26 results about "Local algorithm" patented technology

The application relates to a ship plate detection positioning method, a storage medium and electronic equipment, which comprises the following steps: labeling a ship picture shot by a port wharf, establishing a ship plate detection positioning model, updating the ship plate detection positioning model through a non-local neighborhood calculation alignment algorithm and a channel fusion enhancement algorithm; using the above data set, training and adjusting the updated ship plate detection positioning model to obtain a trained ship plate detection positioning model; shooting a monitoring video stream through a camera arranged at the port wharf, inputting the monitoring video stream into the trained ship plate detection positioning model, detecting a ship plate on the port wharf, and calculating a positioning coordinate of the ship plate appearing in the video stream. Differing from the prior art, the application performs non-local alignment on multi-level features in a feature pyramid through a multi-scale non-local algorithm, fuses channel features between feature maps before and after alignment, and further improves the detection precision of the algorithm.

The invention discloses a technical algorithm for estimating the SOC of a lithium battery in real time, which relates to the technical field of lithium batteries and comprises a hardware acquisition and control module, a cloud service platform module and a core algorithm module. According to the technical algorithm for estimating the SOC of the lithium battery in real time, the MIAUKF of the core algorithm module is responsible for accurate estimation of the SOC, and the I-EKF is responsible for dynamic updating of the SOH and closed-loop linkage of data, so that the problem of estimation deviation of a single algorithm is solved; through self-adaptive correction of Q and R, battery characteristic changes under different working conditions are adapted; the local algorithm ensures real-time performance, the cloud platform provides computing power support and historical data optimization, and both speed and precision are considered; therefore, the terminal voltage and the cell temperature of the single lithium battery in the lithium battery pack can be monitored in real time with high precision, the common hidden dangers of the lithium battery such as overcharge, overdischarge and overhigh temperature rise of each single lithium battery are effectively prevented, the service life of the battery is prolonged, the use economy of the battery is improved, and the energy utilization rate of the battery pack is improved.

The invention discloses a fault diagnosis method based on integrated empirical mode decomposition and a manifold structure, and aims to research an algorithm model capable of realizing effective fault diagnosis for an early fault with weak characteristics. The main core of the method is to integrate eigenmode function components obtained by empirical mode decomposition, judge the sensitivity of the eigenmode function components to early faults so as to provide a variable reconstruction strategy more sensitive to the early faults, and meanwhile, extract local features and manifold structures by using a neighborhood preserving embedding algorithm so as to improve the robustness of the early faults. And high-order statistical features more sensitive to early faults are constructed in combination with a statistical local algorithm, so that the high-order statistical features are input into a Bayesian classifier, and finally early fault diagnosis is realized. Compared with a traditional method, the method can more effectively distinguish different types of early faults, obtains higher accuracy, and is a more excellent early fault diagnosis method.

The application relates to the technical field of path planning, in particular to a path planning method based on target characteristics; by adopting a window shrinkage and a strengthened path cost for a large obstacle, motion instability caused by steering mutation is effectively inhibited; a small obstacle is relaxed in constraint to improve local obstacle avoidance flexibility. Newly added brake safety verification ensures no sudden stop risk under sudden obstacles, and experimental data show that the scheme synchronously optimizes path length, task time, path smoothness and safety in a complex scene, meets the core needs of efficient and smooth operation of a commercial scrubber, adopts a shrinkage strategy for a large obstacle, relaxes constraint and reduces path penalty weight for a small obstacle, and allows more aggressive small-range maneuvering, the mechanism significantly improves scene adaptability; the technical problems of non-smooth trajectory, path detour and insufficient brake safety when a scrubber adopts a traditional DWA as a local algorithm for path planning are solved.

The embodiment of the invention discloses a big data processing method and system based on cloud computing, and the method is realized based on a cloud computing platform, and the method comprises the steps: decomposing a big data processing task into a plurality of micro-tasks, and distributing the micro-tasks to distributed computing nodes; each computing node selects an adjacent node based on a hyper-local algorithm, and constructs a local task processing model only by using local data and metadata of the adjacent node dynamically selected based on a fast game mechanism, the metadata including micro-task data semantic features and task dependency relationships of the adjacent node, and the task dependency relationships of the adjacent node are calculated; semantic features are extracted through a natural language processing technology, and a task dependency relationship is determined based on input and output association of the microtasks; according to the local task processing model, independently executing the micro-task on each computing node, and dynamically adjusting the distribution of the micro-task through a point-to-point local negotiation mechanism; and aggregating the micro-task processing results of the computing nodes to generate a global data processing result. According to the method, the operation and communication overhead is reduced.

This invention discloses a method for calculating the bearing capacity of shallow foundations by integrating machine learning and nonlocal algorithms, belonging to geotechnical engineering. 1. The type of foundation soil is determined through indoor triaxial tests; strain-hardening foundation soils are analyzed; 2. Strain-softening foundation soils are analyzed; 3. A finite element model is established for refined simulation, followed by 7. 3. A constitutive model combined with a nonlocal algorithm is used for refined simulation, then proceeds to 4. 4. The influence range (DL) of the shallow foundation analysis is calibrated. f DL is determined by the softening rate control parameter β. f If suitable, proceed to step 7; otherwise, proceed to step 6. Step 6 involves building and training a machine learning agent model to ultimately obtain a suitable deep learning (DL) algorithm. f Shallow foundation analysis was performed with β1 and 7 to obtain the predicted value of the foundation bearing capacity. The present invention uses the above method to achieve efficient inversion of nonlocal parameters, thereby obtaining a reasonable predicted value of the shallow foundation bearing capacity.

The invention discloses a variable-scale satellite group orbit planning method based on a graph near-end optimization algorithm. Real-time and efficient control under the conditions that the number of nodes dynamically changes and communication links are limited is achieved. Firstly, the orbit and communication relation of a satellite group on time is abstracted into a dynamic graph; then, a reinforcement learning structure based on superposition of the graph neural network and the recurrent neural network is constructed; a strategy and value function is asynchronously iterated on a satellite by adopting a reinforcement learning near-end strategy optimization algorithm, all calculation and parameter updating are completed at a satellite end, and a central master control satellite is not needed. After local algorithm updating is completed, each satellite exchanges parameter differences with the neighbor satellites meeting the reliability condition, and weighting is carried out according to the satellite link quality and physical distance mixed weight. The method can be widely applied to application scenes needing constellation-level collaboration such as earth imaging, global communication and navigation, and a complete technical system is provided for safe, efficient and autonomous operation of a large-scale satellite group.

The embodiment of the application discloses a kind of based on the collaborative control method, system, equipment and medium of light storage fusion system;The method comprises: obtaining various data and sending to cloud platform by central controller;Various data are cleaned by cloud platform, and photovoltaic power and load power are initially predicted;Set the benefit formula of light storage fusion system and the calculation method related to its influence factor on cloud platform;Get strategy algorithm under different working conditions;According to strategy algorithm, charge-discharge control instruction and energy storage system control instruction are issued to central controller;The charge-discharge control method of different working conditions and different characteristic attributes of user, subsequent cloud platform can be adjusted according to different needs Intelligent collaborative method, remove the coupling between local algorithm, intelligently adjust to make light storage fusion system operating condition and energy consumption optimal, fully utilize the charge-discharge capacity of energy storage system, improve comprehensive income, reduce energy consumption.

The invention discloses a detection algorithm packaging method and apparatus, a device, a medium and a product. The algorithm packaging method comprises the steps of obtaining a mapping relationship between a physical address of an algorithm module structure in an algorithm pool and algorithm content bytes; according to the mapping relationship, establishing an encryption mapping relationship between the virtual relative address and the algorithm content bytes, and obtaining a package of the encryption algorithm table; loading the package to a memory, decrypting the package to obtain an encryption algorithm table, and obtaining a memory algorithm structure according to the encryption algorithm table; obtaining an algorithm demand, and obtaining a hijacked virtual relative address according to the algorithm demand; and obtaining a corresponding result in the memory algorithm structure according to the algorithm requirement and the hijacked virtual relative address. According to the technical scheme, the confidentiality problem of deploying an algorithm on a client in industrial software can be solved, and meanwhile, time is replaced by space, so that the delay problem caused by the fact that a local algorithm structure needs to load a disk physical address algorithm during model training is avoided.

A distributed vector processing method, device, system and medium are disclosed. The system includes at least two clusters, each cluster includes a plurality of nodes, and the number of nodes in each cluster of the at least two clusters is the same. In each cluster, the plurality of nodes include two types of nodes, aggregation nodes and computing nodes. The aggregation nodes in each cluster are used to manage the data in the cluster, and the computing nodes are used to store and calculate the data of at least one partition. Each computing node in each cluster in the scheme provided by the application can perform parallel storage and calculation on data, can quickly transplant local algorithms without a distributed foundation, can realize distributed calculation, and can accelerate the development efficiency of federated algorithms.

The invention belongs to the technical field of voice recognition, and provides a voiceprint recognition method, system, application and device and a storage medium, and the method comprises the steps: obtaining voiceprint original data of a target object; calling a local algorithm library, and performing feature extraction on the voiceprint original data to obtain high-dimensional voiceprint feature data; performing compression processing on the high-dimensional voiceprint feature data to obtain voiceprint recognition features of which the data volume is within a preset range; and inputting the voiceprint recognition features into a chip for operation, comparing the voiceprint recognition features with registered voiceprint features pre-stored in the chip for verification, and outputting an identity recognition result. According to the method, the voiceprint feature data are processed into the high-dimensional voiceprint feature data, on one hand, the quality of the voice data can be improved, and a chip can compare the voice data more conveniently; and on the other hand, the data volume can be reduced and is adaptive to the computing power of a chip.

The invention discloses a deep learning-based path planning method for a factory park, and belongs to the technical field of image recognition and navigation. Comprising the steps of performing global path planning based on an improved A * algorithm; carrying out multiple constraints such as local path planning, speed sampling range comprehensive sign speed limit, load acceleration and obstacle safety distance by using an improved DWA algorithm, and introducing a trajectory evaluation function containing a compliance cost item; a global-local switching detection model taking YOLOv8n as a basic framework is constructed, a GAM attention mechanism is embedded into a neck network, a Wise-IoU loss function is adopted, and laser radar point cloud features are fused, so that the detection precision of small-scale traffic signs and obstacles is remarkably improved; and by calculating a comprehensive weight and comparing the comprehensive weight with a preset threshold value, triggering local algorithm activation or global path re-planning. According to the invention, the problems of difficult sign identification, low planning efficiency and high switching false triggering rate in a factory park environment are effectively solved, and safe, compliant and efficient autonomous driving of the unmanned vehicle is realized.

The application discloses a magnetic susceptibility identification method for deep structure analysis of a mining area, and relates to the technical field of geological mineral exploration. The magnetic susceptibility identification method for deep structure analysis of a mining area comprises the following steps: data acquisition and preprocessing, model construction and inversion, magnetic susceptibility anomaly identification, and structure analysis and classification. According to the initial three-dimensional geological model and through the layered mixed inversion method, the magnetic susceptibility is inverted, the advantages of different inversion algorithms can be fully exerted, the shortcomings of a single algorithm can be made up, the accuracy of the magnetic susceptibility inversion result is improved, the efficient local algorithm and the strong global algorithm can be respectively used according to the geological characteristics of the shallow area and the deep area, the calculation efficiency and stability of the magnetic susceptibility inversion are improved, the set buffer area can smoothly transition and fuse the results of different magnetic susceptibility inversions, the continuous and reliable inversion result is ensured, the subsequent magnetic susceptibility anomaly identification is facilitated, and the identification effect of the magnetic susceptibility identification method is improved.

The invention relates to the technical field of education informatization, and discloses a middle school electrical experiment-oriented intelligent measurement and AI analysis system, which comprises a hardware measurement terminal and an upper computer processing platform, the hardware measurement terminal comprises a power supply and interface module, a microcontroller master control module, a voltage and current acquisition module, a temperature acquisition module and a communication module. The microcontroller master control module is connected with the voltage and current acquisition module, the temperature acquisition module and the communication module. Two complementary analysis modes of an on-line large model and an off-line local algorithm are set on an upper computer processing platform, so that processing logic can be automatically switched according to a network environment and an API key state, and the adaptability of the system in different teaching environments is remarkably improved; at the same time, the cooperative work of the data visualization module of the upper computer software and the mobile terminal display subsystem can improve the measurement, analysis and practice diagnosis efficiency of the middle school electrical experiment, and enhance the pertinence and consistency of teaching feedback.

The application discloses a MOS device key electrical performance prediction method based on a KNN algorithm, which comprises the following steps: obtaining MOS device related parameters, and then using the KNN algorithm to predict the key electrical performance of the MOS device. The prediction logic of the local KNN algorithm adopted by the application is naturally transparent, the performance prediction result is directly derived from the real data of the nearest neighbor sample in the parameter space, the basis of each prediction conclusion can be traced back, meanwhile, the Gaussian kernel weighting mechanism is adopted to further give higher weights to the neighbors with closer distances, which is intuitive and conforms to the physical intuition, so that the prediction process is completely traceable and verifiable, and the scheme does not have an explicit training stage, does not need any parameter iteration optimization process, the model can be deployed and used immediately, and when new data is accumulated, the database can be directly expanded without retraining, when different MOSFET device performance predictions are performed, the scheme has great efficiency advantage, extremely high process adaptability and extremely low maintenance cost.

The invention discloses an optical proximity correction model calibration method, a storage medium and a terminal, and relates to the technical field of semiconductor manufacturing, and the optical proximity correction model calibration method comprises the steps that an optical proximity correction model is established, the optical proximity correction model comprises a photoresist model, and the photoresist model has a linear coefficient to be calculated and a nonlinear variable; searching the linear parameters based on a local algorithm; and searching the nonlinear variable based on a global algorithm. And the linear parameters are searched based on a local algorithm, so that the model calibration efficiency can be effectively improved. According to the method, the nonlinear variable is searched based on a global algorithm, so that the nonlinear variable of the optical proximity correction model is prevented from falling into the difficulty of local optimization in the process of searching the nonlinear variable by adopting a local algorithm, and the calibration precision of the optical proximity correction model is improved.

The application discloses a variable scale satellite group orbit planning method based on a graph proximal optimization algorithm, and realizes real-time and efficient control under the conditions of dynamic change of node quantity and limited communication link. Firstly, the orbit and communication relationship of the satellite group at a time point are abstracted into a dynamic graph. Subsequently, a reinforcement learning structure based on a graph neural network and a recurrent neural network is constructed. A proximal policy optimization algorithm is adopted for policy and value function asynchronous iteration on the satellite, and all calculation and parameter updating are completed on the satellite, without a central control satellite. After each satellite completes local algorithm updating, the satellite exchanges parameter difference with neighbor satellites meeting a reliability condition, and is weighted according to a satellite link quality and a physical distance hybrid weight. The method can be widely applied to application scenarios, such as earth imaging, global communication, navigation and the like, which need constellation level cooperation, and provides a complete technical system for safe, efficient and autonomous operation of a large scale satellite group.

1. Name of the product in this design: Graphical User Interface for Small Model AI Platform Algorithm Management in Electronic Devices. 2. Intended use of this design: for use in an electronic device. 3. The key design features of this product are: the graphical user interface content that displays information. 4. The image or photograph that best illustrates the design's key features: the front view. 5. Purpose of the graphical user interface: This graphical user interface is used for the management of algorithms in the small model AI platform. The main view is the algorithm management interface of the small model AI platform. In the main view, when the user clicks the "Import Local Algorithm" button, they enter the interface change state diagram 1, which displays the local algorithm import interface. In Interface Change State Diagram 1, when the user clicks the "Create Algorithm Type" button, they enter Interface Change State Diagram 2, which displays the Create Algorithm Type pop-up window. In interface change state diagram 2, the user enters the algorithm type name and clicks the "OK" button to enter interface change state diagram 3, which displays the algorithm import interface. In the interface change state diagram 3, the user clicks the "Import Algorithm Package" button for the algorithm type that needs to be imported, and enters the interface change state diagram 4, which displays the algorithm import pop-up window. In the interface change state diagram 4, after the user uploads the corresponding algorithm and fills in the relevant information in the pop-up window, scrolls down the page and clicks the "OK" button in the pop-up window to enter the interface change state diagram 5, which displays the interface where the algorithm import is complete. In the interface change state diagram 5, the user clicks on the algorithm type bar that needs to be edited, and enters the interface change state diagram 6, which displays the detailed information of the selected algorithm type. In the interface change state diagram 6, the user clicks the "Edit" button for the algorithm type that needs to be edited, and enters the interface change state diagram 7, which displays a pop-up window for modifying algorithm information. In the interface change state diagram 7, after the user fills in the relevant information in the pop-up window, they click the "OK" button to enter the interface change state diagram 8, which displays the algorithm editing completion interface. In the interface change state diagram 8, the user clicks the "Delete" button for the algorithm type to be deleted, and enters the interface change state diagram 9, which displays a pop-up window for deleting the algorithm type.

This application discloses a cloud computing-based big data processing method and system. The method, implemented on a cloud computing platform, includes: decomposing a big data processing task into multiple micro-tasks and distributing them to distributed computing nodes; each computing node selecting neighboring nodes based on a hyperlocality algorithm and constructing a local task processing model using only local data and metadata of neighboring nodes dynamically selected based on a fast game mechanism. The metadata includes the semantic features and task dependencies of the micro-task data of neighboring nodes. The semantic features are extracted using natural language processing techniques, and the task dependencies are determined based on the input-output association of the micro-tasks; according to the local task processing model, the micro-tasks are executed independently on each computing node, and the allocation of micro-tasks is dynamically adjusted through a point-to-point local negotiation mechanism; the micro-task processing results from each computing node are aggregated to generate a global data processing result. This method reduces computational and communication overhead.

The invention relates to a time sequence prediction method and system under a nonlinear factor interference condition. The method comprises the following steps: performing normalization, wavelet-Fourier feature extraction and sparse hash coding on a second-level sampling sequence, calculating the change dominance of each index, and constructing a difference matrix; and K-Medoids clustering is executed by taking the difference degree as a distance, and a cluster to which the current period belongs is determined. Then comparing local algorithm residuals such as ARIMA and LSTM in the cluster, dynamically fusing linear and nonlinear prediction by means of linear discriminant analysis and a gating coefficient g, and outputting a rough prediction value; and inputting the residual error into GRU-VAE to generate a reverberation kernel, and performing reversible cancellation on abnormal trailing to obtain a correction result. And the system incrementally updates the disturbance mode library and the meta learning network every 10 minutes to realize rapid self-adaption to the novel peak and reverberation mode. According to the method, the five-step average error is reduced to 0.8% in a composite interference scene, the end-to-end delay is lower than 2ms, and the method can be widely applied to the fields of intelligent operation and maintenance, industrial monitoring, high-frequency financial prediction and the like.

This invention relates to an algorithm-driven process safety decision tree generation and traceable output method and system for standard gas production, belonging to the interdisciplinary field of artificial intelligence and industrial control. The method includes: acquiring input data related to the standard gas production task, wherein the input data includes at least standard clauses retrieved from a local knowledge base, gas property parameters obtained from a property database, and order requirements and equipment parameters; calling a local algorithm function to accurately calculate key physical quantities; mapping the calculation results to a rule base to dynamically generate a process safety decision tree, and binding calculation evidence and standard clause references to each decision node; generating a structured operation list and emergency plan based on the decision tree, and embedding traceable metadata. This invention transforms process decision-making from experience-driven to evidence-driven, achieving traceability and verifiability of the decision-making process, and significantly improving the safety, consistency, and automation level of standard gas production.

The invention relates to the technical field of computer vision, and discloses a Risc-v local algorithm-based electric power equipment nest and suspended matter detection system and method, and the method comprises the following steps: S1, collecting the image data of an electric power equipment site through electric power inspection equipment, and generating an original image data set, s2, preprocessing the original image data set, including image denoising, enhancement and size standardization, generating preprocessed image data, deploying a lightweight deep learning model through a local edge computing device based on a Risc-v architecture, and combining image denoising and enhancement preprocessing operations to obtain a preprocessed image data set; the problem of image quality degradation caused by field environment illumination change and weather interference can be effectively solved, so that the system can improve the feature definition of an input image in real time, sufficient extraction of bird nest and suspended matter target features in a complex environment is ensured, and the accuracy and reliability of a detection result are improved.

The application relates to a language model-function calculation anti-illusion method and system based on a model context protocol, and belongs to the technical field of cross artificial intelligence and industrial control. The method comprises the following steps: performing semantic understanding on a user query through a locally deployed first language model, identifying a calculation task type, and extracting key parameters; converting the key parameters into standardized structured parameters and performing legality verification; calling a local algorithm function according to the calculation task type, combining a local physical property parameter database to perform accurate calculation, and outputting a calculation result with an intermediate process; and performing language interpretation and format layout on the calculation result through a locally deployed second language model. Each step is decoupled through a model context protocol module, and data exchange is performed through a standardized protocol. The application separates numerical calculation from large model reasoning, fundamentally avoids illusion risks, and retains the advantages of natural language interaction and document generation of a large model.

This invention provides an adaptive localization method for microseismic events in deeply buried tunnels. It includes an optimization initialization mechanism based on target gradients to accurately locate the initial point. Building upon existing local algorithms, it explores and analyzes the relationship between the seismic source location and the distribution of the sensor array in real engineering environments, as well as the spatial morphology of the objective function value. This summarizes the patterns of the initial point and the objective function gradient, determines the initialization region, optimizes the initial point selection, and maintains the algorithm's convergence rate. A data evaluation strategy is used to process low-quality data affected by the engineering environment. Based on the actual environment of deeply buried tunnel engineering and the two-wave theorem, data filtering and weighting rules are set to ensure that the strategy conforms to the actual engineering background, providing a sufficient and reliable theoretical basis for data elimination and weighting strategies. This invention solves the problems of difficult and complex microseismic event localization in tunnels, achieving accurate localization of microseismic sources and ensuring safe production in deeply buried tunnel engineering.

The invention relates to the technical field of data processing, in particular to an intelligent cleaning control system and method for a medical medicine basket, and the system comprises a synchronous collection module, a rule decision module, a collaborative execution module, a disaster recovery processing module and an interactive feedback module. The synchronous acquisition module outputs temperature, pressure, turbidity and moisture thickness data with synchronous timestamps; the rule decision module outputs a fault pre-judgment instruction, and the federal learning optimizer generates a heat recovery efficiency instruction; the cooperative execution module executes water pressure adjustment, heat exchange control and gradient drying operation; when the disaster recovery processing module network delay exceeds a threshold value, the federal learning optimizer is switched to load local data; and the interaction feedback module generates an AR interface and analyzes the voiceprint instruction to correct the control instruction. Through spatio-temporal data synchronization, local algorithm decision, disaster recovery data closed loop and a man-machine interaction backflow mechanism, full-process autonomous control in a network fluctuation environment is realized, and a medical cleaning process is guaranteed to meet a preset standard.

This invention discloses a multi-mode collaborative method and medium for allocating distribution network inspection tasks, involving the intersection of power system automated inspection technology and UAV scheduling technology. Addressing the low inspection efficiency of existing UAV inspection methods, this application first divides independent coverage areas and boundary areas using K-Means clustering and energy consumption constraints. Then, it employs an ant colony optimization algorithm to solve the sequence planning problem for independent areas, utilizing dual pheromone collaborative optimization of target allocation and inspection order, and combining a negative feedback mechanism to quickly eliminate suboptimal solutions. Finally, it uses a local PI algorithm to perform secondary allocation of towers in the boundary area, achieving optimal cost and thus complete allocation of all towers. This application breaks through the capability limitations of a single inspection mode, achieving full-scenario coverage and optimized resource allocation, significantly improving inspection efficiency.