632 results about "Filter methods" patented technology

The invention discloses a text and image fused online comment toxicity detection and filtering method and system, and relates to the technical field of natural language processing, and the method comprises the steps: extracting a semantic vector of a comment text through a RoBERTa-Marge model; the visual features of the image are extracted through an OfficientNet-V2 model; aligning heterogeneous modal features by adopting a double-flow contrast loss function; self-adaptive decision making of culture sensitivity: loading a regional sensitive rule table according to a user IP address, and dynamically adjusting symbolic semantics; calculating an intimacy correction factor based on the social relationship between the publisher and the receiver; and performing context weighted toxicity scoring, calculating a user historical behavior weight, and outputting a final toxicity probability. According to the method, a deep dynamic mapping mechanism of text and visual features is constructed, heterogeneous features are extracted through RoBERTa-Large and OfficientNet-V2 double-flow architectures, semantic space alignment is forced by utilizing comparative learning, and the problem of image-text splitting detection in the traditional technology is solved.

The embodiment of the invention discloses a digital key ranging value filtering method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring ranging information of a plurality of UWB anchor points in real time; obtaining the position change trend of the digital key according to the distance measurement information; under the condition that the effective distance measurement value does not exist, a preset target position is obtained according to the positioning result corresponding to the moment t of the previous effective distance measurement value, the position change trend and the unlocking and locking state of the vehicle terminal, and a predicted distance is obtained according to the distance relation between the effective distance measurement value at the moment t and the preset target position; obtaining a prediction duration T according to the prediction distance; in the T, when an effective distance measurement value is not detected, state one-step prediction of a Kalman filtering algorithm is executed to obtain a filtered distance measurement value, and when the effective distance measurement value is detected, the state one-step prediction is terminated, and a filtering estimation value is obtained according to the Kalman filtering algorithm to serve as the filtered distance measurement value; and obtaining a positioning result of the area outside the vehicle according to the filtered distance measurement values of the plurality of anchor points.

The invention relates to the technical field of sensitive word filtering, in particular to a sensitive word filtering method and device based on multiple modes. The method comprises the steps of obtaining to-be-detected data; wherein the to-be-detected data at least comprises one piece of to-be-detected image data and one piece of to-be-detected text data; performing feature extraction on the to-be-detected data to obtain image features and text features; mapping the image features and the text features to a multi-modal causal knowledge graph to obtain a to-be-detected causal knowledge graph; wherein the to-be-detected causal knowledge graph is used for representing a sensitive relationship between a to-be-detected image and to-be-detected text data in the to-be-detected data; according to the method, sensitive word filtering is performed on the to-be-detected data based on the to-be-detected causal knowledge graph, so that the sensitive words in the text are accurately filtered.

The invention relates to the technical field of satellite internet, and discloses a low-orbit satellite phased array multi-beam interference modeling and suppression method and system, and the method comprises the steps: selecting a Kaiser window as a core filtering method, and achieving the optimization of beam characteristics through the dynamic adjustment of a shape parameter beta; generating an initial beam directional diagram based on a digital phase matching method, and multiplying the Kaiser window function coefficient by the excitation weight of the 64-array-element linear array element by element to realize spatial domain weighted filtering; the method comprises the following steps: constructing a training data set containing multi-scene interference characteristics, calculating a corresponding covariance matrix and an accurate inverse matrix thereof to form a sample pair, designing a deep neural network architecture, inputting a flattened covariance matrix vector, and learning a complex nonlinear mapping relation from the covariance matrix to the inverse matrix through a multi-layer full-connection structure; a mean square error is used as a loss function to constrain network output precision, and a multi-beam interference system model is constructed; according to the invention, stable and efficient communication of the low-orbit satellite system in a complex electromagnetic environment and under rapid channel change is ensured.

The invention relates to the technical field of intelligent control, in particular to a coke oven gas collecting tube pressure dynamic adjusting method and system based on intelligent control, and the method comprises the steps: obtaining pressure data from a gas collecting tube through a sensor, and storing the pressure data in a time sequence format to obtain a pressure deviation initial sequence; according to the pressure deviation initial sequence, calculating a data difference value and carrying out smoothing processing to obtain a change rate sequence; analyzing the change trend of the change rate sequence, and calculating a secondary difference value to obtain a change acceleration sequence; according to the change rate sequence and the change acceleration sequence, predicting a pressure deviation development trajectory by adopting a filtering method to obtain a predicted deviation trajectory; and if the predicted deviation trajectory exceeds a preset threshold range, obtaining a deviation influence evaluation value through historical data comparative analysis. The problem that the pressure control precision of the coke oven gas collecting pipe is poor in the high-temperature and high-pressure environment is solved, and the pressure adjusting precision of the gas collecting pipe in the high-temperature and high-pressure environment is improved.

The invention discloses a GRACE and Swarm time-varying gravity field fusion filtering method based on a state space model, and the method comprises the steps: taking a spherical harmonic coefficient as a state quantity, constructing a random walk process equation, introducing three types of observations, employing a quantization parameter for observation noise and process noise covariance, constructing according to an order / power law, and carrying out the self-adaptive updating along with the monthly; a Nelder-Mead method is adopted to search for spectral index parameters, and an EM algorithm and a statistical method are utilized to update other parameters in a closed / quasi-closed mode; obtaining the state and posterior covariance of a full time sequence by using Kalman filtering and RTS smoothing; in the GRACE and GRACE-FO window period, continuous reconstruction is carried out by means of a process model and Swarm; and outputting quality evaluation information including monthly gravity field coefficients, posterior covariance, innovative variance ratio, residual whitening test, space power spectrum, uncertainty band and the like. According to the method, while physical rationality and calculation feasibility are ensured, a continuous and stable monthly time-varying gravitational field sequence with quantifiable uncertainty is realized.

The invention discloses a geometric constraint fitting point cloud filtering method for sea surface three-dimensional reconstruction, and belongs to the technical field of computer vision and three-dimensional reconstruction. The objective of the invention is to solve the problem of insufficient subsequent three-dimensional reconstruction precision caused by interference of reflection noise, mismatching points and the like in sea surface point cloud. The method specifically comprises the following seven steps: firstly, acquiring sea surface original point cloud through three-dimensional data acquisition equipment; a filtering technology is adopted to obtain a to-be-fitted point cloud; fitting a quadric surface through an improved RANSAC (Random Sample Consensus) algorithm to solve an initial parameter; constructing a comprehensive error function, and optimizing the model through gradient descent; effective inner points are screened through quadratic term coefficient constraint and a distance threshold value; and finally, iterating until a termination condition is met, and outputting an optimal effective point cloud. The method is high in noise rejection rate, the point cloud fits the sea surface form, and high-quality data support can be provided for sea surface fitting, sea wave simulation and unmanned ship control.

The invention discloses a robustness tracking filtering method for group targets, and belongs to the field of radar and signal processing. The implementation method comprises the following steps: establishing a kinetic model of a group target and a three-coordinate ground-based radar detection probability and measurement model; bayesian recursion of a multi-target state random finite set is realized by a Poisson random finite set, and an intensity function of the Poisson random finite set is described in a Gaussian mixture form. Estimating the maximum motion distance of the target between two adjacent filtering steps based on the dynamic characteristics of the group target, and achieving the detection of the target; the detection probability is described by using Bernoulli distribution, under the condition that the clutter intensity does not exceed the real target intensity, the prior clutter intensity is taken as a threshold value, the detection probability is adaptively adjusted in the filtering iteration process, the respective weights of the measurement irrelevant part and the measurement relevant part are updated, and the strong-robustness multi-target tracking under the condition of sensor leak detection is realized. The method also has the advantages of strong robustness, high tracking precision and small calculation amount.

The invention belongs to the technical field of hidden danger detection, and more specifically relates to a power transmission hidden danger filtering method and system fusing a visual model and spatial logic correction, and a medium. The method comprises the following steps: hidden danger candidate generation based on a high-performance target detection model: adopting a reference segmentation image dynamic multiplexing strategy to complete background semantic analysis of a power transmission channel inspection image, and generating a final reference segmentation image as a semantic mask; performing spatial mapping and logic correction on hidden danger candidate boxes and semantic masks; performing confidence weighting and soft filtering on boundary cases; identifying the boundary cases, constructing a three-dimensional confidence fusion model to calculate fusion confidence, realizing soft filtering according to the fusion confidence, and outputting a high-credibility hidden danger alarm list. The problems that in the prior art, due to the fact that a background semantic base is unstable and total image re-segmentation is frequently triggered, computing resources are wasted, scenes violating common sense cannot be recognized, and the uncertainty of a boundary area is lack of fine-grained description are solved.

The invention discloses an abnormal data real-time filtering method and system based on edge computing in a dynamic environment system, the method is executed by an edge computing gateway, sliding window weighted average preprocessing is equivalently realized by adopting integer shift operation, and the single computing overhead is controlled within 50 clock cycles; dynamically and adaptively adjusting dynamic reference model parameters based on the ratio of the network load to the sensor sampling frequency; performing multi-stage anomaly filtering of a hard threshold value, a mutation rate and a statistical interval on the data through a three-stage pipeline judgment structure executed by atomization; an FPGA hardware queue manager independent of a main processor bypasses a TCP stack to push abnormal data at the highest priority, and redundant data is stored in a zero-copy annular buffer area managed by DMA. According to the invention, the problems of intranet congestion, server I / O bottleneck and alarm delay under the centralized architecture of the traditional dynamic loop system are solved, and the fault, fault, fault, fault, fault and fault are realized on resource-limited platforms such as Cortex-M4 and the like; the average alarm delay is 8 milliseconds; and the data compression rate is more than 90%.

The present invention relates to the field of real-time data filtering technology, and in particular to a real-time data filtering method and system with multi-dimensional feature fusion. The method comprises the following steps: obtaining data transmission logs and extracting original transmission data features, parsing data frame field structure information; identifying multi-dimensional abnormal coupling of data transmission based on field time gradient change rate and dynamic offset status; further detecting abnormal dilution of transmission data, predicting the data transmission complexity exceeding limit state, and detecting the overload condition of data filtering chip structure; then evaluating the attenuation trend of data real-time filtering efficiency, and optimizing the data real-time filtering path accordingly, completing real-time filtering processing of data transmission complexity exceeding limit condition, and outputting filtering results; the present invention achieves higher credibility of data filtering by filtering data in real time.

The invention provides a low-altitude target monitoring, positioning and tracking method and system based on a civil network, and the method comprises the steps: enabling a plurality of terminal devices at different positions to point to and align at a low-altitude target, collecting target measurement data, and uploading the target measurement data to a background server through the civil network; performing abnormal point detection and elimination by adopting a Z-score detection and dynamic time window filtering method, judging a target batch and a target type according to uploaded data, states and time intervals, and storing the target batch and the target type in a database in an original data form; performing fusion processing on a plurality of pieces of uploaded data through a trajectory fusion algorithm to form a target trajectory, and monitoring, positioning and tracking a low-altitude target; and the target track is sent to a front-end server for target fusion result display, and track information of all the low-altitude targets is completely displayed. According to the invention, data are collected through the terminal equipment, low-altitude targets are monitored, positioned and tracked based on a civil network, low-altitude, slow and small targets are detected and found in time, the detection efficiency is improved, and the system equipment cost is reduced.

A traffic filtering method includes a network edge node that provides a cloud service that receives target traffic; obtains a filtering rule, where the filtering rule is for filtering, based on a filtering action, target traffic that meets a filtering condition; invokes a rule engine to parse and execute the filtering rule, where the rule engine is deployed in the network edge node or in a network edge processing system connected to the network edge node; and obtains filtered target traffic based on an execution result, where the filtered target traffic is traffic that meets a filtering requirement corresponding to the filtering rule. The target traffic filtered according to the filtering rule includes at least one of traffic sent by a network side to a user terminal and traffic sent by the user terminal to the network side.

The invention discloses a method for identifying interference signals of weld defects under different lifts-off conditions, which comprises the following steps of: S1, de-trending processing: carrying out de-trending processing on original detection signals; s2, Gaussian wavelet transform: carrying out Gaussian wavelet transform on the detrended signal; s3, optimal wavelet basis selection: by calculating correlation coefficients or energy ratios of different wavelet basis and defect signals, selecting the wavelet basis with the highest matching degree for reconstruction; s4, envelope processing: extracting a signal envelope based on Hilbert transform; s5, mean filtering: applying sliding window mean filtering to the envelope signal; and S6, threshold processing: setting a self-adaptive threshold screening signal, and retaining the feature points of which the amplitudes exceed the threshold. According to the method, de-trending and Gaussian wavelet transform are used for de-noising enhancement. According to the method, wavelet functions of different orders are constructed and matched with defects, secondary signal enhancement is carried out by selecting a filtering method, finally, threshold stripping interference is calculated, reliable defect identification is carried out, and technical support is provided for uneven welding seam quality monitoring.

The invention discloses a big data real-time calculation abnormal value filtering method and a terminal. The method comprises the following steps: acquiring scene basic data, establishing a relevance rule of master data and slave data in the scene basic data, determining upper and lower limit parameters of the master data, the slave data and the scene basic data according to data attributes, and calculating a basic data value of the scene basic data; determining a preset range of the scene basic data by using machine learning in combination with upper and lower limit parameters of the master and slave data and the scene basic data; if the basic data value of the scene basic data is within the preset range, calculating the fluctuation range and the fluctuation value, and if the fluctuation value exceeds the fluctuation range, determining that the scene is abnormal; if the fluctuation value does not exceed the limit, the deviation proportion is calculated, and if the deviation proportion exceeds a preset abnormity judgment coefficient, abnormity is judged; if the deviation proportion does not exceed the limit, finally judging the deviation proportion through an abnormal rule in a preset decision tree; abnormal data are eliminated, filtered data and basic data values thereof are obtained, and the adaptability and reliability of abnormal value detection are improved.

The invention relates to the technical field of point cloud data processing, and provides a ground filtering method and system based on cloth simulation parameter dynamic optimization, and the method comprises the following steps: obtaining point cloud data of a to-be-processed region, carrying out the preprocessing, constructing a complexity index, and calculating a complexity index value based on the preprocessed data; clustering the obtained complexity indexes by adopting a clustering method fusing an elbow rule and a spatial neighborhood constraint to obtain a terrain complexity category; and selecting distribution parameters of a distribution algorithm based on the terrain complexity category and the dynamic mapping of the constructed complexity category-optimal parameter group, and further performing ground point filtering and ground extraction by adopting the distribution algorithm. According to the invention, distribution simulation parameter configuration is guided through terrain complexity analysis, accurate adaptation of different terrain areas is realized, and the accuracy and global consistency of ground point extraction are improved.

The invention relates to a PID (Proportion Integration Differentiation) drawing element intelligent identification and topology reconstruction method based on visual detection, which comprises the following steps of: cooperatively extracting multi-modal information, detecting components based on improved YOLOv11, identifying all text labels in a graph by adopting PaddleOCR, and carrying out pipeline identification algorithm and merging filtering method based on probability Hough transform of pixel points. Identifying and defining the T-shaped connection point as a special topological node to obtain positioning information of components and characters in a PID drawing, associating the components with the characters by using a regular expression and an Euclidean distance, and converting image elements into a topological graph model with a semantic relationship; according to the method, the end-to-end automation process of detection-association-reconstruction is achieved, finally, display is conducted in a graphical interface mode, a data basis is provided for subsequent application such as drawing analysis, system simulation and equipment management, end-to-end automatic conversion is achieved, and efficiency is greatly improved.

Context-aware content filtering adapted for a knowledge domain is provided. In certain examples, a classification confidence score by a classifier indicates a level of confidence that a prompt from a user is associated with the knowledge domain. The classification confidence score is compared with a threshold. When the score is below a threshold, a violation notice is provided to the user without submitting the prompt to a generative artificial intelligence (GAI) model. When the classification confidence score is above the threshold, the prompt is further processed to determine, according to rules, whether the prompt should be submitted to the GAI model. In various examples, the rules are applied to contextual information, safety score information, and intent information derived from the prompt.

The invention discloses a dynamic self-triggering filtering method under incomplete semi-Markov kernel information, relates to the technical field of automatic control and signal processing, and aims to solve the problems that full acquisition of semi-Markov kernel information is difficult, communication burden is heavy and peak control is insufficient. A self-triggering mechanism containing a dynamic variable is designed to determine a data transmission moment by classifying a known / unknown set of semi-Markov information and utilizing known parameters and an unknown lower bound, a modal-dependent filter is constructed, and the mean square index stability and robust performance of a filtering error system are verified in combination with the Lyapunov theory and a linear matrix inequality technology. The problems that an existing method is high in conservative property and wastes resources are solved.

The invention provides an improved density-based iterative voxel downsampling method, and more flexible and adaptive point cloud downsampling is realized by introducing point cloud density information and iterative voxel size. According to a traditional voxel filtering method, a fixed voxel size is used, important details can not be effectively reserved in an area with large density change, and excessive downsampling can be caused in a sparse area. According to the method, the point cloud number of each voxel is calculated, and the voxels are dynamically segmented and subdivided according to the preset threshold value, so that more point cloud information is kept in a high-density region, redundant data is reduced in a low-density region, and the downsampling precision and the calculation efficiency are improved. Specifically, when the number of points in a voxel exceeds a set threshold value and the size of the voxel is larger than the minimum size, the algorithm divides the voxel into eight sub-voxels, and recursive subdivision continues until conditions are met. The method is high in adaptability, can dynamically adjust the sampling strategy according to the density change of the point cloud, and remarkably improves the precision and efficiency of point cloud processing. Compared with a traditional voxel filtering method, the method has the advantages that details of a dense region can be better reserved, redundant points of a sparse region are reduced, and an efficient and accurate point cloud downsampling solution is provided.

The invention relates to the technical field of intelligent traffic, in particular to a road traffic incident intelligent identification and filtering method and system, and the method comprises the steps: collecting video image data, radar point cloud data and environment sensing data, and generating the track information of a traffic target; performing space-time alignment and feature fusion on the trajectory information and the radar point cloud data to form multi-modal features; the method comprises the following steps: preliminarily identifying a traffic abnormal event based on multi-modal feature analysis, and generating event report information comprising an event image and associated features; performing deep semantic understanding on the event report information, analyzing the event image, the track information and the environment sensing data according to the associated features, and determining the semantic category of the traffic abnormal event; executing intelligent filtering processing based on the semantic category; and recording the traffic abnormality subjected to intelligent filtering processing as an effective traffic event, and generating a traffic recognition result. According to the method, deep semantic understanding is carried out through rich multi-modal features, and the overall accuracy and processing efficiency of traffic incident recognition are improved.

The invention discloses a semiconductor heterojunction interface thermal resistance prediction method based on machine learning, and the method comprises the steps: collecting semiconductor material data, obtaining semiconductor intrinsic attribute data through a semiconductor public database, and carrying out the preprocessing of the data; calculating statistics of element attributes through a Magpie algorithm to obtain feature descriptors, eliminating redundant features by adopting a variance filtering method and recursive feature elimination, and normalizing the redundant features; model building and training are carried out by designing CNN and XGBoost algorithms; performing parameter optimization on the trained model through forward propagation, back propagation and a DBO algorithm; and predicting the thermal resistance of the semiconductor heterojunction interface based on the optimized model. The method solves the problems that in an existing semiconductor heterojunction interface thermal resistance prediction method, the experimental measurement period is long, the cost is high, environmental parameters are difficult to control accurately, the theoretical calculation complexity is high, and the deviation between a prediction result and an actual working condition is large due to the dependence on ideal interface conditions.

The invention provides a satellite-borne single-photon laser radar point cloud self-supervised filtering method and system, and relates to the technical field of laser radar data processing, and the filtering method comprises the steps: carrying out the point cloud self-supervised filtering through a core hypothesis that a source point cloud and a target point cloud which are randomly split from a same coarse filtering point cloud slice should have similar distribution; a completely self-supervised deep learning filtering framework is constructed, and efficient denoising of the satellite-borne single photon point cloud can be realized without manual annotation or external prior knowledge. The statistical consistency of the point cloud is used as a supervision signal, and the limitation that structural noise with a specific mode cannot be effectively filtered out through a traditional threshold value method is overcome. Distribution characteristics of the point cloud are automatically learned through a deep learning model, coordinate correction is carried out, full-automatic processing is achieved, and the processing efficiency of mass satellite-borne point cloud data is improved; according to the method, the fine filtering point cloud after coordinate correction is directly output, and terrain details can be better kept.

The invention discloses a fusion filtering method, system, equipment and medium for power transformer winding state monitoring, and belongs to the technical field of power equipment state monitoring, and the method comprises the steps: obtaining an original state monitoring signal of a transformer winding; performing multi-scale decomposition on the original state monitoring signal by adopting self-adaptive wavelet transform; an improved particle filter is embedded, and a suggested distribution function of relative entropy optimization and a dynamic weight updating mechanism are combined; the filtered sub-signals are reconstructed through inverse wavelet transformation; weight parameters are adjusted in a self-adaptive mode; and outputting the key state parameters for transformer winding health assessment. Under the complex electromagnetic environment and the dynamic load fluctuation working condition, the signal-to-noise ratio of the state data of the transformer winding can be remarkably improved, the comprehensive accuracy rate of state evaluation is improved, and reliable data support is provided for health management and preventive maintenance of the transformer winding. And the operation reliability of the transformer and the intelligent level of state evaluation can be obviously improved.

The invention provides an FPGA-based millimeter-wave radar data parallel pipeline filtering method, device and equipment and a storage medium, and the method comprises the steps: receiving echo data of a millimeter-wave radar in real time, and alternately writing the continuous echo data into a first buffer region and a second buffer region through a double-buffer structure; the method comprises the following steps: receiving full frame data through a three-line queue, and dividing a currently processed scanning line in the three-line queue into n blocks; performing parallel filtering on the jth sampling point in any block and the front scanning line and the rear scanning line at the corresponding position, and generating filtered data; and calculating a self-adaptive threshold based on the statistical characteristics of each block, comparing the self-adaptive threshold with the filtered data, suppressing noise signals lower than the self-adaptive threshold according to a comparison result, and selecting a result with the highest energy as final output by comparing signal energy of corresponding positions of the three lines. The problem that resource occupation is too high when echo signals of the millimeter wave radar are processed on an existing FPGA is solved.

The invention discloses a swimming skeleton point coordinate data denoising method and device based on spatio-temporal topological structure learning, and belongs to the field of data processing. The method comprises the following steps: acquiring and splicing a swimming video of a target object; detecting a target object in the video, and obtaining skeleton point coordinates and corresponding confidence scores of the target object by adopting a posture estimation model; preprocessing the coordinates to obtain feature vectors; adjacency information of each skeleton point is obtained based on the human body skeleton topology, the feature vectors of all the skeleton points are input into a spatial domain noise removal network, the adjacency information of the skeleton points is aggregated, multi-scale pooling is carried out, and denoised spatial domain features are obtained; sampling the high-dimensional spatial-temporal characteristics by adopting a deformable time convolutional network; and carrying out deformable convolution operation and decoding operation on the sampling features to obtain skeleton point coordinates after time-space domain denoising. Linear and nonlinear noise in data is processed in a time-space domain in a cooperative manner, so that the limitation of a traditional low-pass filter method and an existing ST-GCN method is effectively overcome.

The present application relates to the field of machine vision. Provided are a non-target of interest filtering method and apparatus, and a device and a storage medium. The method comprises: acquiring a learning sample image, which is an image captured with respect to a target region, wherein the learning sample image comprises a target of interest; on the basis of the learning sample image, determining pixel filtering parameters, which are used for representing the imaging sizes of the target of interest at different positions of the target region; and on the basis of the pixel filtering parameters, filtering out a non-target of interest from the target region in an image to be processed. The method is applicable to a target detection process, and is used for reducing labor costs incurred by manual calibration for filtering parameters when a non-target of interest is filtered out.

The invention provides a thermal elastic structure joint topological optimization method based on MATLAB and ABAQUS, and belongs to the technical field of thermal elastic structure optimization design. Firstly, finite element models of mechanical loads and thermal loads are established in ABAQUS, and files are exported; then calling an ABAQUS operation model in MATLAB, extracting core data such as nodes, units and a stiffness matrix, and constructing a parameterized thermodynamic finite element model based on an SIMP method; calculating a function value and sensitivity by taking volume minimization as a target and taking global stress and temperature constraint as conditions; and finally, iteratively optimizing the design variables by adopting a moving asymptote algorithm, and obtaining a final topological configuration through a filtering method. Through deep cooperation of MATLAB and ABAQUS, core finite element data of commercial software is extracted at a time for optimization iteration, repeated data generation and interaction are avoided, the calculation efficiency is greatly improved, and an efficient solution is provided for processing the topological optimization problem of stress and temperature constraints under complex heat-force loads.

The invention provides a zero-trust gateway encrypted traffic non-inductive filtering method and device based on an eBPF multi-level mapping architecture and a computer program product. The method comprises the steps that a data packet is intercepted in a network driving layer through an XDP mode of an eBPF, and quintuple and encryption protocol visible meta information are extracted; performing hierarchical strategy matching on the meta-information by adopting a multi-stage mapping architecture, and dynamically optimizing a search path through a performance model; and executing real-time verification and access decisions based on a zero-trust policy engine to realize non-perceptual filtering of encrypted traffic. According to the method, the problems of weak detection capability and high matching delay of a traditional scheme under encrypted traffic are solved, the method has the advantages of high speed, low delay and high precision, and meanwhile, data privacy and compliance are guaranteed.

The invention discloses a yellow storm terrorism content identification and filtering method and system, and relates to the technical field of digital content auditing, and the method comprises the steps: carrying out the modal decomposition and preprocessing of a to-be-detected content, including the calculation of a single-modal sensitive probability and confidence, and recording the coordinates and confidence of a local image region; calculating a dynamic weight and performing normalization processing based on the confidence and the prior credibility of each mode; calculating a final fusion score by combining the sensitive probability and the weight of each mode, and performing content auditing decision according to a preset threshold value; and the modal priori credibility is updated through manual feedback data. According to the method, the detection rate of cross-modal violation contents is remarkably improved through a multi-modal collaborative analysis and confidence-driven dynamic fusion technology; and the system adopts a three-level decision-making mechanism to automatically allocate auditing resources, so that the manual rechecking cost is greatly reduced, meanwhile, an online learning function is introduced to continuously optimize the model, the sensitivity to novel violation contents is kept, and the processing efficiency and detail identification are both considered.