665 results about "Adaptive method" patented technology

The invention provides a traffic control strategy self-adaption method and system based on simulation feedback, and belongs to the technical field of traffic prediction and control, and the method comprises the steps: firstly obtaining traffic control scene data containing traffic flow data and road condition information, then constructing a simulation evaluation environment, configuring scene parameters based on the traffic control scene data, and carrying out the simulation evaluation environment; the method comprises the following steps: simulating traffic operation states under different traffic control strategies, calling a pre-trained reinforcement learning model to perform simulation evaluation on each strategy in a traffic control strategy set, generating a strategy effect feedback set comprising a traffic operation efficiency index and a traffic order stability index, and according to the strategy effect feedback set, calculating the traffic order stability of the traffic control strategy. And performing parameter adjustment on the strategy in consideration of the index association relationship to obtain an adjusted strategy, and finally outputting the adjusted strategy to the traffic control system to realize strategy updating, thereby effectively improving traffic operation efficiency, ensuring traffic order stability, and realizing adaptive optimization of the traffic control strategy.

The invention discloses a self-adaptive NL2SQL method and system based on a large language model and retrieval enhancement. The method comprises four key steps of mode linking, example enhancement and SQL generation, multi-path thinking chain fusion reasoning and multi-round self-correction. The system comprises the following modules: a knowledge base management module, an input analysis and preprocessing module, a mode link module, an example enhancement module, an SQL (Structured Query Language) generation module, an SQL fusion and optimization module, an SQL execution and feedback module, a multi-round self-correction module, a Prompt construction and context management module, a system interface module and a content generation module. The method has the advantages that multiple large models are supported, a data source is quickly accessed, a user-defined prompt structure is configured, good engineering maintainability and scene adaptability are achieved, and SQL generation accuracy, performability and universality are improved.

The invention discloses a hierarchical traffic prediction-based multi-terminal video display adaptive method, which comprises the following steps of: acquiring network traffic, user behaviors and equipment state information in real time by constructing a multi-dimensional feature acquisition module, and generating a dynamic feature matrix containing traffic, interaction and performance features; processing the feature matrix by using a deep learning model, and generating and dynamically adjusting a network traffic grading prediction result; formulating an optimized video content distribution strategy based on a prediction result and equipment performance parameters, and coordinating multi-terminal resource distribution; through user behavior analysis and network state monitoring, a self-adaptive video display control strategy is generated, dynamic adaptation of video resolution, frame rate and playing logic is realized, a full-link collaborative optimization module is constructed, terminal, network and application layer resource allocation is coordinated, a video transmission and display strategy is optimized, and network flow prediction and multi-terminal interaction requirements are combined to realize multi-terminal interaction. And generating a dynamic video layout and content switching strategy.

The invention provides a dual-antenna attitude and orientation and robust self-adaption method based on integrated navigation, which comprises the following steps: combining an INS (inertial navigation system) and a dual-antenna GNSS (global navigation satellite system), and enabling the system to output continuous and stable high-precision carrier attitude information when a GNSS signal is interfered by using dual-antenna integrated navigation. According to the method, most of satellite end clock correction, ionosphere and troposphere errors and receiver end clock correction are eliminated by using double-antenna GNSS pseudo-range and carrier phase double difference, robust statistics are calculated through heading prediction residual vectors to obtain robust factors, an observation noise covariance matrix is expanded to reduce the influence of system noise and observation noise, and the system performance is improved. A Mahalanobis distance based on a prediction residual vector is introduced to detect whether a system is abnormal or not, a state prediction covariance matrix is adjusted through a self-adaptive factor, weight reduction of an abnormal INS dynamic model is achieved, and the stability and reliability of system attitude information are improved through a noise covariance self-adaptive control mechanism.

The invention relates to the technical field of wind power generation. The invention provides a yaw static and dynamic error adaptive method based on big data backtesting, which comprises the following steps of: constructing a multi-dimensional feature vector through multi-source data fusion data, and establishing a dynamic feature data set; based on the dynamic characteristic data set, a space-time diagram convolutional network is adopted to establish a wind power plant dynamic error prediction model, and space-time evolution rules of wind shear, turbulent flow and wake flow effects are captured; constructing a variational self-coding reference model based on historical full wind speed section data, calculating a residual error between a current working condition and the variational self-coding reference model in real time, and taking the residual error as a static error prediction value; performing adaptive weight fusion on the static error prediction value and the dynamic error prediction value to obtain a fusion error; and the fusion error is converted into the yaw angle correction amount, and the wind facing action is executed. The problems that an existing yaw error recognition technology is high in data dependence, lack of dynamic analysis, poor in model adaptability and difficult in complex wind field processing are solved.

The invention relates to the technical field of large model algorithms, in particular to a knowledge distillation-based field large model lightweight adaptive method and system, and the method comprises the steps: obtaining knowledge distillation parameters and student model performance parameters, and building a nonlinear mapping relation between the knowledge distillation parameters and the student model performance parameters; the optimal parameter combination is optimized and solved based on the mapping relation, and target knowledge distillation parameters are generated; issuing the target parameters to a training engine, monitoring performance deviation in real time and triggering re-optimization; in the reasoning process, performance fluctuation is monitored, and model characteristics are managed and controlled; target domain data characteristics are collected, a mapping relation is corrected in combination with big data analysis, and the domain adaptation capacity is improved; a knowledge base and a case base of historical distillation data are constructed, a standardized adjustment scheme is formed, and self-adaptive matching is achieved. According to the scheme, through precise modeling, dynamic optimization, real-time monitoring and knowledge reuse, the knowledge distillation efficiency, model robustness and field adaptability are remarkably improved, and systematic technical support is provided for large model lightweight.

The invention relates to the technical field of multi-agent cooperative control, and discloses a multi-agent reinforcement learning formation adaptive method fusing a graph attention mechanism. According to the method, a dynamic graph state space is constructed, and a speed self-adaptive communication mechanism and a multi-target return function are combined, so that the problem of insufficient generalization ability during intelligent agent number change or topological structure dynamic adjustment is solved. The scheme comprises the following steps: establishing a virtual leader-following formation model and a motion dynamics model; designing an adaptive communication radius formula based on speed change; constructing a graph state space, and dynamically aggregating neighbor information by using a multi-head attention mechanism; designing a strategy network and a value evaluation network with residual connection; and a centralized training distributed execution framework is adopted to update model parameters. Adaptive extraction of key neighbor information is realized through a graph attention mechanism, and adaptability and stability of a formation system in a complex environment are improved by combining dynamic communication range adjustment. The method is suitable for unmanned aerial vehicle group control, robot cooperation and other scenes.

The invention discloses a fire point self-adaption method based on multi-source synchronous satellite data, which relates to the technical field of remote sensing and geographic information, and comprises the following steps of: aiming at satellite remote sensing images with different resolutions, respectively selecting fire point detection algorithms matched with the resolution characteristics of the satellite remote sensing images to independently finish preliminary detection of fire points, extracting fire point space distribution and feature data corresponding to respective resolution scale; preprocessing the extracted multi-source fire point spatial distribution and feature data, and establishing a multi-source fire point data set; according to the method, the scale consistency and fusion precision of multi-source satellite fire point detection are improved, and the problems of fire point boundary breakage, false splitting and the like caused by resolution difference are solved. Through scale difference quantification and machine learning evaluation, in combination with scale correction of high-resolution images, adaptive dynamic adjustment of fire point boundaries is realized, and the reliability of fire monitoring and emergency decision making is improved.

The invention discloses a data interaction self-adaptive method for a photovoltaic protocol converter, and particularly relates to the technical field of power grid dispatching cooperative control. The method comprises the following steps: monitoring and calculating a voltage instantaneous change rate of a grid-connected point of a power grid in real time, and comparing the voltage instantaneous change rate with a threshold value; when a threshold value is exceeded, synchronously collecting a current phase and generating a voltage and current phase difference change direction; extracting a voltage high-frequency component, analyzing frequency band energy distribution through wavelet packet decomposition, and calculating a Shannon entropy representing an energy concentration ratio; when the phase difference change direction accords with a preset power grid short-circuit fault feature and the Shannon entropy value is lower than a preset entropy threshold value, determining that a power grid transient event occurs; at the moment, a data caching mechanism of the protocol converter is stopped immediately, and a real-time stream transmission channel is activated; an emergency control instruction issued by a power grid dispatching system is directly transmitted to the photovoltaic inverter through the channel; and when the voltage instantaneous change rate is continuously lower than the preset threshold value for a preset duration, recovering the data caching mechanism of the protocol converter.

An adaptive transformation matrix generation is used simultaneous compression and encryption of data. The system continuously monitors input data streams to detect changes in distribution patterns, dynamically updating transformation matrices in response to these changes. Using performance evaluation criteria, the system selects and deploys optimal matrices that maintain both compression efficiency and cryptographic security as data characteristics evolve. The system implements configurable adaptation policies that govern when and how matrices are updated, ensuring system stability while maximizing performance. By employing sliding window analysis and distribution variance metrics, the system can quantify data drift and trigger appropriate adaptations. The system maintains compatibility with various operational modes including lossless, lossy, and modified lossless configurations, applying mode-specific optimization criteria to each scenario. This adaptive approach offers a resilient solution for data transmission and storage scenarios where both data reduction and security must be maintained across diverse and evolving data streams.

The invention discloses an offshore wind turbine foundation scouring full-period intelligent monitoring method and system based on multi-source data fusion and a dynamic coupling model, and relates to the technical field of intelligent monitoring, and the method comprises the steps: deploying a multi-source monitoring module, and constructing a finite element model; carrying out load calculation and parameter inversion; and training full-cycle dynamic updating of the washout failure function model. According to the method, a self-adaptive Kriging-Bayesian method is adopted, a Bayesian inversion framework and a self-adaptive agent model are fused to solve optimal soil body parameters, full-period model dynamic updating based on dynamic monitoring data is achieved, a multi-fidelity deep kernel learning model is adopted, three types of data are fused into a training set, full-period intelligent monitoring of offshore wind turbine foundation scouring is achieved, and the method has the advantages of being simple in structure, convenient to operate and high in practicability. The dynamic identification of soil parameters is realized by combining a self-adaptive inversion framework with a displacement error closed-loop optimization mechanism, and the technical problem that a traditional static model cannot adapt to the spatial-temporal variability of seabed geology is solved.

The invention provides a power distribution network single-high-frequency resonance rapid tracking and adaptive suppression method, a multi-mode dynamic cooperative adaptive suppression system is established based on a PCCVF adaptive method, and the damping characteristics of a power distribution network are remodeled by injecting compensation current into a power distribution network system so as to realize broadband resonance suppression of the power distribution network. The compensation current injection method comprises the following steps: step 1, detecting resonant frequency deviation of a power distribution network through real-time FFT (Fast Fourier Transform); 2, performing dynamic phase angle correction based on a phase prediction residual error of real-time frequency deviation; step 3, constructing a layered impedance remodeling module, and ensuring stable power transmission; 4, establishing a frequency-variable impedance model of the power distribution network based on resonance energy spectral density analysis, dynamically optimizing parameters through fuzzy logic of a bell-shaped membership function, and feeding harmonic compensation current into the power distribution network by using a space vector pulse width modulation technology; according to the invention, single high-frequency resonance can be effectively suppressed, the response and tracking performance of the system is improved, and the spectrum analysis efficiency and bandwidth occupation are optimized.

The invention discloses an asymptotic tracking control method for a mobile double-flexible mechanical arm network. The method comprises the following steps: constructing the mobile double-flexible mechanical arm network; the uncertainty of a leader is considered, and a self-adaptive distributed switching observer is constructed; a servo system is introduced, transverse displacement is generated, and a tracking error model is constructed; an adaptive neural network controller is constructed by considering unknown gain faults and parameter uncertainty; on the basis of a distributed switching observer and a neural network controller, asymptotic consistency tracking control over a mobile double-flexible-mechanical-arm network is achieved. According to the method, asymptotic fault-tolerant consistency tracking control of the mobile double-flexible mechanical arm network can be effectively realized under heterogeneous linear leader and denial of service attacks, and the problems of unknown gain faults and unknown parameters are solved by utilizing a self-adaptive method and a neural network technology; the moving position of the moving double-flexible mechanical arm and the angle position of the two flexible mechanical arms reach the specified transient performance, and asymptotic consistency tracking is achieved.

The invention discloses a cross-modal remote sensing image unsupervised adaptive method based on unreliable pseudo tag guidance. The method comprises the following steps: step 1, acquiring data of a source domain and a target domain and preprocessing the data; 2, constructing a teacher-student self-training framework, generating a pseudo tag for a target domain by a teacher network, and filtering noise through confidence evaluation; 3, dividing the image into reliable pixels and unreliable pixels according to the confidence coefficient, and generating a mask; 4, constructing positive samples, negative samples and anchor point features; 5, designing an unreliable sample guide pixel contrast loss function; 6, optimizing the loss function training model until the optimal performance is achieved; and 7, predicting test set data by adopting the trained model to obtain a semantic segmentation result. According to the method, the potential of pseudo labels of which modals are difficult to label is fully mined, the problem that unreliable pixels are insufficient in utilization during training is solved, and finally more effective cross-modal domain alignment and better cross-modal unsupervised domain adaptive semantic segmentation precision are realized.

The invention discloses a reinforcement learning-based quadrotor unmanned aerial vehicle attitude preset performance control method, which comprises the following steps of: establishing an unmanned aerial vehicle attitude kinetic equation by considering a time-varying inertial parameter and external unknown interference; a sub-channel preset performance controller is constructed, and a robust control quantity is generated in combination with preset performance and a non-singular fast terminal sliding mode control technology; designing a reinforcement learning parameter generator, and dynamically optimizing 12 time-varying parameter estimated values through a time sequence feature extraction network and a residual network; and establishing an online reinforcement learning training mechanism, constructing a multi-target reward function, and optimizing the output of the parameter generator through the multi-target reward function. According to the method, a reinforcement learning method is adopted to replace a traditional self-adaptive method to estimate time-varying parameters, multi-degree-of-freedom decoupling optimization is achieved through a sub-channel control architecture, tracking error preset performance constraint is achieved under the working condition of time-varying inertial parameters, and the dynamic adjustment capacity and anti-interference robustness of an attitude system are remarkably improved.

The invention relates to the technical field of satellite navigation and positioning, in particular to a zenith troposphere wet delay calculation method based on deep learning, and the basic steps of segmented ZWD multi-source data fusion based on an improved Transform model are as follows: establishing a ZTD model, establishing a TCA-Transform model, and fusing data. The task of establishing the ZTD model is to calculate the atmospheric refractive index according to meteorological data so as to obtain ZTDPP; the TCA-Transform model establishment comprises the following steps of: improving a Transform model by using a domain self-adaptive method; according to the data fusion, the ZWDPP and the ZWDMODIS are fused by using a TCA-Transform model. And finally, a ZWD product with high temporal-spatial resolution and high precision can be obtained. The generalization ability of an original Transform model is improved, so that the troposphere delay correction ability in a specific area is improved, and a zenith troposphere delay (ZWD) product with high resolution and high precision is obtained.

The invention provides a hierarchical ethical adaptive method and system based on a development stage, and the method comprises the steps: collecting the multi-modal behavior data of a user, and recognizing the cognitive development stage of the user; constructing an ethical rule knowledge graph with a hyponym inheritance and context activation mechanism; fusing user acceptability and ethical conflict strength, and performing strategy balance based on a game model; calling the generative language model and the hierarchical template library to generate matched ethical feedback content; different culture expression styles are adapted through a culture migration network; and a user stage transition window is predicted based on the cognitive evolution trend, and an intervention mechanism and strategy adjustment are triggered. The system supports personalized ethical guidance strategy configuration, is compatible with various user types and cross-culture situations, realizes adaptive optimization and expression intellectualization of ethical decision, and improves user understanding degree, acceptability and ethical guidance effect. The method is suitable for multiple scenes such as education guidance, value intervention and man-machine ethical interaction.

The invention relates to the technical field of wireless communication, and discloses a wireless communication adaptive method and system based on a data transmission state, and the method comprises the steps: obtaining multi-source data, carrying out the preprocessing of the multi-source data, obtaining a CSI compression matrix, predicting the channel coherence time, dynamically adjusting the CSI sampling interval, and defining a load-channel coupling factor; obtaining cross-layer data based on the CSI compression matrix, performing fusion through a rotation matrix to obtain a fusion matrix, and extracting a physical layer fusion feature and an application layer fusion feature to calculate a multi-target state score; establishing a 5G power compensation mechanism based on cross-layer data, calculating a four-dimensional influence tensor, performing tensor decomposition and optimal action selection, and decomposing T into a core tensor and a factor matrix; selecting an optimal parameter combination through modular product calculation; based on cross-layer data, a quantum entanglement feedback mechanism is introduced, data is fed back, entanglement state association cross-layer indexes are designed, a quantum gate is adjusted through entanglement state design, and a model is updated in real time in combination with incremental learning.

The invention relates to the technical field of data storage and management, and discloses a self-adaptive storage capacity adjusting method and system for an industrial-grade solid state disk. The method and the system can intelligently analyze and predict future storage requirements by collecting key performance indexes and user data access modes of the solid state disk in real time, remarkably improve the efficiency of storage management, and can automatically identify behavior modes of storage use and formulate corresponding storage strategies through a machine learning algorithm, thereby improving the efficiency of storage management. According to the self-adaptive method, hot data is ensured to be quickly accessed, cold data is reasonably migrated, so that the utilization rate of storage space is optimized, in addition, the storage performance is further improved and the risk of data loss or damage is reduced by adopting a garbage collection algorithm and a TRIM command, and the self-adaptive method not only improves the response speed and the reliability of storage equipment, but also reduces the maintenance cost, and is suitable for popularization and application. And a more efficient and flexible storage solution is provided for enterprises.

The invention discloses a multi-modal remote sensing image registration method based on domain self-adaption and an optimizer set algorithm, and relates to the field of image processing, and the method comprises the steps: respectively extracting the feature vectors of a reference image and an image to be registered, carrying out the feature matching of the feature vectors according to the class label containing conditions of the feature vectors, and obtaining a registration result; obtaining a feature matching vector based on a matching result; estimating transformation parameters corresponding to the affine transformation model by using a random sampling consistency algorithm, and optimizing the affine transformation model according to an estimation result and similarity measurement; and performing spatial transformation on a to-be-registered remote sensing image by using the optimized affine transformation model to realize spatial alignment between the reference image and the to-be-registered image. According to the method, firstly, the feature points of the multi-modal remote sensing image are extracted, the difference of feature vectors is reduced by using a domain adaptive method, the precision and robustness of feature matching are improved, transformation parameters are optimized through an optimizer set algorithm, and search is prevented from falling into local optimum.

The present invention provides an adaptive drone image matching pair selection method and system, comprising: S1: acquiring drone images, performing feature extraction on the drone images using the SIFT algorithm, and obtaining a local feature set of the drone images; S2: constructing a vector codebook, performing feature aggregation on the local feature set using the vector codebook, and obtaining a feature matrix set; vectorizing the feature matrix set to obtain a global feature descriptor vector set; S3: performing image retrieval on the global feature descriptor vector set based on a graph index structure to obtain a drone image matching pair set; S4: performing three-dimensional reconstruction on the drone image matching pair set to obtain a three-dimensional reconstruction model. The present invention uses global feature descriptor vectors to replace word frequency calculation based on local features, and provides a graph-based indexing strategy to achieve efficient overlapping image search, making drone image matching pairs more efficient and accurate.

According to the thin-wall part turning parameter self-adaption method and system, the background vibration signals are collected in the non-cutting period to generate the background frequency spectrum base line, the cutting force signals are purified through the base line in the cutting period, cutting state recognition and parameter adjustment are conducted on the basis of the purified signals, and therefore the machining precision of the thin-wall part is improved. The problem that the background vibration of the machine tool interferes with the cutting force signal is effectively solved, and therefore the method has the advantages that the interference of the background vibration of the machine tool on the cutting force signal can be effectively removed, the accuracy of cutting state recognition is improved, more accurate self-adaptive adjustment of machining parameters is achieved, and the machining quality and stability of thin-wall parts are improved.

The invention belongs to the field of electromagnetic field numerical solution, and particularly provides a finite element mesh adaptive method for network parameter analysis of a multi-port microwave device, which comprises the following steps: firstly, establishing a geometric structure of the multi-port microwave device, setting simulation frequency, excitation and boundary conditions, and establishing an electromagnetic simulation model; secondly, performing preliminary division and finite element electromagnetic simulation calculation on a computational domain by adopting a tetrahedral mesh to obtain a numerical solution corresponding to each group of excitation, and synthesizing results based on a posterior error estimation method for posterior error estimation; and finally, according to a posterior error indication grid encryption parameter, carrying out simulation calculation again after encryption, and repeating the process until the solution meets the precision requirement. In conclusion, by combining the characteristics of a finite element algorithm and the characteristics of an error estimation algorithm, in the microwave device network parameter analysis process, multiple sets of excited numerical solutions are comprehensively used for posterior error estimation and grid adaptive encryption indication, so that the analysis precision of network parameters is improved.

The invention discloses an IEC61850 multi-protocol version oriented communication server-side self-adaption method and device, and the method comprises the steps: carrying out the protocol version dynamic recognition, service interface dynamic configuration and service interface dynamic response for IEC61850 standard protocol versions including ED1.0, ED2.0 and ED2.1, and forming a differential processing mechanism for different protocols, and the automatic switching of the ED1.0 / ED2.0 / ED2.1 protocol can be realized. According to the invention, through compatibility adaptive processing of different versions of difference communication services, system configuration complexity caused by protocol version difference and multi-version maintenance cost of device software are reduced, so that diversified requirements of a power system in multi-scene application and internationalized markets are met; and the flexibility of system configuration and the reliability of device operation are improved.

The invention discloses a passive domain self-adaptive encrypted traffic detection method based on self-training Mama. The method comprises the following steps: detecting encrypted traffic by adopting a detection model of the encrypted traffic; the domain adaptation of the detection model comprises offline adaptation and online adaptation. The off-line adaptive method is used for carrying out robust self-adaption on encrypted traffic offset in a static traffic data set under a passive domain and semi-supervised domain adaptive SF-SSDA normal form. The method comprises the following steps of: 1) pre-training a source domain based on a mask auto-encoder MAE; and 2) target domain adaptation training. According to the online adaptation method, on the basis of offline adaptation, an offline batch screening mechanism is replaced with an online cache queue, and online adaptation of the flow type threat flow is achieved. Experimental results on a real world and a public reference data set show that the method realizes faster reasoning with fewer parameters, and is significantly superior to a representative passive baseline method in the aspect of cross-domain detection accuracy.

The invention discloses a contact network state evaluation method and system based on adaptive empowerment optimization. The method comprises the following steps: collecting dynamic and static parameters of a contact network, calculating subjective and objective weights of each parameter by adopting an analytic hierarchy process and an entropy weight method, and calculating a static composite weight by adopting a combined weight method; historical degradation data of dynamic and static parameters of the overhead line system are collected, and the expected degradation rate of each parameter is obtained by adopting a time sequence neural network; calculating a parameter real-time degradation rate, and dynamically optimizing the static composite weight by adopting an adaptive method based on the real-time degradation rate and an expected degradation rate to obtain a dynamic composite weight optimization result; optimizing and updating the gain in the adaptive method; based on a dynamic composite weight optimization result, performing state evaluation on the contact network; the method and the system provided by the invention are not only suitable for contact network state evaluation, but also can be popularized and applied to other state evaluation fields, and have wide application prospects.

The application discloses a short video code rate self-adaptive method based on meta learning, relates to the technical field of streaming media, and comprises the following steps: S1, offline training, a model is established to represent user characteristics and network prediction information; S2, online learning, according to the characteristics of the current user environment, the model parameters are adjusted and optimized. The short video code rate self-adaptive method based on meta learning is adopted, a new SABR framework based on meta learning is successfully realized, the framework can quickly adapt to different user demands, the practicability and the calculation speed of the system are improved, and the framework has industrial application; the offline training and the online learning technology are successfully combined, the generalization and the stability of the model are enhanced; the idea of action masking is introduced in pre-training, the rationality and the reliability of decision are enhanced, the data amount required by meta learning is effectively reduced, the learning efficiency and the accuracy are improved, and the data demand and the training time in the industrial environment are significantly reduced.

The invention discloses a robust noise tag domain adaptive method based on confidence and information entropy. The method specifically comprises the following steps: preparing data in a double-domain data set of a source domain and a target domain; in the preheating stage, the model is trained by adopting source domain labeled data; confidence modulation adaptive learning is carried out; a confusion alignment mechanism based on entropy guidance; establishing a joint optimization loss function; and outputting an adaptive result. By introducing a joint optimization strategy of confidence adjustment learning and entropy guide confusion alignment, the learning intensity is dynamically adjusted, and the inter-class confusion is reduced, so that the problems caused by noise labels and cross-domain distribution differences are effectively solved, and the robustness and stability of field self-adaption are improved; under the condition that there is no label in the target domain, optimization is carried out by means of prediction confidence and entropy information, and the target domain classification accuracy under the condition of high noise label proportion and unbalanced category distribution is improved.

The invention discloses an intelligent OCR dynamic self-adaption method and system based on multi-modal fusion, and relates to the related field of multi-modal information fusion process.The method comprises the steps that multi-modal information is uploaded to an OCR recognition platform, first-modal information is selected and subjected to fuzzy scanning, and layout analysis is executed to determine a layout structure; a hierarchical progressive fusion condition is introduced according to content complexity, a dynamic fusion normal form is set, and an OCR engine array deployed by an OCR recognition platform is initialized; and according to the layout structure, carrying out multi-step focusing fusion planning, triggering directional focusing based on the first modal information and entity alignment focusing aiming at the additional modal information, dynamically configuring an OCR engine and a fusion normal form, and carrying out scanning identification management under multi-modal fusion. The technical problems of low recognition accuracy and poor adaptability of the existing multi-modal information OCR processing are solved, and the technical effect of improving the recognition accuracy and robustness of the multi-modal information OCR is achieved.

The invention provides a low-rank personalized blood pressure estimation method, which comprises the following steps of: performing pre-training on a large-scale PPG-blood pressure data pair on a group level based on a UniTS model of a Transform backbone network, and learning a mapping relation from a PPG signal to blood pressure; carrying out personalized fine tuning on the pre-trained model by adopting a low-rank adaptive technology to realize model adaptation under the condition of few samples; in the personalized fine tuning process, a pulse pressure segmented penalty loss function is introduced, total training loss is formed by combining mean square error loss, and prediction results of the systolic pressure and the diastolic pressure are restrained to conform to the physiological law; the problem of sampling rate difference is solved by adopting a low-rank self-adaptive method with a stable sampling rate, and stable blood pressure estimation across equipment is realized; the invention aims to realize high-precision and cross-device robust blood pressure estimation according with physiological rules under the condition of few-sample calibration by introducing a physiological constraint loss function and sampling rate robust adaptation mechanism only depending on PPG signals and through a framework combining group pre-training and low-rank personalized fine tuning.