16results about How to "Reduce double counting" patented technology

The invention discloses a multi-modal disambiguation and knowledge graph embedding-based power grid Internet of Things equipment plug and play method and system, and the method comprises the steps: analyzing an ambiguous user instruction through a multi-modal module, and fusing context and sensor data to generate a clear configuration intention; a knowledge graph containing equipment configuration and topological resources is constructed, after the knowledge graph is converted into a vector through embedding, an optimal access scheme and a physical model are matched through machine learning, meanwhile, power grid operation rule verification is conducted on a matching result, and it is ensured that the scheme can be directly executed. According to the method, the problems of inconsistent object model matching, manual configuration migration, weak ambiguous instruction analysis and the like are solved, automatic access of novel switches, photovoltaic equipment and other equipment is realized, and the automation level of the power grid internet of things is improved.

The invention discloses a method and system for constructing a comprehensive evaluation index of a soil water source conservation function, and relates to the technical field of soil water source conservation, and the method comprises the following steps: measuring a soil water retention performance index and a soil infiltration performance index; based on the measured indexes, determining the weight of each index by adopting a factor analysis method, and carrying out membership standardization processing on each index; and weighting the weight of each index and the standardized membership degree to obtain the comprehensive evaluation index of the soil water conservation function. According to the method, the soil water source conservation function can be quantitatively evaluated on the sample plot scale only through one-time or few-time soil index determination, dependence on a long-term hydrological sequence is avoided, and the data acquisition cost is remarkably reduced.

The invention relates to a content generation method combining a knowledge graph and a large language model. The method comprises the following steps: based on multi-type target content data, extracting an entity, relationship and attribute triple of a multi-source corpus, dividing a theme cluster, constructing a knowledge graph, and carrying out standardized packaging on knowledge units to obtain a standardized knowledge set; analyzing a user input demand, obtaining a multi-dimensional demand feature, performing similarity matching with a theme cluster to obtain a target cluster, screening candidate knowledge through multi-hop retrieval, and obtaining an associated knowledge subset through three-dimensional evaluation index empowerment score sorting; converting the associated knowledge subset into an analyzable structured text, and adjusting and optimizing model generation parameters to obtain initial generation content; and comparing the initial content with a knowledge graph to obtain a result, optimizing text features and performing adaptive adjustment in combination with multi-type target content data and audience features, and obtaining standard content of target quality. Content generation in combination with the knowledge graph and the large language model is realized.

The embodiment of the invention discloses a communication signal multi-parameter intelligent parallel extraction method. The method comprises the following steps: firstly, acquiring a to-be-measured signal and extracting an I / Q component; processing the I / Q components through a multi-scale feature extraction structure to obtain high-dimensional features; extracting shared features from the high-dimensional features through a plurality of residual block groups, inputting a plurality of parallel task branches, and enhancing the shared features in the branches by using a channel attention mechanism to obtain output features; the output features of all task branches are transformed through a learnable linear layer to form a key matrix and a value matrix, the current task features are transformed through a learnable linear layer to form a query matrix, and the input features are segmented into a plurality of heads by means of a multi-head attention mechanism; each head obtains output of a single head based on zoom dot product self-attention operation, and results of all the heads are integrated to obtain enhanced features of the current task; and determining a communication signal parameter prediction label according to the enhanced features, and realizing multi-parameter efficient parallel extraction.

The application provides a vulnerability repair method and system, and relates to the technical field of network security vulnerability verification and enterprise vulnerability management. The application obtains a task containing a vulnerability identifier, asset information and to-be-verified repair content, processes the task in parallel according to a pipeline split according to business execution steps, and configures a resource pool matching external dependencies for each stage; a stage fingerprint is generated when the task is transferred, a result is reused when a subtask is executed, otherwise the execution is scheduled according to vulnerability harm, asset importance and threat intelligence, and the vulnerability repair is completed according to the verification result. The application solves the problems of low throughput of the existing general pipeline, slow response of high-priority tasks, and easy starvation of low-priority tasks, and achieves the effect of collaborative optimization of throughput improvement, emergency task response acceleration and low-priority task starvation prevention without changing the original business and hardware.

The invention discloses a voice text bidirectional conversion system and method based on asynchronous processing and a message queue. The system comprises a server, an OBS object storage service, a business system, a conversion gateway and a plurality of universal gateways. The conversion gateway is connected with the server, a voice-to-text module and a text-to-voice module are configured in the conversion gateway, and the voice-to-text module is constructed based on a FunASR streaming recognition framework, performs communication through a WebSocket protocol, and performs a voice recognition process in combination with an asynchronous task processing mechanism; the text-to-voice module is constructed based on ChatTTS service, and realizes an asynchronous task distribution and result return mechanism in combination with a RabbitMQ message queue. According to the scheme, efficient conversion between the voice and the text is realized through unified interface design and modular processing flow, and the response speed, the accuracy and the expandability of the voice processing system are improved.

The invention discloses a batch data processing method and device suitable for a database, equipment and a medium, relates to the technical field of data processing, and is applied to a processor. Monitoring whether a target instruction meeting a preset stride condition exists in a data processing instruction stream corresponding to the preset database in the target assembly line or not, if yes, entering an instruction pre-execution mode, vectorizing the target instruction and a subsequent instruction having a dependency relationship with a register of the target instruction, and outputting the vectorized instruction; performing instruction pre-execution based on the vectorized instruction and a first execution unit, and storing an instruction pre-execution result to a first physical register file and a result cache region; and when the data processing instruction stream in the target assembly line reaches the input end of the second execution unit, matching the instruction pre-execution result in the result cache region, and completing batch data processing operation based on an information matching result. According to the invention, the efficiency and accuracy of batch data processing are improved.

The invention provides a font recognition method, device and equipment based on self-adaptive preprocessing and a medium, and the method comprises the steps: carrying out the self-adaptive preprocessing of a collected font image, the preprocessing comprises dynamic noise suppression, illumination unevenness correction and sub-pixel-level geometric correction, and obtaining a corrected character image; performing multi-dimensional feature extraction on the corrected character image, and constructing a fusion feature vector including local detail features, global style features and geometric correction features; and performing dynamic weight matching based on the feature contribution degree evaluation model, calculating the similarity between the to-be-matched character and the candidate fonts, and outputting the recognized font type according to the similarity result, thereby ensuring the font recognition accuracy.

The application relates to the technical field of data processing, in particular to a code reordering processing method and system. A plurality of candidate codes are generated based on requirements; a plurality of indexes are set for each candidate code; each index is tested based on an execution channel to generate an execution result vector of each candidate code; each index is tested for deviation to set a deviation value of each index; an equivalent cluster of each candidate code is constructed; an execution score of each candidate code is calculated based on the deviation value and the execution result vector; each candidate code is tested based on an inference channel to output an inference score of each candidate code; a mixed score is calculated, and each candidate code is reordered based on the mixed score and the equivalent cluster. The application overcomes the inference illusion problem of a large language model when generating codes through a double-channel scoring system, and improves the quality of code generation.

Embodiments of this disclosure relate to a method, apparatus, and storage medium for video processing. The method includes: reconstructing a first video frame in a video by decoding a bitstream, the first video frame corresponding to a first resolution; providing the first video frame to a super-resolution model to determine a first intermediate feature corresponding to a first processing stage of the super-resolution model; determining a feature reuse mode corresponding to the first video frame; in response to a feature reuse mode indication, performing super-resolution processing using cached feature information, acquiring a second intermediate feature corresponding to a second video frame without performing a second processing stage of the super-resolution model; determining a fusion feature based on the first and second intermediate features; and generating a third video frame corresponding to the first video frame based on the fusion feature, the second resolution of the third video frame being higher than the first resolution. In this manner, embodiments of this disclosure can improve video processing efficiency while maintaining image quality.

The invention provides a sequential circuit hybrid fault simulation method, which is used for simulating a sequential circuit fault coverage rate by injecting a fault point into a sequential circuit, and comprises the following steps: S1, obtaining a to-be-simulated circuit netlist, identifying a strong connection component in the to-be-simulated circuit netlist, and marking the strong connection component meeting a preset small-scale strong connection loop constraint condition; s2, performing simulation of multiple periods on the marked circuit netlist to be simulated, and executing in each simulation period: S21, performing code compilation on the marked strong communication components so as to fuse each marked strong communication component into a compilation node, performing compilation type simulation on each compilation node until the state converges, and performing simulation on the compilation node until the state converges; the topology is updated, and the output of each compiling node is used as an event to be transmitted to other areas in the circuit netlist; and S22, taking other areas except the compiling node as event driving nodes, simulating residual circuits in an event scheduling mode, and updating the state of the circuits.

The application provides a blind detection method and user equipment for a 5G wireless communication system, which comprises the following steps: S1, acquiring configuration information of a control resource set and configuration information of a corresponding search space, and determining a candidate PDCCH set composed of multiple candidate PDCCHs according to the configuration information of the control resource set and the configuration information of the corresponding search space, wherein each candidate PDCCH contains at least one control channel element; S2, performing centralized symbol level processing on the control resource set according to the configuration information of the control resource set, so as to extract soft bit information of subcarrier positions related to the candidate PDCCH set at one time, and form a soft bit information library required for current blind detection; and S3, selecting a candidate PDCCH from the candidate PDCCH set by using a pre-set selection rule to perform blind detection, wherein blind detection is performed according to the frequency domain resource block corresponding to the control channel element contained in the candidate PDCCH, and the soft bit information of the subcarrier positions of the corresponding frequency domain resource block is obtained from the soft bit information library to perform blind detection.

The embodiment of the invention provides an image classification method and device, electronic equipment and a storage medium, and belongs to the technical field of image classification. The method comprises the following steps: determining single-mode image data as target image data, and calling a preset image classification model to perform high-frequency feature fusion on image high-frequency features of the target image data and a plurality of single-mode image data to obtain image high-frequency fusion features; calling a preset image classification model to perform low-frequency feature fusion on the image low-frequency features of the target image data and the multiple pieces of single-mode image data to obtain image low-frequency fusion features; calling a preset image classification model to perform multi-modal feature fusion on the image high-frequency fusion features and the image low-frequency fusion features of the plurality of single-modal image data to obtain multi-modal image fusion data; and calling a preset image classification model to perform image classification on the multi-modal image fusion data to obtain image classification result information. According to the embodiment of the invention, the accuracy of classifying the objects in the image can be improved.

The application discloses a kind of theoretical spectral library library construction method, measurement method and its electronic equipment, medium, program product. Among them, the theoretical spectral library library construction method includes: the geometry modeling of a semiconductor device is carried out, the structure parameters for describing the geometry model and the floating range of the structure parameters are set;The geometry model is divided into multiple main modules, and the structure parameter set of each main module is selected for each main module;For each floating value of each structure parameter in the structure parameter set of the main module, define the module structure set corresponding to each main module to construct a main module structure library;According to the condition corresponding to OCD measurement, the scattering matrix / transmission matrix of each module structure in the main module structure library is calculated respectively, and is saved one by one with the main module, to construct main module database.The application greatly reduces the calculation time by reducing repeated calculation.

The application discloses a kind of external key point matching and tracking trajectory calculation method and device based on VGGT.The method comprises: loading configuration file and initializing model;Load external key point coordinates and carry out standardization and quantity alignment;First, input image is converted into tensor format, then image features are extracted, and finally L2 normalization is carried out;Depth feature vector of external key point is extracted from feature map;Calculate cosine similarity to evaluate matching quality;The average amplitude and standard deviation of feature vector are calculated to evaluate feature quality, and the matching threshold is dynamically adjusted;The coordinate difference of matching point is calculated to eliminate the matching that does not meet geometric constraint, and tracking trajectory is generated;Tracking quality analysis and visualization are carried out.The matching success rate of the application is greatly improved, the false matching rate is obviously reduced, the time consumption of processing image is obviously shortened, the efficiency is also greatly improved, the tracking stability is obviously improved under dynamic interference, and automatic high-precision inspection without manual intervention is realized.