaluation result is greatly influenced by subjective experience of doctors, and the dynamic change of symptoms cannot be captured.

This application relates to a method for coupled modeling of the wake field and alternating electric field of a ship propeller, comprising: observational data of measured fluid velocity field and wake electromagnetic field; establishing a spatial model of the fluid velocity field under the assumption of fluid incompressibility and modeling it as a spatial geometric constraint; completing the spatial superposition of the electric dipole core model under the spatial geometric constraint, and modeling the rotational speed of the current loop and the electric field intensity as quantities to be estimated; inverting the quantities to be estimated from the observational data of the wake electromagnetic field using a probabilistic inversion method; obtaining the optimal value of the quantities to be estimated using a machine learning dimensionality reduction method, and using it as the calculation parameter of the core model to complete the numerical modeling of the electromagnetic field; and using a machine learning model to optimize the parameters by taking the observed wake electromagnetic field as input and the equivalent model parameters as output. This application achieves numerical equivalent modeling and efficient inversion inference across physical fields, balancing physical interpretability and computational real-time performance.

The invention discloses a networked heritage space experience data analysis method, and belongs to the technical field of virtual reality application and cultural heritage digital protection crossing. The method comprises the following steps: S1, acquiring a virtual reality scene of a heritage space, and constructing a grid heritage space model; s2, collecting physiological data, behavior data and feedback data of tourists in the gridded heritage space model to obtain multi-modal data; and S3, processing the multi-modal data to obtain an experience data analysis result. The method provides a complete process from data acquisition and processing to result output, supports multi-scene and multi-user parallel analysis, has good engineering applicability and expansibility, and can be widely applied to cultural heritage digital protection and experience design.

This invention discloses a gridded visualization method for scanning lidar detection data, belonging to the field of lidar detection and geographic information system visualization technology. The method comprises the following steps: S1: Constructing a three-dimensional grid spatial model; S2: Collecting raw detection data from the scanning lidar, performing spatiotemporal synchronization and coordinate transformation, and mapping each detection point to the corresponding grid cell of the three-dimensional grid spatial model; S3: For each grid cell, statistically analyzing the historical detection hit count and theoretical scan count, and combining the lidar equation and atmospheric extinction model to calculate at least one detection capability index; S4: Generating visualization rendering data with optimized color mapping, transparency, and hierarchical details based on the detection capability index; S5: Rendering the visualization rendering data using a three-dimensional graphics engine and providing interactive analysis functions to achieve gridded visualization of detection capabilities. This invention improves the accuracy, intuitiveness, and interactive efficiency of lidar detection performance evaluation.

The application discloses a freezing design method for a strong water-permeable pebble layer through a freezing process of an inclined shaft, and particularly relates to the technical field of freezing construction. The rock and soil parameters and the regional historical environment temperature of the pebble layer through which the target inclined shaft passes are obtained; a three-dimensional geometric model of the inclined shaft is established to determine the position, inclination and spacing of the freezing pipes arranged along the shaft axis; a spatial model of the freezing area coupled with the pebble layer is constructed; a freezing-heat-seepage multi-field coupling model is constructed according to the historical environment temperature and the freezing pipe arrangement information, the formation process of the freezing body is numerically simulated, and the evolution functions of the freezing boundary expansion range and the freezing thickness at different times are obtained; the freezing coverage is analyzed to determine whether each depth section meets the design freezing thickness and safety factor requirements. The method can realize multi-physical field coupling modeling of the freezing process and accurate control of the freezing body evolution, has the advantages of strong predictability, uniform freezing effect and the like, and is particularly suitable for the freezing design and construction of the inclined shaft under complex geological conditions.

The invention discloses a terrain-adaptive rainstorm waterlogging spatio-temporal joint prediction method and system, and belongs to the field of alarm responding to disaster events. In order to overcome the defects of insufficient terrain adaptability of fixed feature weight and prediction splitting caused by independent operation of a space-time model in the existing method, the technical scheme of the invention is as follows: on one hand, a terrain adaptive dynamic feature weight mechanism is constructed, the topographic relief is quantified to generate an adjustment coefficient, and the feature factor weight of a geographic space-time weighted regression model is dynamically adjusted; a complex lower cushion surface is adapted; on the other hand, a space-time joint weighted fusion framework is provided, a random forest space model and a PredFormer time sequence model are combined, a space prediction result is used as a time sequence initial condition, and submerging range and water depth time sequence collaborative prediction is achieved. The method is suitable for rainstorm waterlogging prediction in different terrains of cities, and has the advantages of high terrain adaptability, high prediction precision and stability, uniform space and time and quick response.