27results about How to "Improve computing performance" patented technology

The invention relates to a three-dimensional instance segmentation method in a paddy field high-density shielding scene in combination with a multi-modal fusion network. The method comprises the steps of collecting a multi-view RGB image and three-dimensional point cloud data of a paddy field for preprocessing; inputting the multi-view RGB image in the multi-source data set into the image segmentation model to generate an image instance segmentation result; matching the image instance segmentation result with the three-dimensional point cloud data in the multi-source data set; and inputting the point cloud data containing the multi-view RGB image information into the improved three-dimensional instance segmentation network model, and outputting a three-dimensional instance segmentation result. According to the method, accurate segmentation in a high-density shielding scene is realized, and the segmentation accuracy is remarkably improved; effective fusion of multi-modal data is realized, the robustness and precision of instance segmentation are effectively improved, and the method is especially suitable for a fine-grained segmentation task in a dense vegetation scene. The calculation complexity is reduced, the processing efficiency is improved, and the long-distance dependency relationship in the point cloud can be captured while the efficient calculation performance is kept.

The invention discloses a scheduling method, device and equipment for a natural resource multimode computational analysis model, and the method comprises the steps: carrying out the virtualization of a computational analysis model, and achieving the model integration across multiple modes; performing calculation mode and calculation resource scheduling on the data in the calculation process, and subdividing the data into a plurality of calculation sub-tasks to support calculation of large-scale data; and executing scheduling strategies under different scales according to scale characteristics of data calculation and analysis. The method aims at solving the problems that in an existing method, calculation and analysis models of different modes are similar in function, parameter setting is not unified, meanwhile, scheduling of calculation resources is not refined enough, and timeliness and accuracy of calculation and analysis are insufficient under the condition that resources are deficient. According to the method, a unified virtualization model is adopted to integrate a multi-modal analysis model, and calculation resources are finely scheduled to improve timeliness and accuracy. Meanwhile, resource use is optimized, large-scale data calculation and analysis are supported, and efficiency can be improved by rapidly utilizing newly-added resources.

The invention relates to the technical field of intelligent traffic, and particularly discloses a pedestrian movement track prediction method and system based on visual information and a storage medium, and the method comprises the following steps: unmanned aerial vehicle data collection, coordinate projection transformation, coordinate sampling and track aggregation, sampling of pedestrian ground real coordinates after projection transformation, and tracking of the pedestrian movement track. Aggregating the coordinate data into trajectory information of a plurality of pedestrians; spatial visual information and time information are combined, a global interactive directed topological graph is constructed, then spatial feature and time feature extraction is performed, time sequence features of pedestrian motion are efficiently extracted through a multilayer convolution structure, and finally the future motion trail of pedestrians is predicted. Through the method, pedestrian motion in a complex scene can be predicted more efficiently, and the application effect of the model in pedestrian flow management and public safety planning is enhanced. The future movement condition is predicted by analyzing the historical trajectory of pedestrians, the pedestrian flow management is obviously optimized, and the safety level of public places is improved.

The invention discloses an SKU intelligent classification and label generation method fused with a visual pre-training model. The method comprises the following steps: obtaining image data and text data of an SKU; extracting multi-modal features based on the image data and the text data; according to the multi-modal features, zero sample classification of the SKU is realized; generating an SKU label through an improved fractal condition variational auto-encoder based on a multi-modal feature; on the basis of newly added SKU data, continuous learning of the model is realized; carrying out compression and mobile terminal deployment on the model; through an innovative fractal attention encoder and a chaotic dynamics condition generation decoder, the classification accuracy and the label generation quality are remarkably superior to those of an existing method, the BLEU scores of 95.8% and 0.76 are achieved respectively, commodities can be classified and described more accurately, and the search and recommendation effects of a platform are improved; and due to the continuous learning ability and the dynamic category extension mechanism, the performance is excellent when emerging commodity categories are processed, and the recognition rate reaches up to 89.2%.

The invention relates to an insect type identification method based on a multi-dimensional radar electromagnetic scattering feature sequence, and aims to accurately identify insect types through electromagnetic scattering features of radar echo signals in combination with a deep learning technology. According to the invention, a multi-dimensional feature sequence is constructed based on a fully-polarized radar scattering matrix, and feature learning is carried out by using an insect identification neural network. And the insect identification neural network extracts spatial features of different scales through multi-scale convolution, updates a time dependency relationship in an effective capture feature sequence through a state input based on a time sequence, and finally identifies insect types based on the multi-scale spatial features and the time dependency thereof. According to the technical scheme, the accuracy and robustness of insect type identification are effectively improved, the average accurate identification rate of nine kinds of migratory flying pests in China can reach 98.1%, and the method has wide application prospects and is particularly suitable for the fields of insect migratory flying monitoring, agricultural pest control and ecological research.

The invention belongs to the technical field of remote sensing, and particularly relates to a nuclear power plant surrounding ground feature element extraction method based on deep learning, and the method comprises the steps: 1, making a nuclear power plant surrounding ground feature deep learning sample; step 2, based on the HRNet model, constructing a Mama attention residual module, an inner convolution residual module and an asynchronous convolution residual module through Mama, inner convolution and heterogeneous convolution, and constructing an improved HRNet model; 3, training of a U-Net model, a Deeplabv3 model, a HRNet model, a Swin Transformer model and an improved HRNet model is completed through the sample set training data and the verification data; 4, evaluating land element extraction results of different deep learning models; and 5, extracting elements of the surrounding land of the large-scale nuclear power plant by using the trained improved HRNet model. The method can effectively improve the extraction precision of complex ground feature elements around the nuclear power plant, remarkably improves the calculation efficiency of the model, and is suitable for efficient monitoring and classification of multiple ground feature elements around the nuclear power plant in a large range.

The invention belongs to the technical field of machine vision automatic identification, and particularly relates to a radioactive article automatic identification method based on machine vision, and the method comprises the steps: labels on radioactive articles are automatically identified and clamped, and the labels are square labels with the same size, different from the background color and different in color; according to the method, the radioactive article is measured, the radiation data of the radioactive article is measured, and the radiation field is reconstructed according to the radiation data, so that the problems that working personnel are affected and the measured and acquired data are insufficient when the space dose rate in the retired environment is acquired are solved. According to the radioactive substance clamping and measuring device, the complexity and danger of radioactive substance processing in the nuclear industry are dealt with, clamping and measuring of radioactive substances are carried out through the mechanical arm, the clamping accuracy is improved, the working efficiency is greatly improved, and the safety of workers is guaranteed.

The invention relates to the technical field of electrical equipment state monitoring, and discloses a converter valve multi-element monitoring platform data processing method and device, a medium and equipment.The method comprises the steps that firstly, simulation data samples of converter valve multi-physical fields under all working conditions are obtained, and then a least square generative adversarial network is constructed and trained; the method comprises the steps of obtaining a simulation data sample, obtaining a multi-physics field parameter sample, combining the simulation data sample with the multi-physics field parameter sample to obtain an extended data sample, training and evaluating a preset evaluation model by using the preprocessed extended data sample to generate a target evaluation model, and finally establishing a fault prediction model based on the extended data sample and the target evaluation model. And processing the multi-physical field data of the converter valve by using the fault prediction model to obtain a converter valve fault prediction result. According to the method, the diversity of the data, the calculation capability of the model and the prediction precision of the model are effectively improved, and efficient processing and fault prediction of the multi-physical field data of the converter valve are realized.

The invention discloses a terahertz SAR (Synthetic Aperture Radar) image self-focusing method without vibration model constraint, relates to a terahertz SAR image self-focusing method, and belongs to the technical field of terahertz SAR. The invention aims to solve the problems that in the terahertz SAR image focusing process, image focusing is insufficient due to limitation of a specific vibration model, and the nonlinear equation solving process in the self-focusing process is complex. The method comprises the following steps: acquiring a terahertz SAR echo signal; the terahertz SAR echo signal comprises a phase error; the image entropy of the terahertz SAR image is obtained; solving a first-order derivative of the image entropy S of the terahertz SAR image about the phase error, and enabling the first-order derivative of the image entropy S about the phase error to be zero to obtain a nonlinear equation to be solved; solving the nonlinear equation based on a Gaussian-Newton method to obtain a phase error estimation value which is not constrained by a specific model; and constructing a compensation function based on the phase error estimation value to obtain a final clear SAR imaging result.

The invention provides an image processing method and device based on element-by-element operation, equipment and a medium, and relates to the technical field of GPUs (Graphics Processing Unit), and the method comprises the following steps: according to the number of a plurality of elements contained in a plurality of input tensors needing to participate in operation and hardware resource information of GPU equipment, determining the number of the plurality of elements contained in the plurality of input tensors needing to participate in operation; determining the number of thread blocks required to be used by the plurality of input tensors in the element-by-element operation process; according to the number of the thread blocks, the number of the elements and the memory layout of the multiple elements, kernel functions of element-by-element operation class operators corresponding to the multiple input tensors needing to participate in operation are determined on the basis of a predefined template structure body, and parameters of the template structure body comprise element types and the number of the elements; and based on the number of the thread blocks, initiating a plurality of threads to enable the plurality of threads to obtain the plurality of elements according to the kernel function, and executing an element-by-element operation process of the plurality of input tensors to obtain an image processing result.

The invention discloses an equidistant offset method of a closed-loop graph, any closed-loop graph and an offset distance d are input, a graph after equidistant offset of the graph can be obtained, the closed-loop graph is a closed contour composed of arcs and line segments, and no graph is arranged in the contour. The offset method specifically comprises the following steps of inputting any closed-loop graph and an offset distance d, performing inspection and merging preprocessing on an input object, performing normal offset on all edges, generating an initial offset curve, performing repair processing on the initial offset curve, gradually obtaining an offset graph and outputting a result. The method has the advantages of being wide in applicability, high in precision and stability, capable of processing complex graphics, good in calculation efficiency and the like, and can be widely applied to various fields of computer aided design, manufacturing, simulation and the like.

This invention discloses a joint optimization method and system for task offloading and resource allocation based on SAGIN. It collects task and scheduling information through a SAG-IoT system; utilizes the task data and scheduling information to achieve optimal association between UAVs and IoT devices based on a matching game algorithm, and makes optimal decisions for task offloading to maximize the computational performance of the SAG-IoT system; based on optimal association, optimal task offloading decisions, and maximizing the computational performance of the SAG-IoT system, it constructs an online offloading framework based on deep reinforcement learning, constructing a path from input X(t) to optimal action x. * The low-complexity mapping strategy (t) is based on repeated interactions between different modules and a random environment, iterating and running sequentially to achieve joint optimization of task offloading and resource allocation. This invention achieves optimal computational performance while stabilizing the system queue. In addition to utilizing binary computation offloading, the optimization framework can also be extended to online partial computation offloading strategies consisting of multiple independent subtasks.

The application discloses a kind of multi-micronet robust game optimization scheduling method and system based on dynamic auction official algorithm, and relates to energy system scheduling and optimization technical field.The method constructs system physical model and double-layer robust optimization model based on Stackelberg master-slave game, upper layer maximizes system operator's profit to formulate price signal, lower layer maximizes the worst scenario income of multi-micronet alliance to optimize resource scheduling, combined with multiple constraint conditions, Stackelberg equilibrium is solved iteratively using dynamic auction official algorithm.The application converges to global optimal solution quickly through three-dimensional hybrid driving price updating mechanism, realizes system economic benefit maximization, operation robust and reliable, and has good explainability and practical application value.

The invention discloses an event data analysis method and device, a storage medium and electronic equipment. The method comprises the steps that event sequence data generated by a target application program are obtained, a first event type is different from a second event type, a group of causal coefficients is determined according to the event sequence data, and the group of causal coefficients comprises a first event type and a second event type which are different from each other under the condition that an event corresponding to the first event type occurs; the target causal coefficient of the occurrence probability of the event corresponding to the second event type is determined according to the central position of the influence of the first event type on the second event type in time and the diffusion amplitude, and the target causal coefficient responds to the determined first event; and displaying a target event associated with the first event and the target influence coefficient according to the target causal coefficient. The technical problem that the analysis result is not accurate enough due to the fact that the analysis result of the event data is not fine in the prior art is solved.

The invention discloses a 2C railway contact network scene depth algorithm based on deep learning, relates to the technical field of image analysis, can realize highly accurate depth estimation in a 2C scene, and can deduce the distance and depth information of an object by analyzing a single image so as to provide accurate scene perception and distance measurement. This is very critical for applications in 2C scenes, such as security monitoring, environmental awareness and the like, a deep learning technology is used to calculate the distance from the entire image to the camera, and a monocular depth estimation technology can use a deep learning model, such as a convolutional neural network or an auto-encoder, to perform training and inference, and the monocular depth estimation technology can be used to calculate the distance from the entire image to the camera. According to the method, the transverse object distance can be calculated more accurately, and the method is particularly suitable for distant objects.

The invention discloses a patella X-ray image recognition method for lightweight network architecture search under constraint conditions, which comprises the following steps of: 1, collecting marked patella X-ray image data, preprocessing, and dividing into a training set and a verification set; 2, constructing a Shenconvolution neural network population through initialization; 3, training individuals in the population by using an X-ray data set, and using fitness function values based on constraints as fitness values of the individuals; 4, parent individuals are selected according to the fitness values of the individuals in the population, a filial population is generated through filter trimming and crossover variation of a self-adaptive trimming rate, and finally environment selection is carried out; and 5, iteratively optimizing the network structure until a preset condition is met, and extracting an optimal framework as a patella X-ray image recognition model. According to the method, the convolutional neural network architecture adapted to the patella X-ray image data set can be automatically searched, so that accurate classification of the patella X-ray images is realized, especially in application scenes with equipment hardware constraints.

The invention discloses a geographic element gridding representation method suitable for a ship traffic management system, which combines the characteristics of the ship traffic management system to carry out fine grid division on an overwater area, converts geographic elements into information taking geographic grids as basic units through high-performance operation, and adopts a double-layer grid compression storage strategy to store the geographic elements. Therefore, the system can realize real-time high-speed access and processing of geographic information under low resource consumption. According to the method, a double-layer grid compression storage strategy is adopted for the geographic elements, real-time high-performance calling and querying of the geographic elements under the condition of low resource consumption of the VTS system are achieved, and the method has the advantages of being low in resource consumption, high in real-time performance, high in flexibility and the like.

The invention relates to a CoTNet-based micro-expression recognition method under space-time feature driving and application thereof. The method comprises the following steps: extracting an optical flow diagram of a micro-expression sample by using an optical flow method; extracting global features and local features by adopting CoTNet based on a context attention mechanism; based on CASMEII, SAMM and SMIC data sets, a micro-expression recognition model is trained and verified for the preprocessed data through a leaving-one cross validation method, and the micro-expression recognition model is used for recognizing medium-high-risk passengers in civil aviation security. According to the method, the CoTNet network and the optical flow method are combined, the micro-expressions are recognized by analyzing facial movement, key information in a micro-expression sequence is captured more comprehensively, the accuracy and robustness of micro-expression recognition are improved, meanwhile, micro-expression recognition is achieved based on the CoTNet micro-expression recognition method under space-time feature driving, and the method has the advantages of being high in recognition efficiency, high in accuracy and high in robustness. And the real emotional state of the passenger is analyzed, the risk of carrying dangerous goods and making dangerous behaviors of the passenger is judged, and related departments are assisted in identifying medium-and-high-risk passengers.

The invention belongs to the field of mechanical calculation simulation, and particularly relates to a method for predicting shear strength of structural surfaces with different roughness in a three-dimensional stress state. The method comprises the following steps: firstly, preparing a standard rock sample and a silicone rubber sample, splitting the standard rock sample along a splitting line to form a split rock sample with a rough surface, scanning the rough wall surface of the split rock sample, constructing a three-dimensional model, and rendering to generate a two-dimensional image; then, under the three-dimensional stress condition, a shear strength test is carried out, and the shear strength of the structural surface is measured. And finally, associating the two-dimensional image with the shear strength to construct a data set, adopting a GoogLeNet deep convolutional neural network as a reference model, performing parameter adjustment through Bayesian optimization, optimizing a training process in combination with K-fold cross validation and an early stop strategy, evaluating model performance by using a test set, and continuously iterating to improve prediction precision. According to the method, the nonlinear mapping relation between the images of the structural surfaces with different roughness and the shear strength is established, and the mechanical behaviors such as the shear strength of the structural surfaces can be predicted.

The invention relates to the technical field of artificial intelligence big data processing, in particular to a user intelligent matching method and system based on a data collision algorithm. The establishment of the user tag is the basis of intelligent matching, the basic features of the feature information are correspondingly associated with the user tag, the behavior features are specific behaviors of the user, the achievement features are results generated by the specific behaviors, and the behavior features and the achievement features are associated with the user tag through the basic features. By using a data collision algorithm, matching points can be quickly found in large-scale and multi-dimensional data. Compared with a traditional matching algorithm, the data collision algorithm can process a complex data set more efficiently, and the matching precision and efficiency are remarkably improved.

The invention discloses an efficient supervised action segmentation method based on frame-level and action-level double-layer time modeling, which relates to the technical field of computer vision, and comprises the steps of learning a frame relationship by using a convolutional neural network, learning an action dependency relationship by using a converter, and carrying out communication between the two relationships through a cross-self-attention mechanism. A frame branch captures temporal information between frames using multiple CNN networks, an action branch initializes action flags using a multi-head cross-attention and self-attention transformer, using cross-attention to allow frame features to utilize high-level action dependencies learned by the action branch while allowing the action flags to access low-level information in the frame branch, and finally, carrying out final prediction through an average value of the features. According to the method, an action mark set with a fixed size can be efficiently learned to clearly represent an action segment, and actions and frame-level time modeling are executed in parallel, so that the method can be superior to a frame-based and two-stage method at lower calculation cost.

The invention discloses a gas-solid coupling vibration simulation method for a ceramic-based composite material. The gas-solid coupling vibration simulation method comprises the following steps: establishing a multi-scale finite element model of the ceramic-based material; the load of a gas flow field is calculated through a computational fluid mechanics method, and physical quantities between gas-solid interfaces are accurately mapped and transmitted through a radial basis function interpolation function; stress-strain distribution in the solid domain is solved through a calculation structure dynamics method, closed-loop feedback is formed by combining fluid loads, and bidirectional coupling is conducted on the fluid domain and the solid domain; constructing a loading and unloading stress-strain hysteresis behavior model of the ceramic-based material, and calculating vibration response under any stress path; and the displacement of the solid node is fed back to the CFD fluid model, the position of the fluid node is updated, dynamic gas-solid coupling vibration is calculated, and synchronous iteration in the two domains is kept. According to the method, numerical instability caused by sudden change of material rigidity and nonlinear hysteresis behaviors in traditional numerical simulation is overcome, and the method has calculation efficiency and robustness.

The invention relates to the technical field of talent management, and discloses a talent ability evaluation and post matching system based on a knowledge graph, and the system comprises a data processing module which collects individual information and target industry information, converts the individual information into an individual ability vector, and converts the target industry information into an industry demand vector; the graph construction module is used for constructing a cross-industry knowledge graph through multi-source knowledge fusion; the skill migration module is used for calculating skill migration complexity and a capability gap vector; the path planning module is used for generating a plurality of transformation paths and selecting an optimal transformation path based on a transformation path scoring function; according to the method, accurate matching of individual capabilities and post requirements is realized by constructing the cross-industry knowledge graph, and cross-industry skill migration paths can be identified, so that the system can discover potential matching opportunities which are difficult to identify by a traditional method, and the most suitable development paths are recommended for individuals.

This invention relates to a knowledge graph completion method in the field of enterprise credit, belonging to the field of artificial intelligence technology. It includes the following steps: S1: data preprocessing and graph structure construction; S2: construction of an efficient multi-relation graph convolutional neural network; S3: dynamic boundary conditions and loss optimization; S4: multi-constraint optimization; S5: training and prediction. This invention significantly improves the completion accuracy of multi-relation knowledge graphs through innovative methods such as multi-relation modeling, dynamic similarity adjustment, and boundary adaptive optimization, resulting in a complete graph structure. It enables the model to maintain high prediction accuracy even with scarce data and low-frequency relationships. Through feature smoothing and cluster consistency regularization, it possesses temporal and dynamic semantic modeling capabilities. It also has the ability to model the "equity chain—supply chain—risk chain" characteristics of the enterprise credit field.

The invention provides a sparse convolution rule table generation method, an operation method and a tensor processing unit, and provides a sparse convolution rule table generation method for realizing sparse convolution on the tensor processing unit, and the method comprises the steps: calculating the offset of an input coordinate; calculating corresponding output coordinates of the input coordinates in each filter kernel, and calculating the offset of the output coordinates; the offset of the input coordinates and the offset of the output coordinates corresponding to the filter core are compared, a sequence mask corresponding to the filter core is obtained, and elements, not equal to 0, in the sequence mask indicate that the offset of the corresponding set of input coordinates is the same as the offset of the corresponding set of output coordinates; and obtaining a rule table corresponding to the filter core based on the sequence mask corresponding to the filter core and a mask screening function, and performing sparse convolution operation based on the rule table. The number of times of data access is effectively reduced, and the operation efficiency is improved.

The invention discloses a java-based road segmentation and coordinate point matching road segmentation method, which comprises the following steps: S1, data preprocessing: collecting data, storing the data into a database, extracting road coordinate point information from the database, converting the format of the coordinate point information, and storing the coordinate point information into a cache for later use; s2, road segmentation: carrying out segmentation processing on the cached data; the segmented road sections are processed and stored in a database for road matching use; s3, coordinate point matching: taking out the road segment information from the database, carrying out format conversion, then carrying out coordinate point matching on the road segment information, and finally outputting a successfully matched road. According to the method, the matching precision is improved, the data capability is enhanced, the algorithm adaptability is improved, and the system expandability is enhanced.

The invention discloses a computing task scheduling method and device and electronic equipment. The method comprises the following steps: receiving a task processing request initiated by a target object; performing multi-round analysis on task demand information of the target computing task and environment information within a preset time period before the cloud computing platform receives the task processing request by using a multi-target optimization algorithm to obtain an optimal scheduling strategy set, the optimal scheduling strategy set comprises a plurality of optimal scheduling strategies with the lowest calculation cost or the lowest calculation energy consumption or the highest calculation performance; feeding back the optimal scheduling strategy set to the target object; and in response to a strategy selection instruction initiated by a target object, selecting a corresponding optimal scheduling strategy from the optimal scheduling strategy set to schedule the target calculation task. The technical problems that the task response speed is low and the computing resource utilization rate is low due to the fact that the task scheduling strategy cannot be dynamically adjusted according to the real-time state of the network in related technologies are solved.