545 results about "Feature generation" patented technology

This invention discloses an intelligent resource allocation method for energy-aware partitions, belonging to the field of computer architecture and operating system task scheduling technology. The method divides computing resources into multiple independent energy-aware partitions based on the processor cores. It then monitors the energy status parameters of each partition in real time. By parsing the instruction code stream of the tasks to be assigned, the proportion of arithmetic and logical instructions is extracted as instruction mixing features, and the address span of memory access instructions is analyzed to obtain cache access density features. Based on the energy status of the partitions and the computation and memory access characteristics of the tasks, a suitability score for the task relative to each partition is generated, comprehensively reflecting the energy matching degree and hardware resource matching degree. Based on the suitability score and the remaining task acceptance capacity of the partitions, a two-dimensional assignment strategy is used to dynamically map tasks to the core execution in the optimal energy-aware partition. This invention achieves fine-grained matching of computing resources and energy status, optimizing system energy utilization efficiency.

The disclosure provides a bird's eye view feature generation method based on multi-modal fusion, comprising: extracting all vehicle-mounted sensor collection information of the current time of the vehicle to obtain at least first modal collection information and second modal collection information; using a first modal feature extraction network to extract first modal features of the first modal collection information, and using a second modal feature extraction network to extract second modal features of the second modal collection information; obtaining the bird's eye view feature after initialization of the current time based on the historical time bird's eye view feature; obtaining the bird's eye view feature with first modal feature information based on the first modal features and the bird's eye view feature after initialization of the current time; obtaining the bird's eye view feature with second modal feature information based on the bird's eye view feature with first modal feature information and the second modal features; and generating the bird's eye view feature with multi-modal fusion information. The disclosure also provides an electronic device, a readable storage medium and a program product.

This invention provides an intelligent operation and maintenance system for prefabricated substations based on big data, belonging to the field of prefabricated substation operation and maintenance technology. The data access unit unifies the time alignment of electrical operation, environmental monitoring, and historical operation and maintenance data, constructing a constrained time-series data structure based on equipment topology; the state evolution modeling unit extracts multi-scale features, generating three sets of evolution parameters: impedance, capacity, and load; the control variable reconstruction unit generates enhanced control input vectors through component mapping and iterative feedback; the real-time control execution unit outputs reactive power regulation and voltage control commands accordingly; and the human-machine feedback unit embeds manually corrected information and updates control parameters. This invention solves the problems of lack of topological constraints in multi-source data and the difficulty of embedding human experience into automated control, improving the intelligence level, control accuracy, and adaptability of prefabricated substation operation and maintenance.

The application provides a semantic extraction method based on model computing power dynamic adjustment, a model and computer equipment. The method comprises the following steps: calling a plurality of feature extraction layers to perform multiple feature extractions on initial semantic features to obtain predicted semantic features; judging the matching degree of the predicted semantic features and preset expected semantic features; if the matching degree of the predicted semantic features and the preset expected semantic features meets a preset matching threshold, generating an output result of a language model through the predicted semantic features; and if the matching degree of the predicted semantic features and the preset expected semantic features does not meet the preset matching threshold, taking the predicted semantic features as new initial semantic features, re-calling the plurality of feature extraction layers to perform multiple feature extractions on the new initial semantic features until the output result of the language model is generated. The embodiments of the application can allocate different computing resources for different tasks, so that the purpose of reasonably using computing resources is achieved.

The application relates to the technical field of artificial intelligence, in particular to a multi-modal human-computer interaction content filtering method, device and medium. The method comprises the following steps: collecting multi-modal interaction data; obtaining a current application scene, calculating single invalidity scores of various data, obtaining joint invalidity scores in combination with multi-modal weights corresponding to the scene, and judging whether the joint invalidity scores are lower than a preset threshold. If yes, scene-specific violation judgment weights are called, multi-modal violation features are extracted from the data, and current fusion features are obtained by combining the weights; the latest N (N>1) historical fusion features are obtained, the current fusion features are combined to generate modified fusion features, the modified fusion features are input into a pre-trained deep learning model, and a risk confidence is obtained. When the risk confidence exceeds a preset threshold, an interaction instruction transmission link is immediately cut off, and multi-modal data is encrypted and stored as evidence, so that accurate filtering and high-risk prevention and control of multi-modal interaction content are realized. The application can effectively improve the judgment accuracy.

This invention relates to the fields of computer vision and road infrastructure inspection, specifically to a method for detecting road surface cracks based on edge perception and global receptive field modeling. The steps include: inputting an image into an encoder-decoder network for multi-scale feature extraction; inputting feature maps of at least one level into a biomimetic edge perception module for edge feature enhancement; inputting feature maps of other levels into a global receptive field module for enhancement processing; and upsampling the enhanced features level by level through the decoder, fusing the previous level's decoded features with the enhanced features output by the global receptive field module to generate a road crack segmentation map. By introducing the concept of biomimetic visual edge perception, the model's ability to perceive road cracks is enhanced. Combined with a global receptive field modeling strategy, it achieves joint modeling of the relationship between local details and the overall spatial structure of cracks, thereby improving the detection accuracy and stability under complex road surface background conditions.

The application provides a construction scheme intelligent generation method and system based on multi-agent cooperation, and relates to the field of building engineering digitization technology. The method comprises the following steps: constructing a project knowledge base containing vector index and knowledge graph index; retrieving similar historical schemes based on project characteristics and learning chapter structure characteristics; generating a construction scheme outline; performing knowledge-enhanced retrieval on the chapter to be generated to obtain structured technical data; using a double-channel generation mechanism to process general chapters and construction method chapters respectively; receiving natural language fine-tuning instructions to complete content correction; converting the generated content into a standard Word document and applying an enterprise template. The system comprises a knowledge base construction module, a template retrieval module, an outline generation module, a knowledge retrieval module, a multi-agent cooperation module, a full-text integration module, and a document output module. The application significantly improves the efficiency and technical accuracy of construction scheme preparation through multi-agent division of labor, knowledge-enhanced retrieval, and double-channel generation, and solves the problems of low efficiency, case residue, and insufficient targeting in traditional scheme preparation.

This application provides a video processing method, apparatus, computer device, and computer-readable storage medium to improve matching accuracy and robustness in various complex scenarios. The method includes: acquiring a video frame to be processed, comprising a left-eye image and a right-eye image, the video frame being included in the video to be processed; performing feature extraction and stitching on the left-eye image to obtain a first target feature, the first target feature being obtained by stitching together the depth features and local features of the left-eye image; performing feature extraction and stitching on the right-eye image to obtain a second target feature, the second target feature being obtained by stitching together the depth features and local features of the right-eye image; generating a cost matrix based on the first target feature and the second target feature; and generating a first disparity map of the video frame to be processed based on the cost matrix.

A material search computing system generates a joint feature comparison space by combining joint image-text features of surface material data objects. The joint feature comparison space is a consistent comparison space. The material search computing system extracts a query joint feature set from a query data object that includes text data or image data. In addition, the material search computing system compares the query joint feature set to the joint image-text features included in the joint feature comparison space. Based on the comparison, the material search computing system identifies a result joint feature set and associated result surface material data objects. The material search computing system generates material query result data describing the result surface material data objects, and provides the material query result data to an additional computing system.

This invention relates to a visual sequence compression method for multi-turn visual question answering in medical images, comprising: extracting an original visual token sequence based on a target medical image; generating a compression matrix using a pre-trained meta-compression module based on the image features of the original visual token sequence; applying the compression matrix to the original visual token sequence to obtain the compressed visual token sequence; concatenating the compressed visual token sequence with the language token sequence of the current question-answering turn, and inputting the concatenation into a large-scale visual language model to generate the question-answering response for the current turn. The beneficial effects of this invention are: by dynamically generating a compression matrix based on the features of the medical image itself, this invention can compress the length of the visual sequence while still retaining key image features that are low-attention but related to lesion identification, thereby overcoming the shortcomings of existing technologies that easily overlook small lesions and weak-contrast abnormal areas.

This application relates to the field of industrial data processing and intelligent control technology, and discloses a method for dynamic modeling and prediction of industrial big data combined with digital twins. The method includes: acquiring operating status data, process execution data, control command records and prediction results, and constructing a sequence of control action events; segmenting the operating status data according to the sequence of control action events to form natural operating data segments and intervention operating data segments, extracting distribution characteristics, fluctuation characteristics and response characteristics, and generating a segmented behavioral feature set; calculating the control reflexive influence quantity and generating a reflexive drift marker, dividing the natural evolution dataset and the control feedback dataset, and predicting them separately; generating control risk constraint quantity, screening, delaying or limiting the control commands to be executed, and iteratively correcting the dataset according to the execution results; this method can distinguish between control feedback changes and natural evolution changes, reduce drift misjudgments, and suppress multi-round oscillations and parameter fluctuations.

The application discloses a current transformer magnetic characteristic test system and method and a computer program product, wherein the system comprises an original data acquisition module, a data feature generation module, a quality index evaluation module, a high load performance test module and a threshold correction determination module; core parameters are obtained by dynamically adjusting a sampling strategy, a feature vector is generated after preprocessing, a multilayer convolutional neural network model is used to mine parameter nonlinear correlation and output a quality index; a quality evaluation correction value is generated by combining temperature rise and noise data under a high load condition, an evaluation threshold is dynamically corrected, and test result determination is realized. The application effectively solves the problems of high misjudgment rate of static threshold determination in the prior art, difficulty in capturing parameter nonlinear correlation and inability to predict long-term stability, significantly improves the precision and reliability of magnetic characteristic test, accurately predicts the long-term full-load operation stability of equipment, and provides a reliable basis for power system core equipment selection and safe operation.

This application discloses a method, device, and storage medium for global fluctuation analysis of text. The method, relating to the field of text parsing technology, includes: processing the input text using a first model to obtain a first parsing result, and processing the input text using a second model to obtain a second parsing result; determining the sentence-level raw evaluation value of each sentence in the input text; determining the paragraph-level raw evaluation value of each paragraph in the input text based on the first parsing result, the second parsing result, and the sentence-level raw evaluation value; connecting the paragraph-level raw evaluation values ​​of each paragraph into an expression intensity curve using the paragraph sequence of the input text as a time axis; and identifying and labeling four geomorphic features—peak, valley, data slope, and data plain—based on the expression intensity curve to generate a visualized scoring geomorphic map. This application achieves accurate prediction and hierarchical selection of training corpus quality before model training.

The application relates to an optimization design method of a polygon-based multi-plane device, comprising the following steps: S1. determining parameters of a polygon-based MPLC device; S2. constructing a topological-mode coupling model to determine a phase compensation amount; S3. generating an initial phase mask according to the phase compensation amount calculated in step S2 and in combination with phase distribution characteristics of a target mode; S4. outputting optimized phase mask parameters based on a genetic optimization algorithm; and S5. simulation verification: building simulation models of different topological structures, inputting the optimized phase mask parameters, simulating a mode evolution process, and analyzing whether a performance threshold value meets a demand. The application solves the problems that a traditional algorithm ignores the influence of a topological structure and the generality of a phase mask is poor, realizes accurate design of different polygon devices, and thus effectively reduces loss and improves shaping effect.

The application relates to a CAD vector data intelligent preprocessing method based on a multiple recognition mechanism, which comprises the following steps: parsing a DXF file to construct a CAD heterogeneous information graph; extracting text features for a layer node and generating a layer embedding vector through adaptive fusion of a gate unit by aggregating high-order semantic structure features through a meta path; introducing a dynamic threshold decision mechanism to generate a dynamic threshold according to global statistical features of a drawing to determine a background layer; constructing a global dependency graph containing cleaned dependent edges and semantic constraint edges, dividing the global dependency graph into weakly coupled connected subgraphs as independent optimization transactions by using a graph cut algorithm, and performing parallel execution to complete cross-dimension collaborative optimization; setting a multi-level logical checkpoint and generating an atomic operation log, and realizing tracing and accurate recovery based on an incremental storage mechanism. The application has the effects of significantly improving background layer recognition accuracy and scene adaptability, improving collaborative optimization efficiency, and guaranteeing operation safety and traceability.

Embodiments of the present disclosure perform training attribution by identifying a synthesized image and a training image, where the synthesized image was generated by an image generation model that was trained with the training image. A machine learning model computes first attribution features for the synthesized image using a first mapping layer and second attribution features for the training image using a second mapping layer that is different from the first mapping layer. Then, an attribution score is generated based on the first attribution features and the second attribution features, where the attribution score indicates a degree of influence for the training image on generating the synthesized image.

This invention relates to the field of air pollution monitoring technology and discloses an air pollution discrimination method and device based on machine learning. The method comprises the following steps: inputting pollution source attribute information and monitoring data from the first and second time periods into a pollution feature model to obtain pollution feature vectors; determining candidate pollution sources similar to the baseline pollution source type based on vector correlation; inputting the feature vectors of the candidate pollution sources and historical periodic monitoring data into a discrimination model trained with the baseline pollution source as a positive sample to obtain the pollution type discrimination result. The pollution feature model extracts multi-dimensional spatial distribution vectors from a spatiotemporal feature model to initialize the network layer and updates parameters dimensionally. The discrimination model combines dynamic monitoring data and feature vectors to achieve discrimination. The device includes units for feature generation, correlation analysis, and type discrimination. This solution improves the accuracy and efficiency of air pollution discrimination through multi-dimensional data fusion and model optimization.

A machine learning based (ML-based) method and system for forecasting cash flow, is disclosed. Initially, the data associated with business units are obtained from data sources. The data are pre-processed to generate pre-processed data. Features associated with financial information are determined for horizons based on the pre-processed data using a plurality of input AI models. Feature combinations are generated by integrating the features associated with the financial information, for each horizon of the horizons. Forecasts are generated for each horizon for a pre-determined time interval, using a stacked AI model including forecasting models. The generated forecasts for the cash flow of the business units, are provided as an output, to the users on user interfaces associated with electronic devices associated with the users.

This invention provides a method and apparatus for generating vocabulary memorization content, relating to the field of data processing technology. The method includes: acquiring multidimensional feature data of a target word to be memorized and a target user; wherein the multidimensional feature data includes at least the target user's personalized image features and learning stage identifier; constructing a target scene corpus based on the semantic information of the target word to be memorized; constructing structured prompt information for driving a generative model based on the target word to be memorized, the target scene corpus, and the multidimensional feature data; inputting the structured prompt information into a pre-trained generative artificial intelligence model to generate a dynamic mnemonic comic, and displaying the dynamic mnemonic comic on a terminal device; wherein the dynamic mnemonic comic includes a character image generated based on the personalized image features, and the character image interprets the meaning of the target word to be memorized within the context defined by the target scene corpus.

Embodiments relate to training and applying a parallel-structured neural network. In some embodiments, a dataset of a plurality of claims is received, where each claim of the plurality of claims includes multiple fields. Features, including numerical features and a plurality of categorical features, are extracted from the plurality of claims and based on the multiple fields. The parallel-structured neural network is trained by (1) selecting, based on a feature selection algorithm and a loss function, one or more numerical features and one or more categorical features, and (2) generating, based on the selected features, a categorical layer, a fully-connected layer, and a concatenation layer of the parallel-structured neural network. A claim is received including numerical data and categorical data, and a probability prediction is generated by inputting the claim into the parallel-structured neural network. Finally, a submission including the first claim is generated based on the probability prediction.

The application discloses a feature extraction workflow execution method and system embedded in a database, and the method comprises the following steps: a feature extraction workflow is defined in an extended SQL syntax; the feature extraction workflow is parsed, structured executable units are extracted as nodes, a topological relationship between the nodes is constructed, and a directed acyclic graph is generated; wherein the executable units comprise data source nodes, calculation / conversion nodes and storage nodes; the nodes in the directed acyclic graph are mapped into feature extraction tasks, an execution plan is generated; according to the execution plan, a task scheduling queue is determined, and corresponding feature extraction tasks are executed according to the task scheduling queue; features extracted by executing the feature extraction tasks are generated into feature vectors, an association between the feature vectors and original data is established, a feature table with vector indexes is formed and stored.

The application provides a judicially disadvantaged group personnel identification method and device and terminal equipment, which are suitable for the technical field of data processing, and the method comprises the following steps: generating personnel identity feature information and personnel socio-economic feature information according to personnel identity information, personnel physiological age information, personnel occupation information, personnel education information, personnel family burden status information, a personnel identity feature generation model and a personnel socio-economic feature generation model; generating initial disadvantaged group personnel identification information according to the personnel identity feature information, the personnel socio-economic feature information, a disadvantaged group personnel feature benchmark value and an initial disadvantaged group personnel identification model; and generating target disadvantaged group personnel identification information according to the initial disadvantaged group personnel identification information, historical disadvantaged group personnel identification information, disadvantaged group division threshold information and a target disadvantaged group personnel identification model. The application is used for intelligent identification of disadvantaged groups, thereby effectively improving the accuracy of identity identification in judicial identification.

This invention relates to the field of intelligent visual inspection technology, specifically to a visual recognition system for the wear state of cutting tools on machined parts. The system includes a liquid film interference suppression module, a multi-scale tool segmentation module, a wear feature generation module, a machining parameter acquisition module, an adaptive decision-making module, and a visualization output module. Specifically: the liquid film interference suppression module generates a liquid film interference-suppressed image; the multi-scale tool segmentation module outputs a tool body mask; the wear feature generation module calculates the cutting edge defect rate and surface mottledness, and generates a tool wear index; the machining parameter acquisition module acquires machine tool machining parameters; and the adaptive decision-making module outputs the wear level through a working condition matching decision tree. This invention, by fusing image feature extraction and working condition parameter determination, achieves accurate identification and visualization output of tool wear state, improving the efficiency of intelligent monitoring and decision-making in complex machining scenarios.

Embodiments of the disclosure provide for improved Bluetooth communication security in vehicles. In the context of a method, the method includes obtaining, from a computing device, at least one feature of a communication specification for a vehicle, wherein: the communication specification defines data sharing protocols between a sensor or computing device aboard the vehicle and a vehicle management system to enable wireless communication by a Bluetooth communication mode without disabling means of the sensor, computing device, or vehicle management system for communication by a second mode. The method includes generating a vulnerability of the communication specification based on the at least one feature; determining a mitigation operation based on the vulnerability; and providing a recommendation comprising the mitigation operation to the at least one computing device to enable modification of the communication specification toward reducing the vulnerability.

The disclosed small sample radar HRRP recognition method based on feature generation comprises the following steps: obtaining the signal of the recognition target and pre-processing; inputting the pre-processed HRRP data, obtaining mixed features through a feature extraction module, sending the mixed features into a feature decoupling module, and decoupling into identity features and angle features; the identity features pass through an identity recognition module to obtain a recognition result and are saved in an identity category fingerprint library; the angle features pass through an angle decoding module to calculate the angle values thereof; inputting an expected angle, entering an angle encoding module to obtain an expected angle feature, extracting the identity features from the identity fingerprint library, splicing the identity features with the expected angle feature, inputting into a feature generation module, and generating mixed features under the expected angle; building an overall network model, training the network using a training sample set, testing the network effect using a test sample, and obtaining a recognition accuracy.

The invention provides a tumor medical mode recognition method and system based on image detection, and relates to medical image processing and artificial intelligence auxiliary diagnosis, and the method comprises the steps: carrying out the partitioning of a digital pathological section image, evaluating the color distribution, light and shade contrast and microstructure geometric features of each block, and generating a local quality evaluation graph to recognize a quality damaged region; a local diagnosis judgment graph is generated based on tumor recognition model analysis, and the diagnosis confidence of the model on each region is visually reflected; performing correlation analysis on a low-confidence area and a quality damaged area in the two images, if the overlapping degree reaches the standard, generating quality correlation interpretation, and determining that uncertainty is derived from image quality; finally, an interpretability report containing the original image, the two evaluation images and interpretation is generated, a clear basis is provided for doctors, traditional compensation defects are avoided, recognition reliability and transparency are improved, precise clinical decision making is assisted, and practical value is achieved.

This invention discloses a container shipping route analysis method based on artificial intelligence analysis, belonging to the field of shipping intelligence technology. The method includes: collecting route operation data during the operation and management of a target container shipping route, performing time alignment and normalization processing to generate an actual route event sequence; embedding route governance rules into corresponding route segments based on the actual route event sequence to construct a route governance semantic representation; outputting the arrival time series of each port of call and the reliability evolution characteristics as the route progresses through an artificial intelligence analysis process based on the route governance semantic representation; and generating a counterfactual route trajectory sequence using the arrival time series and reliability evolution characteristics. This invention reconstructs the theoretical evolution process of a shipping route under given operational constraints by constructing a counterfactual route trajectory with route governance semantics, providing a predictable and alignable expressive basis for the operational state of the route in both time and space dimensions.

The application relates to an adaptive multi-channel remote gesture interaction method and system, which comprises the following steps: generating a working mode identifier according to a sensor available state; collecting sensor original data streams and solving to generate standardized gesture vectors; inputting the vectors into a trajectory prediction model trained based on historical data to generate prediction trajectory data; combining current channel bandwidth characteristics and the prediction data to perform frame rate adaptation and motion smoothing processing on the gesture vectors to generate optimized instruction data packets; acquiring a user operation event, performing hierarchical local feedback while sending the instruction data packets; when it is monitored that a network packet loss rate exceeds a threshold value, generating a session state snapshot and synchronizing the snapshot to a backup channel to complete a hot switching; analyzing instructions and snapshot rendering cursor trajectories at a remote end, and generating an updated session state snapshot based on trajectory behavior characteristics; and storing the snapshot into a trajectory tracking database for subsequent state recovery. The application can improve user remote interaction experience in a complex scene.

The present application relates to the field of instruments and process measurement technology, in particular to a multi-range pressure gauge adaptive switching system and method based on time sequence feature perception, the system comprising a data acquisition module that generates a smooth pressure time sequence signal in real time; a time sequence feature perception module that receives the smooth pressure time sequence signal and generates a pressure prediction trend signal for a future short time window based on pressure change trend characteristics; an adaptive decision module that compares the pressure prediction trend signal with a preset current range threshold range; a switching execution module that responds to a pre-switching trigger signal when a switching lock signal is in an unlocked state and outputs to a multi-range sensor group to complete switching of a working channel; and a data fusion module that performs weighted fusion processing on two output signals according to a switching process state signal provided by the switching execution module. The present application can solve the problem of switching lag in existing pressure measurement systems when facing pressure transient working conditions.