125 results about "Cloud atlas" patented technology

The invention provides a cloud atlas segmentation method based on a FCN and a CNN, and belongs to the field of image segmentation of computer vision. The method is characterized in that firstly, a near neighborhood of each pixel point in a cloud atlas is correspondingly clustered through an ultra-pixel, the cloud atlas is input to fully convolutional networks FCN32s and FCN8s with different step lengths, and pre-segmentation results of the cloud atlas are realized; in a FCN32s result map, a black area is a part of a non-cloud area in the cloud atlas, and in a FCN8s result map, a white area is a part of a cloud area in the cloud atlas; and remaining uncertain areas are grey areas which need to be determined through a deep convolutional neural network (CNN), key pixels in an ultra-pixel area need to be selected to represent characterizes of the ultra-pixel area, and whether the characteristics of the pixels are cloud or non-cloud is determined through the CNN network. According to the method, the precision is equivalent to that of MR-CNN and SP-CNN, and the speed is increased by 880 times compared with MR-CNN and increased by 1.657 times compared with SP-CNN.

The invention discloses a power-grid-GIS-based lightning analysis and early warning method and a system thereof. The method comprises the following steps that: a basic power grid equipment geological information graph layer is established according to the coordinate of the power grid equipment; real-time monitoring data of lightning weather of the power grid equipment coverage area as well as meteorological cloud atlas data are obtained; a lightning weather data model is established; and the real-time monitoring data and the lightning weather data model are matched, a lightning early warning grade is determined, the basic power grid equipment geological information graph layer and the lightning weather chart layer are associated, and the power grid equipment information and lightning early warning data are overlapped and displayed on the power grid GIS platform. The method and system have the following beneficial effects: the lightning falling area is analyzed and a professional power lightning disaster early warning model is established; power equipment is associated; and precise early warning for the single line and equipment is realized. Targeted measures and specifications for geological disaster prevention can be formulated; and the losses of the power grid can be reduced.

The invention provides a method for typhoon monitoring and evaluation of the monitoring precision based on multi-source satellite data, belongs to the technical field, and solves the problem of limitation of a single satellite data source in the practical application in the prior art. The method for typhoon monitoring and evaluation of the monitoring precision based on the multi-source satellite data includes: (1) typhoon information data acquisition: obtaining typhoon visible light or infrared image information data by employing a Fengyun geostationary meteorological satellite, and obtaining typhoon microwave scattering data by employing a microwave scatterometer of an ocean satellite; (2) information data standardization processing: performing cloud atlas grayscale information extraction on a read base data source of visible light or infrared images of a Fengyun geostationary meteorological satellite partition map, performing geo-statistical interpolation on the microwave scattering data of the ocean satellite to obtain a final regional wind field map; and (3) typhoon center interpretation: determining the typhoon center. According to the method, accurate typhoon information is obtained by employing mutual confirmation and mutual supplementation of different satellite data.

The invention discloses a coal and gas outburst forecasting method which comprises the following steps that: (1) a geometric model is built; (2) material parameters, initial values and boundary conditions of an area to be tested are given; (3) a theoretical model is built; (4) the material parameters, the initial values and the boundary conditions are substituted into the theoretical model to calculate the coal and gas outbursts in the area to be tested; (5) the coal and gas outbursts calculated in step (4) is marked to generate a cloud atlas; and (6) according to the number of the coal and gas outburst marks of the area to be tested in the cloud atlas, whether the area has coal and gas outbursts is judged. The coal and gas outburst forecasting method of the invention first builds the theoretical model for judging the outbursts, and then adopts a finite element technique to calculate, to realize the forecasting of the coal and gas outbursts. The forecasting method of the invention is based on the complete theoretical model, realizes the coal and gas outburst forecasting by numerical calculation, is easy to operate in actual situations, and can greatly reduce the engineering cost.

The invention relates to a terrain geometric parameter extraction method based on a hierarchical elevation cloud atlas, which comprises terrain classification and recognition, terrain simplification and terrain geometric parameter extraction processes. For a 401*401 DEM grid terrain specified in the digital map of the GeoTIFF format of the ASTER GDEM type, the relative height is first calculated to divide the terrain into five types: the plain, the hill, the low mountain, the middle mountain, and the high mountain through programming; the hierarchical elevation cloud atlas of the specified terrain is obtained through elevation layering and color mapping; the eight neighborhood boundary tracking algorithm is used in image processing to obtain the terrain bottom surface boundary, the bottomsurface area and the center of the bottom surface, and the color representing the maximum elevation interval is identified to obtain the average elevation of the highest area of the terrain; finally,the geometric parameters of the plain, spherical crown, cone and wedge shape of the terrain simplified geometric model are obtained by combining the terrain classification, the calculation of the bottom shape factor and the slope calculation. The invention has the advantages of simple method, small calculation amount, real-time performance and practicality.

The invention provides a typhoon positioning detection method based on a Himawari meteorological satellite. A meteorological satellite cloud atlas in 0.64-micron short-wave infrared window frequency band is adopted, the detection precision is higher, and the instantaneity is good; since the form of the typhoon is highly irregular at the beginning and the end of the typhoon, a detected center location and a CMA optimal path are large in deviation; and when the typhoon forms obvious eye of wind or is at a mature stage, the detection method has high accuracy on the centralized positioning of thetyphoon.

The invention discloses a finite element simulation analysis method of a power wire clip. The finite element simulation analysis method comprises the following steps of S1, establishing a power wire clip model by three-dimensional software; S2, guiding the power wire clip model obtained in S1 into finite element simulation analysis software; S3, establishing a coordinate system, dividing meshes of the power wire clip model, and forming tetrahedral meshes; S4, applying the boundary conditions of the power wire clip model to simulate and calculate; S5, calculating to obtain results; S6, processing the simulation and calculation results into a simulation cloud atlas or an applicable data format, observing the whole and local states of stress distribution via the simulation cloud atlas, and determining the stress number of each part of the power wire clip on the simulation cloud atlas. The finite element simulation analysis method has the advantages that the wire clip can be quickly modeled, the stress of the wire clip is performed with simulation analysis, the stress centralizing part is obtained, and the key optimizing and defect monitoring on the stress centralizing part can be performed.

The present invention discloses a cloud cluster automatic detection method based on a foundation cloud atlas. The cloud cluster automatic detection method comprises: an image feature space conversion step for normalizing a ratio of a blue channel value and a red channel value of each pixel point in the foundation cloud atlas to obtain an NBR value serving as a color feature value of the pixel point; extracting the pixel points (called uncertain pixel points) of which the NBR values are within the interval of [0, 0.3], and performing minimum cross-entropy calculation to obtain a segmentation threshold with the minimum cross entropy; and comparing the NBR value of each pixel point with the segmentation threshold, if the NBR value is less than the segmentation threshold, judging that the pixel point is cloud cluster, and otherwise, judging that the pixel point is blue sky, thereby achieving detection of the cloud cluster. By only calculating the minimum cross entropy of the uncertain pixel points to obtain the optimum segmentation threshold, the cloud cluster automatic detection method of the present invention can well solve the problem of segmentation threshold offset caused by rendering of a background for the color of the cloud cluster in the case of an extremely blue sky or an extremely bright sky, thereby greatly improving detection precision of the cloud cluster in the foundation cloud atlas under a complicated background.

The invention discloses a probability distribution based short-time rainfall forecast method and system, and belongs to the technical field of rainfall forecast. The system comprises a cloud atlas collection module, a meteorology data acquisition module, a data transmission module, a master control module, a rainfall probability generation module, a precipitation estimation module, a rainfall characteristic determination module and a data display module. The precipitation estimation module can obtain rainfall data of higher precision and confidence by using advantages of multi-source data; therainfall characteristic determination module uses a rainfall rain type characteristic matrix and time characteristic matrix of a rainfall station to describe the rainfall and movement condition of the rainstorm center in the aspects of time and space; and a mathematic matrix is used to describe the time-space distribution characteristic of rainfall, the rainfall characteristic is researched lateron the basis of the mathematic matrix, and the accuracy and comprehensiveness of rainfall characteristic research are improved.

The invention provides a creep fatigue life prediction method based on a crystal plasticity theory. The method comprises the following steps: establishing a representative unit model of ABAQUS based on an electron back scattering diffraction technology; correcting the back stress model and writing the back stress model into a subprogram UMAT to obtain a creep fatigue hysteresis loop; fitting a creep fatigue hysteresis loop through a test parameter method to obtain material parameters; calculating a stress-strain value of each integral point and averaging the stress-strain values to obtain a creep fatigue hysteresis loop and a post-processing cloud atlas; extracting maximum plastic slippage and energy dissipation from the creep fatigue hysteresis loop and the post-processing cloud atlas, analyzing the change rule of the maximum plastic slippage and energy dissipation along with circulation cycles, and providing creep and fatigue indication factors; and predicting the creep fatigue crackinitiation life according to the indication factor. According to the creep fatigue life prediction method, plastic slippage and energy dissipation are used as fatigue and creep indication factors, the creep fatigue damage evolution rule can be better reflected, the crack initiation position can be accurately predicted, and the method has the advantages of being visual, high in applicability and high in accuracy.

The invention relates to a method for calculating the strength of a main shaft of a wind turbine generator set, modeling is implemented for a platform through existing finite element software, material attributes of all parts are finally defined, calculation is performed, and a deformation and stress cloud atlas is obtained; the modeling process mainly comprises the steps of using a solid unit to set and merge the main shaft, a bearing and a locking disk into an integrated structure in a simulating manner, arranging four gear box elastic supports outside the locking disk, using a spring unit group to simulate each gear box elastic support, connecting nodes of a wind wheel with all flange bolt holes of the main shaft through a rigid beam unit I, and connecting the locking disc with the spring unit group through a rigid beam unit II; and applying external loads on the nodes of the wind wheel, applying all constraints on the spring unit group and applying position constraints on the bearing. The method has the benefits of being conductive to precisely calculating the strength performance of the main shaft, being capable of simultaneously calculating multiple working conditions, being conductive to saving time and saving cost, and being very applicable to large-scale popularization.

The invention discloses a system and a method for stimulating and predicting a melt thermocapillary convection process and belongs to the technical field of semi-conductor materials. The system comprises a melt data acquisition module, a thermocapillary convection simulation and prediction module, a simulation and prediction result storage module and a result image display module, wherein the thermocapillary convection simulation and prediction module comprises a grid partition sub-module, a thermocapillary convection calculation sub-module and an interface recognition module. The method comprises the steps of acquiring geometric parameters of a melt liquid bridge area, medium physical parameters, physical parameters of an fluid on the side of an environment where the melt is located and initial parameters of a melt thermocapillary convection process; performing thermocapillary convection simulation and prediction; and classifying a velocity field, a temperature field, a flow field, a pressure field and two phase boundary positions, storing results in chronological order, and displaying the results in a form of a graph, a vector diagram or a cloud atlas in a classified manner. The system and the method can be used for predicting the development of internal thermocapillary convection in a crystal growth process as well as the distribution of the temperature field and the velocity field in a microgravity environment and under magnetic field control, and a prediction means is provided for preparing high-quality single crystals.

The present invention relates to a short-term load forecasting method based on a cloud model. First, a three-layer classification model is established based on seasons, day types and meteorological factors, and the third-level index is extracted through the correlation coefficient method, that is, the characteristic quantity of meteorological factors affecting the load size. According to the different mechanisms of the influence of characteristic quantities on the load, the corresponding scoring standards are formulated, and the scores of each three-level index are obtained by using the membership function. The larger the score, the greater the load of the index. Then according to the importance of each index, the weight value of each index is obtained by using the AHP, and based on the cloud model, the weighted deviation degree is obtained, and the cloud map is drawn, and the load is classified through the cloud map. Finally, the score obtained by the feature quantity of the forecast day is calculated, classified according to the load, and classified into its category. Based on the bp neural network, the load data of the category to which the load belongs is used as a training sample to predict the load of the forecast day. The invention has higher classification recognition accuracy and stronger adaptability.

The invention provides an optical power prediction algorithm based on a satellite cloud atlas. The optical power prediction algorithm uses a visible light satellite cloud picture as a prediction datamodel, predicts the cloud shielding condition above a certain specific photovoltaic power station in a plurality of hours in the future by analyzing and judging the cloud picture change condition in aplurality of hours in the past, and corrects the irradiation intensity prediction value of the specific photovoltaic power station through the empirical model, so that the problem that the optical power cannot be predicted due to the fact that the photovoltaic power station is shielded by the cloud is solved, and fluctuation of the cloud layer above the photovoltaic power station is effectively predicted, and the irradiance prediction precision is improved, and then prediction of the generating capacity is improved.

The invention discloses a detection method of a printed circuit board and a detection device. The detection method comprises the following steps: three-dimensional morphology parameters of the printed circuit board are obtained on the basis of a three-dimensional digital image correlation method; and according to the three-dimensional morphology parameters, defect detection of the printed circuit board is carried out. In comparison with existing AOI detection methods, the technical scheme of the invention has the following advantages: the three-dimensional morphology of the printed circuit board can be accurately measured, and visual three-dimensional cloud atlas and accurate three-dimensional morphology data can be obtained, thus solving the problem of defect detection in height direction. In comparison with existing AXI detection methods, the technical scheme of the invention helps raise detection speed and can satisfy on-line real-time detection of the printed circuit board.

The invention discloses a power transmission tower structural mechanical analysis and App (Application) establishment method which comprises the following steps: on the basis of COMSOL Multiphysics software, modeling according to a structure of a power transmission tower and a cross section and a material of a rod, applying external loads such as self weights, wind and ice loads of power transmission towers, leads and insulators, performing stress analysis on the power transmission tower, and generating stress cloud atlas, deformation cloud atlas and result reports; secondly, through an App developer inside the software, designing an App of power transmission tower structure stress analysis by using a form editor and a method editor. The method can be widely applied to finite element modeling and structure stress analysis on a power transmission tower, and by using the App established simultaneously, non-professional operators without skills of finite element modeling and simulation analysis can be enabled to obtain calculation results by rapidly modifying parameters of models.

The invention provides a downhole flow metering device and method for a layered water injection well. A valve of the device is located is located on a Christmas tree; by adjusting the valve, requiredpressure waves are formed, and pressure codes are sent to downhole differential pressure water distributors; a pressure meter meters wellhead pressure data, a flowmeter meters the injection flow, thedownhole differential pressure water distributors receive pressure codes and decode the pressure codes, and an adjusting instruction is obtained, so that a motor of the device is controlled accordingto the adjusting instruction to rotate forward or reversely to a certain open degree, the change that the open areas S of water nozzles can be adjusted is generated, and flow adjusting is conducted; and a ground data processing device receives injection flow data and the pressure data during allocation, the change of the injection flow is collected in real time according to the pressure change, and thus a pressure and flow relation cloud atlas of the layer is formed. According to the downhole flow metering device and method for a layered water injection well, the injection amount of each layercan be obtained precisely without downwards-putting of a cable instrument; and the collecting frequency can be adjusted according to the flow change, and the working efficiency and flow recording accuracy are improved.

The invention discloses an end position judging method of a non-consumable gas shielded arc welding wire. An electric arc is built through a welding control system; voltage between a non-consumable welding gun and a workpiece is acquired by dint of a voltage sensor, and is converted to a corresponding space electric field; an electric field distribution cloud atlas of the electric arc is built; voltage between the non-consumable welding gun and a welding wire and voltage between the welding wire and the workpiece are acquired, and are converted to corresponding space electric fields; the two electric fields are combined with the electric field intensity distribution atlas of the electric arc space; the electric field wire directions at the end parts of the welding wire are respectively obtained through the non-consumable welding gun and the workpiece end; and the intersection of the two is determined as present position of the end parts of the welding wires to realize judgment of the end positions of the welding wire. The method facilitates improvement of the stability in the welding process, and further improves the welding line quality.

The invention discloses a foundation cloud atlas classification method based on spatial pyramid random mapping. The method comprises the following steps of firstly, extracting local features from each training foundation cloud atlas sample in a denseness sampling manner; then, carrying out dimensionality reduction on each local feature by applying random mapping, and mapping an original high-dimensionality feature set to a low-dimensionality subspace; then, clustering features which are subjected to dimensionality reduction in the low-dimensionality subspace, so as to obtain a codebook; then, dividing a sample image into different areas according to a spatial pyramid model, obtaining area features of the different areas according to the codebook, combining the area features, and taking the combined area features as final feature representation of the sample image; finally, obtaining a classification result of a tested foundation cloud atlas by applying a support vector machine classifier. According to the method, spatial information of the image can be obtained through applying the spatial pyramid model, so that information in the cloud atlas can be better represented; meanwhile, the local features of the image are subjected to dimensionality reduction by adopting random mapping, so that the efficiency of a foundation cloud atlas classification system can be increased, the time expense is reduced, and the dimensionality disaster can be avoided.

The invention discloses an engine timing cover cap lightweight CAE analysis method. The method comprises the steps of S1, establishing a timing cover cap initial scheme finite element model; S2, performing modal analysis calculation and frequency response analysis calculation on the finite element model of the initial scheme of the timing cover cap; S3, performing vibration response optimization on the finite element model of the initial scheme of the timing cover cap; S4, establishing a topological structure of the timing cover; S5, establishing a 1D shell unit model of the timing cover and performing morphology optimization; S6, obtaining a timing cover optimization scheme based on the vibration response optimization and morphology optimization result cloud atlas; and S7, a timing coveroptimization scheme: establishing a timing cover optimization model, and carrying out modal and frequency response analysis and local optimization check. According to the lightweight CAE analysis method for the engine timing cover cap, designers can be helped to obtain a lightweight design scheme of the timing cover cap structure on the premise of considering the performance, and the product development speed is increased.

The invention discloses an interdynamic navigation method based on a hotspot vehicle, comprising the following steps of: sending real-time cloud atlas data through a navigation terminal of each driving motor vehicle to a data service center to form an overall road traffic flow cloud atlas; displaying a traffic flow cloud atlas of a target area on the navigation terminal at real time according to a cloud atlas data request of the navigation terminal; selecting a vehicle that interests a user from the traffic flow cloud atlas of the target area as the hotspot vehicle to mark, and tracking the real-time driving information of the selected vehicle, so as to accomplish the interdynamic navigation based on the real-time traffic information sharing.

The invention provides a simulation satellite cloud atlas generation method based on a generative adversarial network and a numerical model product, and the method comprises the steps: collecting andarranging satellite cloud atlas data and a numerical model analysis field product at the same time and in the same region, and constructing a training data set; constructing a simulation satellite cloud picture generation model based on a deep generative adversarial network according to the training data set, and training the model; and according to the simulation satellite cloud picture generation model, reconstructing a simulation satellite cloud picture corresponding to the forecast time by taking the numerical model forecast field product as model input, and outputting the simulation satellite cloud picture as a result. According to the invention, the purpose of generating the simulation satellite cloud atlas by the numerical model product is achieved.

The invention discloses an absolute water potential calculation method suitable for grain condition monitoring. The absolute water potential calculation method comprises the following steps that: temperature and humidity measurement devices are arranged at multiple measurement points in the granary space so as to acquire the temperature value t and the humidity value M of the grain of the measurement points; and the absolute water potential value E<g> of the grain of each measurement point is calculated: E<g>=RT<alpha>ln(EAH<g>)=[8.31x(t+273)xln(EAH<g>x133.3)] / 18, wherein R is the universal gas constant 8.31J / Mol, T<alpha> is absolute temperature of the grain, and EAH<g> is the balance absolute humidity value of the grain and the unit is mmHg. The invention also provides an absolute water potential cloud atlas generation method suitable for grain condition monitoring. The absolute water potential values of multiple prediction points on a prediction section are calculated by using an interpolation algorithm based on the absolute water potential value E<g> of the grain of the measurement points; and an absolute water potential cloud atlas on the prediction section is drawn. The moisture and thermodynamic parameters of the grain bulk are fused to obtain an absolute water potential field and form the cloud atlas so that multiple complex physical fields of the grain bulk can be integrated in one atlas to be expressed in a unified way, and thus analysis of the current state of the grain bulk and the historical evolution rule is facilitated.

The invention discloses a photovoltaic output ultra-short-term prediction method and system based on multi-source heterogeneous data. The method comprises the steps of extracting cloud atlas featuresfrom a foundation cloud atlas through a deep neural network; extracting historical data features from the historical photovoltaic output data; splicing the cloud picture features and the historical data features; and finally, based on the spliced features, generating predicted photovoltaic output data through a one-dimensional convolutional network. According to the invention, the deep neural network and the ultra-short-term photovoltaic output prediction are combined, the advantages of the deep neural network in image feature extraction are utilized, the features are extracted from the foundation cloud atlas and then fused with the photovoltaic output historical data, and the photovoltaic output prediction is realized. The combination of the image and the historical data overcomes the defects of single input data and low information amount of the prediction model, and the deep neural network automatically extracts the cloud image features to overcome the defects of low information utilization rate and weak generalization ability of the artificially designed image features.

The invention discloses a photovoltaic power prediction method based on ground-based cloud picture image features. The method comprises the following steps: S1, extracting illumination intensity, a high-frequency component, transmissivity, a zenith distance and cloud factor features based on a ground-based cloud picture; and S2, establishing a GBDT photovoltaic power prediction model, and constructing the GBDT photovoltaic power prediction model by taking image features such as the illumination intensity, the high-frequency components, the transmissivity, the zenith distance and the cloud factors extracted from the foundation cloud atlas as important feature quantities influencing photovoltaic power, thereby realizing accurate prediction of the photovoltaic power. The photovoltaic power prediction result obtained through the method has high precision and certain stability, and the method is simple and easy to implement.

The invention discloses a system and a method for dynamically constructing a regional natural disaster risk cloud picture. The method comprises the following steps: acquiring environmental data of a target region; performing standardization processing on the environment data; determining evaluation indexes and influence factors according to the environment data; determining each evaluation index and influence factor weight; calculating a regional risk index according to the evaluation index and the influence factor; and drawing and displaying a risk cloud picture according to the regional risk index. According to the invention, the original simple defense of natural disasters is changed into comprehensive management of risks, the prevention and control capability of the natural disasters is greatly improved, and the huge loss of the disasters to the country and people is reduced.

The invention is suitable for the field of temperature monitoring, and provides a method and device for rendering a three-dimensional temperature cloud atlas. The method includes the following steps: establishing a three-dimensional temperature cloud atlas according to positions of temperature sensors in a monitoring machine room; and rendering a three-dimensional temperature cloud atlas of the monitoring machine room according to temperature values of the temperature sensors. The method and device provided by the invention improve the degree of accuracy of the temperature cloud atlas, and facilitate monitoring of the machine room temperature by a user.

The invention relates to a thermal design method used for radar incubator TEC (Thermo Electric Cooler) model selection. The method comprises the following steps that: through theoretical calculation,finishing the model selection of a TEC, a fan and a radiator in a radar incubator; establishing the three-dimensional model of the radiator incubator; establishing a three-dimensional gridding calculation domain; carrying out simulation calculation on the radar incubator to obtain an initial simulation result; and establishing the contour line cloud atlas of temperature distribution and the flow path line of fluid, and improving the internal structure and layout of the radar incubator which does not conform to job requirements. By use of the method, a research and development period can be shortened, and economic benefits are improved. In a practical use process, the radar incubator obtained through the method of the invention has a good heat dissipation effect, and meanwhile, the radar incubator has the advantages of compact structure, miniaturization and good waterproof sealing and electromagnetic shielding effects.

The invention relates to an atmospheric pressure influence considered mine goaf temperature field simulation method which is used for obtaining a total pressure value p<in> at an air inlet throat and a total pressure value p<out> at an air return throat. A set of simultaneous plane temperature field model equations is formed, and a wind pressure boundary condition equation and a thermodynamics boundary condition equation are added; a numerical solution is obtained in computational fluid dynamics software FLUENT to obtain a goaf temperature field simulation cloud atlas under certain ground atmospheric pressure; and a rule that if the atmospheric pressure is increased, the absolute pressure of air in the goaf is increased, the flow velocity is caused to decrease, and therefore low heat convection and conduction efficiency and small temperature distribution gradient is caused due to multi-field coupling is obtained, and the rule conforms to a practical situation of a mine. The method has the advantages of small error and high precision.

The invention relates to a typhoon intensity remote sensing inversion method based on deep learning, and the method comprises the following steps: determining the geographic coordinate information ofa typhoon center at a to-be-inverted moment, and obtaining a cloud atlas of a satellite at the to-be-inverted moment; reading geographical coordinate information in the satellite cloud picture and preset brightness temperature data of a plurality of wave bands; determining a coordinate point of a typhoon center position in the satellite cloud atlas, and constructing a three-dimensional matrix based on waveband brightness temperature data by taking the coordinate point as a center; constructing eight binary classification CNN models based on a focaliss loss function; inputting the three-dimensional matrix data into eight binary CNN models to obtain and output a corresponding numerical value, the typhoon level corresponding to the maximum numerical value being the typhoon intensity categoryat the moment; and calculating the maximum wind speed of the typhoon at the moment according to the obtained values. The method can effectively solve the problems that in an existing deep learning method, it is difficult to find the optimal channel combination for typhoon intensity inversion, and side effects are caused by unbalanced typhoon sample distribution.