38 results about "Observation parameter" patented technology

The invention provides a carbon emission calculation method based on adaptive Kalman filtering. According to the carbon emission calculation method based on adaptive Kalman filtering, a data matrix is generated by acquiring multi-source data, and a unified data basis is provided for subsequent calculation; a dynamic system model and an error measurement matrix are initialized, a state transition model indicates a carbon emission evolution law, an observation model establishes a mapping relation between multi-source data and carbon emission, and the error measurement matrix dynamically indicates a model effect and adaptively adjusts calculation parameters to ensure the adaptability of the model to the dynamic change of the system; a carbon emission prediction parameter and an error measurement matrix are calculated in real time through adaptive Kalman filtering, a prediction value is corrected through an observation parameter, an optimal estimation parameter and an observation residual error are obtained, meanwhile, the error measurement matrix is updated in real time, and dynamic fusion and noise suppression of multi-source data are achieved through the closed-loop process. And the precision and the real-time performance of carbon emission calculation are improved.

The invention relates to the technical field of bamboo forest ecological environment monitoring, in particular to a bamboo forest ecological environment collaborative observation method based on multi-source Internet of Things sensing, and the method comprises the steps: synchronously collecting a multi-source collaborative observation parameter set of a bamboo forest slope through an Internet of Things sensor network, and then constructing a point-surface scale fusion bamboo forest moisture dynamic model; correcting evapotranspiration point cluster data by using a random forest gradient lifting algorithm to generate terrain weighted evapotranspiration, aggregating runoff data according to terrain weight to obtain a total amount of runoff sub-regions, then executing dynamic water balance constraint calculation, optimizing infiltration capacity in combination with a variation assimilation algorithm, redistributing infiltration capacity space proportion according to soil hydraulic conductivity when a residual error exceeds a threshold value, and calculating the total amount of the runoff sub-regions. Finally, according to the residual space distribution thermodynamic diagram, the evapotranspiration point cluster is migrated to the micro-topography representative area, and observation network adaptive optimization is achieved, the problems that in the prior art, the data scale is disjointed, the observation precision is insufficient, and the network is difficult to adapt to change are solved, and the bamboo forest ecological observation precision and reliability are improved.

The invention relates to a shadow generation method and device for a three-dimensional model, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring light source information of a virtual light source and observation parameters of a virtual camera; according to the light source information and the observation parameters, a multi-view profile diagram corresponding to the three-dimensional model is generated, and the multi-view profile diagram is used for representing an initial shadow; projecting the multi-view profile diagram to a reference plane according to the light source information and the observation parameters to obtain a hard shadow image corresponding to the three-dimensional model; obtaining shadow features of the three-dimensional model according to the light source information and material attributes of the three-dimensional model; and inputting the hard shadow image and the shadow features into a trained shadow prediction model based on a convolutional neural network, and outputting a soft shadow image corresponding to the three-dimensional model through the shadow prediction model according to the hard shadow image and the shadow features. By adopting the method, the shadow of the three-dimensional model can be quickly generated.

The invention provides a key test factor determination method and device based on a Bayesian causal network, and relates to the technical field of reason tracing. The method comprises the following steps: acquiring a multi-dimensional observation parameter of a preset intelligent system, and generating an initial data set based on the multi-dimensional observation parameter; based on an improved Bayesian causal structure learning algorithm and the initial data set, determining a causal structure diagram between an input factor and a target output in the initial data set; based on the causal structure diagram, performing intervention operation on an input factor pointing to the target output to determine an average causal influence degree of the input factor on the target output; and calculating importance degree scores of the input factors according to a preset weight fusion mode and the average causal influence degree, and determining the input factors corresponding to the importance degree scores meeting a preset key factor selection condition as key test factors. In the intelligent system, the real causal-driven effect between the variables is effectively determined, and the accuracy of the key test factor is improved.

A method includes obtaining satellite information, including a plurality of navigation satellite identifiers, and satellite data, observed values of an observation parameter, and weight values that are separately associated with the plurality of navigation satellite identifiers; determining that an inaccurate observed value exists in a set of the observed values; determining, based on the satellite information, a check vector representing accuracy of the set, and a plurality of eigenvectors representing accuracy of an observed value of the observation parameter that is associated with a corresponding navigation satellite identifier; performing cluster analysis on the check vector and the plurality of eigenvectors, and determining a target navigation satellite identifier corresponding to an eigenvector that is of a same type as the check vector; and performing selection in the plurality of navigation satellite identifiers and / or adjusting a weight value associated with the target navigation satellite identifier based on the target navigation satellite identifier.

This invention discloses a crop-adaptive agricultural meteorological observation system, belonging to the field of agricultural meteorological observation technology. It integrates multiple devices to collaboratively measure various meteorological variables to indicate weather conditions, deploys them in a grid-like layout, and collects and transmits data regularly. It collects key growth parameters according to crop growth stages and synchronously correlates them with meteorological data; optimizes data and accurately matches it to construct a fusion dataset; establishes a correlation model to identify key meteorological factors; adjusts observation parameters according to crop needs; sets two-level early warnings and pushes information through multiple channels; retrieves data from multiple dimensions and securely stores it; and provides visual operation and parameter setting functions. This invention improves the accuracy of meteorological data through multi-variable collaborative observation, dynamically adapts to the needs of different crop growth stages and varieties, provides comprehensive regional collaborative observation coverage, timely and targeted early warnings, supports linkage with external devices, promotes the refinement and intelligence of agricultural meteorological observation and planting management, and helps ensure crop growth.

This invention discloses a method for inverting sea surface wind speed from GNSS reflected signals based on mask perception and image-based Mamba-CNN. The method includes: performing spatiotemporal matching and interpolation processing on multi-source spaceborne GNSS-R observation data and ERA5 reanalysis data; gridding the processed observation parameters according to geographical location to obtain a multi-channel grid image and acquiring a binarized occupancy mask; inputting the multi-channel grid image and the binarized occupancy mask into a mask perception Mamba-CNN model; calibrating the grid features through a mask conditional encoder and a feature linear modulation layer, and performing spatial modeling using feature extraction blocks combining convolutional and Mamba state space paths to obtain the wind speed inversion result; aggregating the extracted features using center-perception pooling, and outputting the final sea surface wind speed inversion image through a regression network. This invention can significantly reduce computational complexity, effectively reconstruct the continuous wind field structure, and improve the inversion accuracy and robustness in high-wind-speed areas.

The invention discloses an elevator operation and maintenance retrieval enhanced question and answer method, device and equipment and a storage medium. According to the method, by introducing the parameterized hyperedge, the limitation of a binary relationship of a common knowledge graph is solved, the method can adapt to a multi-entity and multi-constraint elevator operation and maintenance retrieval enhanced question-answer scene, and the accuracy of elevator operation and maintenance retrieval enhanced question-answer and elevator maintenance work orders is improved. Besides, the comprehensive score of each parameterized hyperedge in the hyperedge candidate set is obtained by integrating the semantic similarity score and the consistency score, and K target parameterized hyperedges of which the comprehensive scores are in the front are determined, so that the semantic similarity between the elevator maintenance consultation text and text items of the target parameterized hyperedges is ensured, and the comprehensive score of each parameterized hyperedge in the hyperedge candidate set is determined. And the working condition represented by the constraint set of the field observation parameters and the target parameterized hyperedge is matched, so that the accuracy of the generated elevator maintenance work order is improved.

The invention provides an accelerator parameter selection method and device, equipment, a storage medium and a program product, and relates to the field of hardware design, core design parameters of target accelerator design and preset weights thereof can be determined according to priori knowledge, and in a design space of the target accelerator design, the number of candidate values of the core design parameters is used as the number of clusters, and the cluster number is calculated. According to the cluster number and the distance between the parameter points in the design space, parameter point clustering is carried out, parameter point clustering is carried out under the guidance of priori knowledge, and a parameter point cluster under each type of core design parameter configuration is obtained. Then, high-value candidate parameter points can be screened from each parameter point cluster. Then, observation parameter points can be sampled from the candidate parameter points, accelerator design simulation is performed on the observation parameter points to obtain real index values, and Bayesian optimization processing is performed on the candidate parameter points in an iteration manner according to the observation parameter points and the real index values thereof to obtain target design points of target accelerator design, so that the parameter point selection effect can be improved.

The invention relates to a small watershed multi-element cooperative monitoring method, device and equipment and a medium. The method comprises the following steps: acquiring satellite remote sensing data, unmanned aerial vehicle observation data and ground observation data of a target drainage basin; preprocessing the satellite remote sensing data, the unmanned aerial vehicle observation data and the ground observation data to obtain multi-element observation parameters; carrying out basin live inversion according to the multi-element observation parameters to obtain land utilization grids, water body boundary-area data, runoff data and sediment data; and according to the land utilization grid, the water body boundary-area data, the runoff data and the sediment data, constructing a rainfall-runoff-sediment-water quality chain response relationship, and obtaining drainage basin multi-element scale coordination time sequence data, an element influence map and land utilization influence evaluation. By adopting the method, multi-element data space-time cooperative monitoring can be realized, and the full-chain demand of small-watershed ecological hydrological management is supported.

The application discloses a method and device for inverting typhoon wind speed by using synthetic aperture radar, equipment, medium and product, and relates to the field of ocean remote sensing and application. The method comprises the following steps: obtaining a group of observation data of the synthetic aperture radar when the synthetic aperture radar inverts the typhoon wind speed; obtaining an inversion value of the typhoon wind speed based on the weight of each observation parameter and the group of observation data of the synthetic aperture radar when the synthetic aperture radar inverts the typhoon wind speed; the determination process of the weight is as follows: constructing a target function for inverting the typhoon wind speed by using the synthetic aperture radar based on elastic network regularization; obtaining the weight of each observation parameter by solving the target function for inverting the typhoon wind speed by using the synthetic aperture radar based on elastic network regularization according to a training data set; and the training data set comprises a plurality of groups of observation data of the synthetic aperture radar and corresponding typhoon wind speeds of each group of observation data when the synthetic aperture radar monitors sample typhoon wind speed. The application can realize high-precision inversion of the typhoon wind speed.

The invention discloses a GNSS-R soil humidity inversion method and device, a storage medium and computer equipment. Comprising the following steps: inputting a delay-Doppler graph obtained by detecting a target area by a satellite-borne GNSS-R into a feature extraction model corresponding to the target area, extracting a global dependency relationship of a feature sequence of the delay-Doppler graph through the feature extraction model, and outputting global scattering feature data of the target area based on the global dependency relationship; inputting the ground environment data of the target area and the observation parameters of the satellite-borne GNSS-R into a geographically weighted regression model corresponding to the target area, performing fitting calculation on the spatial correlation of the ground environment data and the observation parameters through the geographically weighted regression model, and outputting prior inversion data of the soil humidity of the target area; and performing nonlinear transformation and feature fusion on the global scattering feature data, the prior inversion data and the observation parameters based on a soil humidity model to obtain a soil humidity inversion result. Therefore, the adaptability and inversion accuracy of the model in different geographical environments are improved.

The invention relates to the technical field of robot algorithm control, in particular to a robot control method and system based on a blind privilege information implicit alignment algorithm, equipment and a medium, and the method mainly comprises the steps that real-time observation parameters and historical observation parameters of real-time operation of a robot are obtained; obtaining privileged observation parameters, wherein the privileged observation parameters comprise a second parameter obtained when the robot runs; and inputting the real-time observation parameters, the historical observation parameters and the privileged observation parameters into an ActorCritic model to obtain a joint action actcion and a corresponding action evaluation value. According to the method, vae is not adopted, the training process is more stable, and the convergence speed is higher; the topographic information estimated by the polyae is more accurate and explainable; end-to-end training does not exist, no complex two-stage training process exists, distillation is carried out while learning is carried out, the phenomenon that only the experience of'successful 'distillation is carried out in a teacher-student framework can be avoided, and the policy is more robust and stable.

The invention relates to a parameter adjustment method for large-scale mooring buoy seaborne comparative measurement, which comprises the following steps: establishing connection with a data acquisition system of a target buoy in a wireless communication mode, acquiring an observation element type and a sampling frequency parameter of the target buoy, and adjusting the observation element type and the sampling frequency parameter of the target buoy based on the acquired parameter; performing observation parameter configuration on each sensor arranged in the comparison measurement device, synchronizing the comparison measurement device with a system clock of the target buoy, monitoring the drifting direction and speed of a water sail at the bottom of the comparison measurement device, and determining the drifting speed of the target buoy according to the collected wave observation data and the height-corrected wind speed and wind direction data. And calculating the flow velocity and the flow direction of the surface ocean current through a preset algorithm model.

A method for authenticating map elements of safety-critical driving functions via a control device is disclosed. After a mapping step is performed, at least one observation parameter for at least one mapping step of at least one map element is obtained via at least one monitoring function, and the observation parameter is compared with an expected value. Based on the comparison of the observation parameter with the expected value for at least one mapping step, at least one result value is calculated via the monitoring function, and the at least one result value is stored as a certificate and associated with the at least one map element. Furthermore, a control device, a computer program, and a computer-readable storage medium are disclosed.

The present application relates to the technical field of safety monitoring, in particular to a hoist safety performance evaluation method, the method comprising: taking hoist motor current and rope tension as observation parameters, recording data under differentiated working states, identifying dynamic characteristics of the hoist through analyzing the data, and generating dynamic parameter record results. In the present application, dynamic parameter recording allows real-time capture of subtle changes of the hoist under differentiated working states, providing a basis for accurately identifying dynamic characteristics of the equipment, and the application of performance fluctuation indicators further utilizes dynamic recording, identifies fluctuations and deviations outside the normal range through analyzing data changes, provides a basis for judging the severity of fluctuations and taking corresponding adjustment measures, improves the monitoring capability and fault prevention efficiency of the hoist operating state, can identify and take measures in time before potential risks develop into serious faults, reduces the probability of accidents, and improves the safety and efficiency of operations.

The invention relates to a normalized region automatic monitoring and updating method and system. The automatic monitoring and updating method for the normalized area comprises the following steps that S101, observation parameters, configured by a user, of a normalized observation area are received and stored, and the observation parameters comprise the area range, the shooting period and the filing address; s102, on the basis of the shooting period, an observation task is automatically generated at regular time; step S103, automatically issuing the generated observation task to a satellite operation unit to execute shooting; according to the normalized region automatic monitoring and updating method and system, automatic operation of the system can be driven through a small amount of configuration, a traditional manual operation mode is greatly replaced, task dynamic generation and efficient circulation from a reporting party to a satellite party are achieved, fundamental transformation of a remote sensing service mode from passive response to active pushing is promoted, and the remote sensing effect is improved. And the overall remote sensing data service capability and the intelligent level are improved.

The invention discloses a method and a system for correcting hygroscopic parameters of non-ideal mixed aerosol and a medium, and belongs to the technical field of atmospheric measurement. The method comprises the following steps: acquiring instrument observation data; estimating instrument observation data through a ZSR ideal mixing criterion, and generating ZSR hygroscopic observation parameters; obtaining an input feature data set according to the ZSR hygroscopic observation parameters, the instrument observation data and inversion hygroscopic observation parameters generated by inversion of the instrument observation data; inputting the input feature data set into a pre-constructed deviation prediction model, and outputting a non-ideal mixed deviation item; calculating a main effect value of the input feature data set, and generating empirical correction terms based on key features according to the main effect value; and obtaining a moisture absorption parameter correction result according to the ZSR moisture absorption observation parameter, a pre-constructed seasonal correction coefficient and an empirical correction item based on the key characteristics. According to the method, the deviation part can be explained, and the contribution degree of each influence factor is quantitatively explained.

The application discloses a kind of non-ideal mixed aerosol hygroscopicity parameter correction method, system and medium, belong to atmospheric measurement technical field.Its method includes: obtaining instrument observation data;Estimate instrument observation data by ZSR ideal mixing criterion, generate ZSR hygroscopicity observation parameter;According to ZSR hygroscopicity observation parameter, instrument observation data and inversion generated by instrument observation data inversion hygroscopicity observation parameter, obtain input feature data set;Input feature data set is input into the bias prediction model pre-constructed, and output non-ideal mixing bias term;Calculate the main effect value of input feature data set, and generate the empirical correction term based on key features according to main effect value;According to ZSR hygroscopicity observation parameter, pre-constructed seasonal correction coefficient, the empirical correction term based on key features, obtain hygroscopicity parameter correction result.The application can explain the deviation part, and quantitatively explain the contribution degree of each influence factor.

The invention relates to the technical field of carbon emission, and further relates to a carbon emission accurate measuring and calculating system based on multi-scale variational Bayesian reasoning, and the system comprises a data collection unit which is used for collecting basic observation parameters at an inlet of each carbon emission monitoring unit, and forming an original basic data sequence which is arranged according to a time sequence; the time alignment and channel classification unit is used for aligning the original basic data sequence into basic time slices according to the minimum adjacent acquisition interval; multi-scale observation structures of a short window, a middle window and a long window are respectively organized for each physical channel; the Bayesian reasoning unit is used for carrying out proportion alignment on the fuel channel and the output channel by taking the exhaust channel as a reference channel; and when a preset convergence condition is satisfied, outputting an emission estimation sequence organized according to the monitoring unit and the basic time slice and the accumulated emission of each window, otherwise, continuing iteration. According to the method, the precision, robustness and applicability of measurement and calculation can be remarkably improved.

The embodiment of the invention provides a scooter rollover processing method, a scooter and a computer readable storage medium, detection equipment is arranged on the scooter, and the detection equipment is used for performing target detection on the detection range of the detection equipment according to a preset period; the method comprises the steps that in the operation process of the scooter, a target detection result of each detection frame in a detection frame sequence of detection equipment is acquired, and the target detection result of each detection frame is used for indicating a parameter value of an observation parameter of each detection target in each detection frame; the following operation control operation is executed based on iteration of the sliding window, and the multiple detection frames in the sliding window in the iteration are multiple intra-window frames: rollover detection is performed on the scooter based on parameter values of observation parameters of at least part of detection targets in each intra-window frame in the multiple intra-window frames; and when it is detected that the scooter is in the rollover state, a motor of the scooter is controlled to stop power output.

The application discloses a positioning method, device and storage medium of satellite navigation. The positioning method comprises the following steps: acquiring observation parameters collected by a mobile terminal at a preset epoch, wherein the observation parameters comprise GNSS parameters and inertial navigation related parameters, the GNSS parameters comprise GNSS prior parameters and GNSS posterior parameters; inputting the observation parameters into a preset positioning accuracy evaluation model to obtain an accuracy evaluation value, wherein the positioning accuracy evaluation model is obtained based on historical observation parameters collected at multiple epochs; determining whether the positioning information represented by the current observation parameters is accurate based on the accuracy evaluation value; when it is determined that the positioning information is inaccurate, determining a positioning result based on a float solution of the integer ambiguity; and when it is determined that the positioning information is accurate, determining a positioning result based on a fixed solution of the integer ambiguity. The application has the technical effect of improving the positioning accuracy and reliability of the global satellite navigation system.

The present disclosure relates to a data collection method, device, medium and product, the method comprising: obtaining first observation parameters of each type of monitoring data of a monitoring device for a preset time monitoring window; wherein the first observation parameters comprise: a first window observation value of each type of monitoring data and a first observation fluctuation threshold of the first window observation value; determining first prediction parameters of each type of monitoring data for the first time monitoring window; the first prediction parameters at least comprise: a prediction value, an upper limit of the prediction value, and a lower limit of the prediction value; determining whether the first time monitoring window satisfies a collection condition based on the first observation parameters and the first prediction parameters; in the case of determining that the collection condition is satisfied, collecting data of the monitoring device to obtain a target observation value. The present disclosure can avoid a large amount of redundant data in the data collection process, reduce the storage and calculation cost of hardware for data, and improve the efficiency of data collection.

The invention discloses an ocean heat wave forecasting method and device, a medium and a product, and relates to the technical field of ocean heat wave intelligent forecasting, and the ocean heat wave forecasting method comprises the steps: obtaining a measurement value of a multi-source observation parameter of a target sea area at a current moment; standardizing the measured value of the multi-source observation parameter at the current moment to obtain a standardized value of the multi-source observation parameter at the current moment; inputting the standardized values of the multi-source observation parameters at the current moment into the ocean heat wave forecasting model, and outputting predicted values of event parameters of ocean heat waves of the target sea area at N future moments; the event parameters comprise occurrence conditions and intensity; the occurrence condition is that ocean heat waves occur or do not occur, and the strength is zero strength, medium strength, high strength or serious strength; the ocean heat wave forecasting model is obtained by training a random forest model, Ngt; 1. According to the invention, the forecasting precision of ocean heat waves is improved.

The application discloses a kind of zinc rotary kiln working condition intelligent monitoring method, device, terminal and medium, comprising: obtaining the flame video data of kiln head crater collected by industrial camera;According to the preset time length, the key combustion characteristics of material in flame video data are extracted;According to the preset time length, the corresponding key monitoring data are read from equipment monitoring system;The key combustion characteristics and the key monitoring data are combined to form observation parameter matrix according to different arrangement order;Based on convolutional neural network, the fusion features of the observation parameter matrix are calculated, and the kiln condition information is output according to the fusion features.The application obtains the running state of volatile kiln by computer vision and artificial intelligence technology, solves the problem of large time lag and randomness in traditional manual observation of flame and monitoring data based on experience to judge working condition.

The present invention provides a method having an improved pneumatically actuated system that enables reliable movement of one or more positioned pin lifting devices and / or vacuum valves. [Solution] A method for providing control parameters for controlling a vacuum system 1, wherein a control unit 30 derives control parameters based on processing a set of observation parameters, the control unit 30 applying a first control signal to a first control valve 21 to change the pressure of a pneumatic actuator 11, thereby moving a first mount 17 from a first start position to a first end position, monitoring the fluctuating pressure of the first pneumatic actuator with a first fluid sensor 22, determining when the first mount reaches the first end position, and processing a set of observation parameters. The set of observation parameters includes at least the first control signal, the fluctuating pressure, and information regarding the arrival of the first mount at the first end position.

The embodiment of the invention discloses a fault diagnosis method, a program product and equipment. The method comprises the following steps: acquiring current related variable data of a related variable associated with a monitoring target and current observation parameter data indicating an observation result of the monitoring target; input data of a predetermined fault diagnosis model is formed based on the current related variable data, the confidence interval range of the observation parameters is determined through the fault diagnosis model, and the fault diagnosis model is obtained through calculation based on a sample data set corresponding to the monitoring target; each piece of sample data in the sample data set comprises sample related variable data of related variables and sample observation parameter data corresponding to the sample related variable data, and one piece of sample data is data obtained from steady-state data obtained when one piece of equipment is in a stable operation state; the sample data set is obtained based on a plurality of different devices; and based on the current observation parameter data and the confidence interval range of the observation parameters, evaluating the current state of the monitoring target.

The invention discloses an ortho-rectification method and system for a high-resolution SAR (Synthetic Aperture Radar) image in a fluctuating area. The method comprises the following steps: acquiring observation parameters of a high-resolution DEM (Digital Elevation Model); generating a simulated SAR image by using the observation parameters, matching the simulated SAR image with a real SAR image, correcting an initial positioning deviation, and extracting a control point of a fluctuating region by performing secondary matching on the corrected real SAR image; acquiring a high-density control point of a flat area in the optical reference image and the initially calibrated real SAR image by using a geometric distortion perception multi-mode matching method; and performing statistical analysis based on the control points of the flat area and the control points of the fluctuating area, constructing a geometric fine calibration model, and correcting positioning errors in the azimuth direction and the distance direction for the corrected SAR image to realize high-precision orthorectification. According to the SAR ortho-rectification method combining DEM simulation and multi-mode matching, the problems of fluctuation distortion, DEM low resolution and the like are solved under the condition of no GCP, and high-precision ortho-rectification is achieved.

The invention belongs to the technical field of power station inspection, and provides a power station inspection preset bit configuration method and system based on a three-dimensional Gaussian model. The power station patrol preset position configuration method based on the three-dimensional Gaussian model comprises the following steps: generating the three-dimensional Gaussian model and pre-labeling patrol path information in the three-dimensional Gaussian model; positioning a 3D bounding box of the target equipment and extracting a multi-view image covering each key part of the target equipment in the 3D bounding box; equipment point locations are obtained and bound to a standard equipment model point location ledger; calculating an optimal patrol terminal for observing the equipment point location, and carrying out association binding on the equipment point location, the optimal patrol terminal and the standard equipment model point location ledger; and calculating observation parameters of the optimal patrol terminal corresponding to each equipment point location according to the spatial position relationship between each equipment point location and the patrol terminal associated with the equipment point location, and forming a power station patrol preset position configuration scheme. The intelligent configuration of the inspection equipment can be realized, and the configuration efficiency and accuracy are remarkably improved.

The application discloses a method and system for intelligent reconstruction and prediction of a marine three-dimensional flow field, and relates to the technical field of data processing. The method comprises the following steps: obtaining multi-source observation data of the marine through a large language model and an edge monitoring device; preprocessing and fusing the multi-source observation data to obtain fused observation data; constructing a spatiotemporal correlation graph by taking a target observation parameter in the fused observation data as a node and taking the spatiotemporal correlation and the physical quantity coupling relationship between the target observation parameters as edges; determining reconstruction data corresponding to the multi-source observation data according to the spatiotemporal correlation graph; and predicting a three-dimensional flow field of the marine according to the reconstruction data and outputting a visualized three-dimensional flow field. The application can fuse and reconstruct the multi-source observation data, thereby improving the accuracy of the prediction of the three-dimensional flow field of the marine.