29results about How to "Increase calculation rate" patented technology

The present invention belongs to the field of quantum navigation and positioning technology, and discloses an inter-satellite ranging system, method, device and data processing terminal. The inter-satellite ranging system includes: a photon transceiver unit, which is used to generate entangled photon pairs, transmit signal photons to the satellite to be measured and the internal optical path, and receive signal photons reflected by the satellite to be measured and the internal optical path; a coincidence counting unit, which is used to associate the received signal photons with the idle photons to obtain their second-order correlation curves; a distance measurement unit, which is used to perform DFT phase difference weighted processing on two second-order correlation curves generated by the optical path difference inside and outside the system to obtain the actual optical path difference from the satellite to be measured to the system. The present invention uses a photon matching method to filter out the cosmic space stray light entering the photon detector to a large extent, thereby improving the signal-to-noise ratio. The present invention adopts an improved coincidence counting algorithm, and the calculation rate is greatly improved compared with the traditional coincidence counting algorithm.

The invention provides an autonomous navigation method and device for an inspection robot, the robot and a storage medium, and the method comprises the steps: obtaining image information, determining the visibility of an inspection environment based on the image information, determining an inspection working condition according to a comparison result of the visibility and a preset visibility threshold value, and carrying out the autonomous navigation of the inspection robot. If the inspection working condition is a daytime inspection working condition, determining a selectable path and first pose information of the inspection robot based on the image information, calculating the driving direction of the inspection robot in combination with the selectable path and the first pose information, and if the inspection working condition is a night inspection working condition, calculating the driving direction of the inspection robot based on a preset path and second pose information obtained in real time; calculating the driving direction of the inspection robot, taking the driving direction as a navigation result, and controlling the inspection robot to autonomously drive according to the navigation result; different driving direction determination modes are adopted under different inspection working conditions, and the flexibility of the autonomous navigation method of the inspection robot is embodied.

The invention discloses a remote sensing image supervision and classification method based on decision tree correction, and the method comprises the steps: obtaining a to-be-classified satellite image and a reference image, the reference image being a satellite image which covers a research region of the same sensor and has a verifiable ground feature type; all the images are preprocessed, and image data DN values are converted into earth surface radiation values; cutting the processed image to obtain image data of the area needing to be researched; respectively synthesizing the preprocessed wave band data of each image and various parameter results calculated based on the wave band data into new synthesized image data; drawing a region of interest for the image data of the reference composite image, and making a classification decision tree of the image through machine learning; classifying other to-be-classified composite image data based on the classification decision tree, and drawing new regions of interest for the composite images by referring to the classification result; and utilizing a support vector machine to re-classify the surface features of the synthesized image data. The method improves the classification precision and speed.

The invention discloses a data compression method and system based on tensor transformation, and the method comprises the steps: converting original data into tensor data of a first dimension according to the features and dimensions of the original data; performing dimension reduction processing on the tensor data of the first dimension through a tensor decomposition technology to obtain a plurality of pieces of tensor data of the second dimension; wherein the first dimension is greater than the second dimension; reconstructing the tensor data of the second dimension to obtain tensor representation of the second dimension of the original data; and based on dictionary coding, performing data coding and compression on the tensor representation of the second dimension to obtain compressed data corresponding to the original data. The tensor transformation technology is used in the data acquisition and compression process of the model, so that the data compression speed and effect are improved.

The invention provides a multi-factor analysis system and method suitable for stock markets and a medium, and the system comprises a data integration module which obtains original financial data, and carries out the preprocessing of the original financial data; the factor analysis module is used for performing factor analysis on the preprocessed financial data through a built-in factor library and a factor calculation engine to obtain an analysis result; the factor analysis comprises single factor analysis, multi-factor analysis and attribution analysis; the factor library is dynamically set based on stock markets; and the factor calculation adopts a distributed calculation architecture. According to the system, the factor library and the factor calculation engines are set according to stock market rules, the practicability of the analysis system in the stock market is improved, the factor calculation engines are set in parallel, the calculation rate is greatly increased, factor design is dynamically adjusted according to market conditions, and it is guaranteed that the multi-factor analysis system can adapt to the rapidly-changing market conditions.

The invention provides a slurry permeation discrete element fluid-solid coupling optimization simulation method based on pore morphological characteristics, which comprises the following steps of: obtaining spatial geometric information of sand through a CT (Computed Tomography) scanning method and a digital three-dimensional reconstruction technology, and establishing an STL (Standard Template Library) three-dimensional graphic file of a sand stratum; and importing the STL file into a computational fluid dynamics-discrete element coupling program to establish a pore flow field grid and a discrete element wall boundary, and establishing a discrete element fluid-solid coupling calculation model, thereby realizing the calculation demand of restoring the formation pore morphology characteristics. In order to reduce the calculation amount and improve the calculation efficiency, the total number of slurry particles calculated in real time is reduced in a discrete element module through a method of circularly inserting belt speed particles and deleting filter-off particles. By utilizing the method, original size simulation can be carried out according to the actual shape of an actual stratum and the actual size of actual slurry particles, and a new thought is provided for the research method of the shield tunnel excavation face slurry permeation micromechanism.

The invention relates to the technical field of blast furnace ironmaking, and discloses a hearth erosion three-dimensional identification method, system and equipment and a storage medium. The method comprises the steps that a hearth is three-dimensionally segmented into a plurality of equal-angle block areas based on thermocouple position distribution; establishing a three-dimensional heat transfer calculation model of the hearth, and generating erosion control boundaries and measuring point temperature data sets under different erosion conditions of the partitioned regions; establishing a neural network model group for different block regions; performing offline model training on the neural network model group based on the erosion control boundary and the measuring point temperature data set to obtain a trained hearth erosion identification model group; and collecting actual temperature measurement data, inputting the actual temperature measurement data into the trained hearth erosion identification model group to output multiple groups of calculation results, and combining the multiple groups of calculation results to obtain a hearth erosion three-dimensional identification result. According to the method, the three-dimensional heat transfer calculation model of the hearth is established, the model training is completed offline by using the measuring point temperature data set in blocks, the hearth erosion identification result is rapidly given online according to the actual temperature measurement data, and the calculation rate is high.

The present invention relates to a GNSS RTK positioning and satellite selection method and system driven by both models and data. A hybrid model of observation residual altitude angle and signal-to-noise ratio is fitted using large sample GNSS data, the theoretical noise of each satellite is calculated and large gross errors are eliminated, and then the pre-selection of satellites is completed according to the contribution value of each satellite to GDOP. The precise velocity information obtained by Doppler velocity measurement is used to construct the kinematic equation of the carrier, the predicted position of the current epoch is obtained, and it is substituted into the double difference observation equation to calculate the pre-test residual, and the Grubbs criterion is used to eliminate outliers and roughly screen the available satellites. A code phase combination double difference model is constructed to estimate the observation noise in real time and obtain the variance covariance matrix of the observation value, thereby calculating the post-test standardized residual to perform precise screening of the ambiguity. The present invention can select high-precision observation data for positioning solution, improve the fixation rate, and thus improve the positioning accuracy, while reducing the consumption of computing resources caused by matrix operations.

The invention discloses a multi-cluster air computing network-oriented transmitting-receiving cooperation method and system, and belongs to the field of air computing in a wireless network, and the method comprises the steps: firstly constructing a multi-cluster air computing scene, and analyzing an expression of an air computing rate and key influence factors of the expression; secondly, decomposing the problem into a receiving beamforming sub-problem and a transmission factor control sub-problem according to physical significance; deriving a closed-form solution expression of the receiving beamforming sub-problem according to the receiving beamforming sub-problem; for the transmission factor control sub-problem, continuous convex approximation is used for solving; then, jointly solving the two sub-problems by adopting an alternating optimization algorithm to obtain an optimal weighting sum rate; finally, system parameters are changed, and the change relation between different parameters and the weighting and calculation rate is studied. The scheme provided by the invention is superior to the existing method in the aspect of calculation rate, and reveals the influence of key parameters on the air calculation performance.

The invention provides an all-optical neural network device and system for speckle image classification and recovery, and the training method comprises the steps: obtaining a training sample set which comprises an original image, a speckle image outputted by the original image through an optical fiber, and a classification label image corresponding to the original image; the training sample is input into the initial all-optical neural network, the training sample comprises a first convolution layer, a second convolution layer and a full connection layer, and the full connection layer comprises an image restoration area and an image classification area. And after optical convolution operation is carried out on the speckle image in the first convolution layer and the second convolution layer in sequence, the speckle image is transmitted to the full connection layer to combine convolution results, and a recovered image and a classified image are output. And constructing a composite loss function by using mean square errors and Pearson's correlation coefficients between the recovered image and the original image and between the classified image and the classified label image. And the loss is minimized to update the initial all-optical neural network parameters until training is completed, and a trained all-optical neural network is obtained. Noise and loss can be reduced, and calculation efficiency can be improved.

The invention discloses a large language model (LLM) floating point operation optimization method based on POSIT, and the method comprises the following steps: customizing an RISC-V instruction, and enabling the RISC-V instruction to adapt to Point format operation; performing weight quantization on the weight parameter of the LLM by using a Point number; a floating-point number used in the LLM training process is converted into a Point number; a floating-point number used in the LLM reasoning process is converted into a Point number; and carrying out model quantization by using a Posit number. The invention aims to solve the current situation that an IEEE 754 floating-point number format is adopted in the floating-point operation of the current LLM, a Point floating-point number is used for replacing the IEEE 754 floating-point number, and by means of the self-adaptive precision and the dynamic range of the Point floating-point number, the size of the LLM model is reduced and floating-point number calculation is accelerated while various calculation anomalies are avoided and high precision is kept, so that LLM reasoning training is accelerated.

This invention provides a method, apparatus, electronic device, and storage medium for determining GeoHash codes. First, the minimum bounding rectangle of the region to be identified is determined. Next, the GeoHash codes corresponding to the vertex grids containing any three vertices of the minimum bounding rectangle are determined. Then, based on the GeoHash codes of the three vertex grids, a preset character table is looked up multiple times to determine the GeoHash codes of all grids within the minimum bounding rectangle. Finally, from the GeoHash codes of all grids, the GeoHash codes of each target grid that overlaps with the region to be identified are selected. In this invention, the preset character table includes 32 characters from a Base32 encoding table, and the positions of these 32 characters in the preset character table are determined based on the encoding principle of GeoHash codes. Therefore, the GeoHash codes of all grids within the minimum bounding rectangle can be determined by looking up the table. Compared to the multiple encoding conversions in existing technologies, this invention uses a table lookup method, reducing the computational load and improving the computational speed.

The invention discloses a numerical simulation method and device for a hydrogen-doped natural gas pipeline conveying process. The method comprises the following steps: determining a first equation expression corresponding to a pipeline element in a hydrogen-doped natural gas pipeline network, determining a second equation expression corresponding to a non-pipeline element, and determining a third equation expression corresponding to a gas source and a fourth equation expression corresponding to an element connection point; aiming at a to-be-simulated moment, solving all the equation expressions to obtain hydraulic parameters and thermal parameters in the hydrogen-doped natural gas pipeline network at the to-be-simulated moment; during solving, a hydraulic equation set composed of the hydraulic equation expressions in all the equation expressions is solved, and then a thermal equation set composed of the thermal equation expressions in all the equation expressions is solved. According to the embodiment of the invention, all equations determined by the hydrogen-doped natural gas pipeline network can be decoupled, so that the numerical solution difficulty of a simulation process is reduced, and the calculation rate is improved.

The present invention provides a multi-sequence seabed intelligent water sampling system and method based on deep learning. The system includes a multi-sequence water sampler, an underwater camera, a sonar, an environmental parameter sensor unit and a power unit arranged on the multi-sequence water sampler; and a central processing unit; the central processing unit is used to receive data collected by the underwater camera, the environmental parameter sensor unit and the sonar in real time, and use a pre-trained deep learning model to perform image recognition on the data collected by the underwater camera, and control the multi-sequence water sampler to perform seawater sampling according to the recognition result and the data collected by the environmental parameter sensor unit and the sonar, and control the power unit to drive the multi-sequence water sampler to achieve adaptive movement; the present invention can realize tasks such as accurate selection of seabed sampling points, construction of three-dimensional seabed terrain, adaptive selection of water sampler sampling timing, and adaptive sample detection around seabed sampling points.

The invention belongs to the technical field of ground plane laser measurement, and particularly relates to a stratum deformation early warning system in the parallel stacking construction process of an overpass existing pipeline, which comprises an acquisition module, a cloud processing center and an early warning response module, the collection module completes data collection through a laser monitoring device arranged at the construction position of the existing pipeline parallel stacking construction tunnel, and the collection module sends collected monitoring data to the cloud processing center; the cloud processing center calculates the feedback monitoring data to judge whether formation deformation occurs, if the formation deformation occurs, whether an early warning threshold value is reached is judged, and if the early warning threshold value is reached, an early warning notification is sent to an early warning response module; and after the early warning prompt module obtains the early warning notification, emergency early warning is carried out through the cloud processing center and the laser monitoring device. Whether the stratum deforms or not is measured through laser, the measurement result is accurate, and early warning prompt is safer.

The present invention discloses a quantitative calculation method for infrared multi-radiation source coupled transmission, comprising: obtaining three-dimensional scene information, and dividing the scene into a number of triangular facets; calculating the angular coefficient between each triangular facet; calculating the spectral radiance of the object in the scene; and calculating the spectral radiance of the scene according to the angular coefficient and the spectral radiance of the object. The present invention realizes the quantitative calculation of infrared multi-radiation source coupled transmission by using a radiance algorithm, and can output the radiance distribution of the scene in different spectral bands in the form of an image, making the algorithm easier to implement, and for multi-radiation source scenes, there is no need to perform a large amount of sampling of light sources and light, and the calculation rate is faster.

The invention discloses a method for establishing a depression diagnosis model based on dry blood near infrared spectroscopy, and belongs to the technical field of depression diagnos.The method comprises the following steps that S1, blood samples of a depression patient and a healthy control are collected and stored, S2, a dry blood sample is prepared, the near infrared spectroscopy of the dry blood sample is collected to form a spectrum set, S3, data of the near infrared spectrum set are preprocessed, and a depression diagnosis model is established. The method comprises the following steps: S1, establishing a plurality of basic models, S4, after screening and combining important features, establishing a model by a machine learning algorithm, and selecting an optimal model as a final spectrogram recognition model, and S5, collecting a near infrared spectrum of dry blood of a patient to be diagnosed, and substituting the near infrared spectrum into the spectrogram recognition model to quickly recognize a result. According to the method, sample pretreatment is not needed, near infrared spectrum recognition of a patient to be diagnosed can be rapidly completed only through trace blood, compared with a spectrum of liquid blood, the collected near infrared spectrum of dry blood has no strong interference of a water peak and has more and stronger characteristic signals, important characteristics are screened out for modeling, and the calculation rate and the model accuracy are improved.

The invention discloses a real-time enhancement processing method, system and device for a radiotherapy CT image and a medium, and relates to the technical field of digital image processing, and the method comprises the steps: obtaining a CT image picture, calculating the gray pixel number and the total pixel number, and outputting the occurrence probability; the occurrence probability is calculated, and a probability accumulated value is output; adjusting the CT image picture, and outputting a balanced CT picture; calculating a pixel reference value, and calculating a neighborhood difference value; marking the pixel points as edge points, performing color enhancement processing on the balanced CT image based on the edge points, and outputting an enhanced image; the method is used for solving the problems that in the prior art, due to the lack of analysis for preprocessing the CT image, when the difference between the edge of the focus of the CT image obtained through scanning and the normal tissue is small, the edge of the focus is difficult to segment only by means of a computer, manual participation in division processing is needed, and the division efficiency is low.

The present invention discloses a broadband electromagnetic simulation method using efficient adaptive frequency scanning. The present invention proposes to use an error formula to judge the convergence of frequency response, thereby being able to obtain appropriate frequency sampling points, which reduces the required sampling points and speeds up the simulation compared to other methods. The present invention proposes to convert a nonlinear problem into two linear problems, iteratively calculate the precise pole and residue values, solve the linear equation, obtain an accurate frequency response fitting value, speed up the calculation rate, and improve the accuracy.

The invention provides a time sequence wave neural network (TSWNN), which is used for solving an optimal cognitive sub-graph query problem (MCSQR), and is characterized in that the optimal cognitive sub-graph query problem (MCSQR) is solved by using a time sequence wave neural network (TSWNN), so that the optimal cognitive sub-graph query problem (MCSQR) can be solved by using the time sequence wave neural network (TSWNN), and the optimal cognitive sub-graph query problem (MCSQR) can be solved by using the time sequence wave neural network (TSWNN). The MCSQP aims to query a path from a starting point to an end point in a specific time in a network, optimize node cognition degree accumulation and realize query of an optimal cognition subgraph. The proposed TSWNN is a novel neural network model, has time dependence and transmission time limitation, and is used for simulating and analyzing the process of information spreading and node activation in the network. All neurons of the method are in parallel calculation, and the model does not need any training. Each neuron in the TSWNN is composed of six parts: an input unit, a wave receiver, a neuron state memory, a wave generator, a wave transmitter and an output unit. The optimal cognitive sub-graph needs to find a network from a starting point to an end point, a network where time sequence waves arrive within given time is obtained, then the final cognitive degree is calculated, and the neural network deals with the optimal cognitive sub-graph problem by relying on the time sequence waves. Experimental results show that the method provided by the invention can solve the globally optimal solution of the problem.

The embodiment of the invention provides a compressible flow simulation method and device, computer equipment and a storage medium, and the method comprises the steps: arranging a multi-core structure of a target chip architecture into a preset two-dimensional structure, and obtaining a plurality of operation cores of the target chip architecture; dividing preset compressible flowing fluid data into a plurality of data blocks, and inputting the divided data blocks into each operation core; in each operation core, fluid parameters at the moment n + 1 in the data with the number k are determined according to the fluid parameters at the moment n in the data with the number k, n is an iteration time step, and the value ranges from 0 to a preset value; an analog calculation result of the target chip architecture on the preset compressible flow is determined according to the fluid parameters at the n + 1 moment under the condition that n is a preset value in all the operation cores, parallel calculation of the multiple operation cores and data transmission among the operation cores are achieved through the multiple operation cores of the target chip architecture, the complexity of intermediate transmission is reduced, and the data transmission efficiency is improved. And the calculation efficiency is improved.

The invention provides a target tracking method and device for a polar region unmanned ship, electronic equipment and a storage medium, and relates to the technical field of target detection, and the method comprises the steps: obtaining distance information, unmanned ship driving information and to-be-processed image data; when the distance information is greater than a preset distance, obtaining a to-be-detected target condition; when the to-be-detected target is in a static state, carrying out gray processing on the to-be-processed image data and then carrying out feature extraction to obtain a first to-be-detected target feature; when the dynamic condition of the to-be-detected target is determined, inputting the to-be-processed image data into the double-layer feature fusion model to obtain a second to-be-detected target feature; when the distance information is smaller than or equal to a preset distance, inputting the to-be-processed image data into a YOLO convolutional network model to obtain a third to-be-detected target feature; and determining the position of a to-be-detected target according to the first to-be-detected target feature or the second to-be-detected target feature or the third to-be-detected target feature. According to the invention, the target tracking precision is improved.

The application discloses a food material recognition method of an intelligent refrigerator and a related device. The method obtains a to-be-processed image by performing a defogging process on a target image. The stored food materials in the to-be-processed image are recognized. If the food material names of the stored food materials are not recognized, the to-be-processed image is compared with a stored food material image set in terms of similarity. In the above process, the to-be-processed image after the defogging process is recognized first to improve the recognition ability of the stored food materials. Since the stored food material image set is a collection of stored food material images corresponding to each food material stored in the intelligent refrigerator, when the stored food materials after the defogging process cannot be recognized, the to-be-processed image can be compared with the stored food material image set in terms of similarity to obtain a target food material image most similar to the to-be-processed image. Then, the food material name of the stored food material in the to-be-processed image is determined according to the food material corresponding to the target food material image, so that the food material misrecognition rate is reduced.

This invention discloses a method for obtaining satellite laser timing pointing. By using the six root numbers of the satellite orbit, the nadir trajectory and foot point trajectory of the satellite are obtained. Combined with the elevation database, the continuous laser pointing of the satellite is calculated, which greatly simplifies the laser pointing calculation process and improves the simulation calculation speed. The input used in this invention does not involve complex parameters such as laser attitude angle. Only TLE ephemeris and elevation data need to be input. The requirements for the function configuration of the scene simulation software are not high, and the calculation cost is low.

The present invention discloses a method, system, medium and program product for detecting cracks in building exterior walls, wherein the method comprises: obtaining an image of a building exterior wall to be detected, inputting the image into a building exterior wall crack detection model, and outputting a detection result of the building exterior wall cracks. The building exterior wall crack detection model comprises an encoder and a decoder; the encoder adopts a VMamba encoder to extract feature maps of different resolutions in four stages; the decoder is constructed using the visual state space block and convolutional neural network of VMamba, and a detail-guided gated fusion mechanism is introduced to fuse the feature maps of different resolutions in each stage of the encoder and decoder, and then a residual module and a classification head are used to obtain the crack detection result. The present invention improves the calculation rate and the ability to extract complete cracks; the introduction of the detail-guided gated fusion mechanism can better fuse the encoding features and decoding features from different stages, and improve the detection accuracy and robustness of the building exterior wall cracks.

The invention discloses a method for rapidly solving harmonic power flow of an MMC (Modular Multilevel Converter), belongs to the field of direct-current power transmission, and aims to overcome the defect that the harmonic power flow is difficult to solve conveniently by a time domain simulation method. A complete MMC harmonic state space model considering control links such as a V / F control outer loop, a current inner loop, a power grid voltage feed-forward loop and a circulating current suppression loop is established. And then aiming at the defect that the time for solving the steady-state power flow is too long in time domain simulation, the invention provides a frequency shift factor based on the characteristics of a periodic time-varying system in a frequency domain, so that the dimensionality of an MCC model can be greatly reduced, and the iterative solving speed of the harmonic power flow is accelerated.

The invention provides a distribution network decision analysis method and device related to a container cloud platform, and relates to the technical field of power grids. According to the method, the decision analysis request is disassembled to obtain a plurality of intermediate parameters such as power distribution network parameters, the container corresponding to the power distribution network parameters and the analysis result is determined, the computing power task architecture is constructed, the project design data is analyzed, and the analysis result is obtained. In the process, the whole calculation task is divided into multiple container calculation, the calculation time of the decision analysis task is shortened, the calculation rate of the power distribution network project design is improved, the problems that the power distribution engineering project design is difficult and low in efficiency can be solved, the analysis efficiency of the power distribution network project design is improved, and the power distribution network project design efficiency is improved. And decision analysis of design experts and review experts is facilitated.

The present invention relates to the technical field of hydrological forecasting, and discloses a universal resource file definition and interaction method, system, medium and device for a hydrological model, which includes: collecting hydrological models and establishing a resource directory set of the hydrological model, wherein the resource directory set includes various hydrological model codes, types and applicable areas; collecting universal resource files of the hydrological model, wherein a resource set is composed of a hydrological model configuration resource file, a model input resource file and a model output resource file; obtaining resource files according to the resource set, establishing a topological relationship between the hydrological model and the resource file, and forming a system configuration file; and obtaining a resource file list based on the hydrological model code by reading the system configuration file, and presenting the list. The present invention is conducive to quickly processing various types of data of the hydrological model.