57 results about "Feature combination" patented technology

The present application relates to a kind of based on multimodal feature combination image uniqueness anti-counterfeiting identification method, belong to image processing and anti-counterfeiting identification technical field.It includes: constructing anti-counterfeiting carrier, include by random microstructure Random texture area And the readable identification area of associated digital identity information;Registration stage, based on texture feature extraction model and morphological feature extraction model respectively, the deep texture feature of random texture area and morphological structure feature are extracted, the unique feature fingerprint is generated by the multi-modal fusion of the two, and is bound and stored with digital identity information;Verification stage, through terminal equipment acquisition image to be measured, locate random texture area and extract measured texture feature, based on metric learning algorithm, measured texture feature and the unique feature fingerprint in database are mapped to the same feature space for comparison, and output true and false determination.The present application realizes the anti-counterfeiting identification of high security and high identification accuracy by multimodal feature combination and asymmetric registration verification architecture.

The application discloses a kind of underwater vehicle degradation characterization method, belong to underwater vehicle technical field, steps are as follows: the full life cycle data collected by sensor carried on underwater vehicle is collected, data is preprocessed;Feature extraction is carried out to the data after pre-processing;Design a kind of layered cooperation particle swarm optimization algorithm based on fuzzy analytic hierarchy process, the selected feature is obtained by screening the extracted feature, and the final optimal feature combination is obtained;Dynamic correlation mahalanobis distance is used to weight and fuse the selected feature, and the fusion feature is obtained;Using the UMAP dimension reduction method with degradation perception, the degradation process is divided into three stages of early, middle and late according to the full life cycle data, an independent UMAP submodel is trained in each stage, and then the degradation index is fused by transfer learning.This application breaks through the limitation of single signal, realizes multi-component coupling degradation recognition, strengthens early degradation capture, reduces the rate of missed judgment and fault risk, has environmental dynamic adaptability, and adapts to complex deep-sea working conditions.

The present application relates to the technical field of gene mutation prediction, and discloses a gene mutation prediction method generated by multi-mode annotation, and the specific process comprises the following steps: carrying out mutation type annotation on input single-base mutation position information to obtain mutation basic information containing mutation types, then using an ANNOVAR annotation tool, SpliceAI splicing effect prediction software and reference mutation information of a function effect database to carry out multi-dimensional feature annotation, using Bayesian PCA based on the obtained multi-dimensional feature mutation data set to fill in the annotation data, then using an automatic engineering feature list and a separated feature selection list to carry out feature combination and screening, and obtaining a gene mutation prediction score after gradient generation tree algorithm. The present application can be used for predicting all non-synonymous exon mutations, has good performance in classifying rare benign mutations, and can identify a small amount of mutations with high pathogenic probability from a large amount of candidate mutations.

This invention provides a fault location method, cloud server, and system for an in-vehicle infotainment system. It determines feature vectors for various fault characteristics based on fault data, representing the importance of each fault characteristic to each fault type with quantifiable numerical values ​​(correlation values), thus eliminating biases from human judgment. Because the fault diagnosis model can learn the complex patterns between feature combinations and fault rationality, model-based diagnosis can more accurately identify fault types. Since the three-dimensional correlation matrix associates fault types with possible root causes and corresponding original fault characteristics, it can accurately locate the most likely root cause of the fault. Compared to existing fault location methods that require manual inspection of vehicle logs, this invention utilizes a fault diagnosis model and a three-dimensional correlation matrix of fault root causes to complete most of the fault location work, significantly shortening fault diagnosis time, improving fault diagnosis efficiency, and thus ensuring driving safety.

A coding / decoding method and apparatus are disclosed, relating to the multimedia field. The method includes: dividing an 11B7T sequence into four groups, namely 3B2T, 3B2T, 3B2T, and 2B1T; using a 2-bit data pattern, setting feature combinations in specific regions of the 7 symbols mapped to the 11 bits; using these feature combinations as identifiers to indicate the encoding method; and at the decoding end, mapping the 7 symbols to 11 bits according to the encoding method indicated by the feature combinations of the specific regions. This divides a long sequence into multiple short sequences, performs PAM3 coding / decoding on these short sequences, and combines the feature combinations of the specific regions to indicate the encoding method, reducing the complexity of long-sequence PAM3 coding / decoding, thereby reducing the circuit implementation complexity of long-sequence PAM3 coding / decoding and improving PAM3 coding / decoding performance.

The present disclosure provides an information sending method and device, computer equipment and a storage medium, wherein the method comprises: obtaining a plurality of candidate interactive objects participating in a game and each feature value of each candidate interactive object under a target feature combination; determining the interactive influence degree information of the candidate interactive object under the target feature combination based on each feature value of the candidate interactive object under the target feature combination; selecting a target interactive object from each candidate interactive object based on the interactive influence degree information corresponding to each candidate interactive object respectively; and sending prompt information of executing a target interactive behavior to the target interactive object. The embodiments of the present disclosure can accurately select the target interactive object which plays an important role in maintaining the interactive relationship, and then the target interactive object can affect other candidate interactive objects to generate more interactive behaviors through the target interactive behavior after sending the target interactive behavior, thereby effectively improving the utilization efficiency of game resources.

The application discloses a kind of based on the fault identification method of double-fed wind farm sending line of multi-feature combination, comprising the following steps: S1.builds double-end transmission line simulation model, sets different fault conditions and generates simulation fault data, according to simulation fault data, the label mark of fault type is carried out, obtains the voltage and current simulation fault data of double-fed wind turbine;S2.according to voltage simulation fault data and current simulation fault data, extract fault voltage feature vector and fault current feature vector, and form multidimensional fault feature vector;S3.construct the neural network of fault identification, multidimensional fault feature vector is input into neural network, neural network is iteratively trained, until the error of the output of neural network and label tends to be stable;S4.acquire double-fed wind turbine real-time fault data, extract fault voltage and fault current feature vector, form multidimensional fault feature vector input neural network, neural network outputs fault type probability, the greatest fault type probability is determined as fault category.

This invention discloses a multimodal, few-sample sausage fermentation day identification method based on Markov meta-learning. The method includes: simultaneously acquiring visible light images and near-infrared spectra of sausages, extracting color, texture, and spectral features, and fusing them into a multimodal feature set; calculating state transition probabilities based on feature similarity to construct a Markov temporal task chain; training the method using a meta-learning trainer incorporating temporal constraints, and selecting the optimal feature combination through a five-dimensional quantification index and a four-stage dynamic iteration process; pairing the optimal combination with multiple basic models, performing transferability verification on a four-dimensional independent validation set, and selecting the final prediction model based on a comprehensive score; extracting corresponding features from the test sample and inputting them into the model, outputting the predicted fermentation day and confidence level. This invention achieves high-precision, strong-generalization fermentation day identification under few-sample conditions, significantly reducing annotation costs and improving adaptability to industrial scenarios.

The application discloses a feature selection method for traffic emission prediction, comprising the following steps: selecting automatic or manual feature grouping, performing missing completion, anomaly detection and normalization processing on traffic emission input data, and initializing feature weights; calculating intra-group and inter-group feature correlation, and updating feature weights; using a machine learning method, combining SHAP analysis and model inherent feature importance method, iteratively updating feature weights; and screening optimal feature combinations and verifying. Through the above steps, the application balances the key feature selection, data dimension reduction and interpretability problem under the premise of ensuring that the traffic emission prediction accuracy is not reduced, and can achieve high efficiency to achieve the purpose of selecting an optimal feature subset.

The invention discloses a northeast ski field accumulated snow depth inversion method based on optimal feature screening and a constraint stacking neural network, and belongs to the technical field of satellite remote sensing image processing and accumulated snow parameter inversion. According to the method, sentry No.1 SAR data is taken as a core data source, multi-source auxiliary data such as weather station snow depth data, atmosphere reanalysis data, geographic information data, vegetation feature data and time data are fused, and an accumulated snow depth inversion initial feature set is constructed; according to the method, an optimal feature combination highly related to the snow depth is screened out by introducing a Spearman correlation coefficient-based feature screening method (FSSCC), and an integrated learning algorithm cNN-Stacking is constructed on this basis, so that adaptive weighted fusion of inversion results of various basic models is realized, and the snow depth inversion precision and stability are improved. An empirical model is adopted in the ski field ski track area, space fusion is conducted on the empirical model and the inversion result of the natural snow area, and accumulated snow depth distribution covering the whole ski field area with the spatial resolution being 20 m is obtained.

The present application relates to an edge scene generation method, device, equipment, storage medium and program product, and relates to the field of vehicle algorithm testing. The method comprises: obtaining a scene data set of automatic driving algorithm test failure and an initial scene, a target feature combination related to test failure can be extracted from the scene data set, based on the target feature combination and the initial scene, the multi-modal processing capability of a large model can be applied to generate an edge scene meeting the test requirements. This way does not need to generate a model by a large amount of historical data self-training, but extracts key failure feature combinations based on the real failure behavior of the automatic driving algorithm, decouples the extraction and processing of the failure features from the traditional generation model, so as to generate an edge scene meeting the training requirements of the automatic driving algorithm by using the inference capability of the general large model, batch acquisition of the edge scene can be realized in less time and at lower cost, and the rapid iteration requirement of the automatic driving algorithm is met.

The application provides a feature learning method integrating genetic algorithm and meta learning, comprising the following steps: (1) adopting a pre-training word segmentation model to perform word segmentation on input text; (2) using a pre-training word vector to encode the word segmentation result; (3) then adopting a genetic algorithm to perform feature screening; (4) adopting an Att-Bilstm model for sentence-level sequence encoding; (5) next, adopting an optimal feature combination as the input of a meta learning model; (6) adopting a ridge regression classifier. The application provides a method integrating genetic algorithm feature extraction, a small sample learning framework and a ridge regression classifier; the genetic algorithm is adopted to reduce the dimension of input text, the small sample learning framework improves the rapid learning ability between multiple tasks, and the ridge regression classifier avoids the disadvantage of insufficient generalization ability.

The application discloses an LLM-enabled multi-agent intermittent demand prediction method and system, which comprises a data monitoring stage capable of converting an input original intermittent demand sequence into a high-quality data set; a feature extraction stage for converting the high-quality data set into a feature combination by analyzing data characteristics through recursive reasoning, autonomously planning steps and strategies of feature engineering, and coordinating feature calculation and fusion tools; a pattern classification stage for converting the feature combination into a demand pattern category according to the feature structure matching a clustering algorithm, and then generating readable labels based on the clustering algorithm result statistics characteristics; and a model selection stage for finally outputting a model recommendation scheme matched for each demand pattern by searching a structured model knowledge base through enhanced generation technology, combining data features and demand pattern categories for multi-dimensional deduction. The application can balance the explainability in the intermittent demand prediction scene, and the decision-making process is transparent and the prediction accuracy is high.

The application discloses a processing method and device for an advertisement conversion event. The method comprises: in response to a received conversion event detection request, obtaining generated advertisement conversion data; based on the advertisement conversion data, determining a user behavior feature combination with a support degree greater than or equal to a first threshold in the advertisement conversion event, the support degree representing a ratio between a number of target conversion events containing the user behavior feature combination in the advertisement conversion event and a total number of the advertisement conversion events; determining a confidence of a candidate association mode created based on the user behavior feature combination, wherein the confidence represents a back transmission integrity ratio of the advertisement conversion event; and determining a back transmission anomaly detection result of the advertisement conversion event based on a back transmission confidence distribution of different user value levels determined based on the confidence. The application solves the technical problem of low accuracy in the back transmission anomaly detection process of the advertisement conversion data.

The application discloses a complex equipment health evaluation method based on an improved Mahalanobis-Taguchi system, belongs to the field of health state evaluation of complex equipment, and comprises the following steps: collecting, preprocessing and dividing complex equipment sensor data; constructing a directed acyclic graph structure and a plurality of MTS classifiers by using Mahalanobis distance and a discriminant matrix; screening effective feature parameters by using orthogonal test and signal-to-noise ratio, and constructing a sub MTS classifier of different feature combinations; determining an optimal threshold value and a classification criterion by using a threshold value sliding and a voting mechanism; and collecting sensor signals and evaluating the health state of complex equipment in an unknown state. The application can effectively realize health evaluation of a complex equipment system.

The application discloses a mouse expression recognition method and device based on multi-view feature combination, and belongs to the technical field of animal behavior analysis and computer vision; the method comprises the following steps: selecting a unilateral target recognition model from a preset mouse multi-view face recognition model library according to a preset view angle event set of a target mouse; inputting to-be-recognized events of multiple shooting view angles of the target mouse into corresponding unilateral target recognition models respectively according to the shooting view angles; judging whether the dominant expression categories of the unilateral target recognition models in the same shooting view angle are the same, and judging whether the unilateral dominant expression categories of the shooting view angles are the same according to the proportion of the dominant expression category in the recognition result set and the matching degrees of the unilateral target recognition models; and through pre-construction of the multi-view face recognition model library and combination of the matching degree calculation mechanism, cross-individual mouse expression recognition is realized, it is not necessary to model each new mouse separately, and the recognition accuracy and efficiency are improved through multi-view weighted fusion.

The present application provides a kind of coral reef remote sensing image multimodal feature generation and repair method and system, it is related to remote sensing image repair technical field, including: based on multi-source sensor obtains coral reef image data, carries out pre-processing;To the pre-processed coral reef image data, utilize convolutional neural network to carry out image morphological feature and spectral feature extraction, and establish feature database;Based on feature combination network, weighted fusion is carried out to multi-source feature, generates comprehensive feature expression and is stored in feature database;The image to be repaired coral reef image is matched with feature database, generates guiding parameter;Guiding parameter is input into generative adversarial network, and the missing area of the image to be repaired coral reef image is reconstructed at pixel level.The present application constructs a complete closed loop system, covers multi-source information scheduling, pixel level guided reconstruction and semantic feedback verification, can supplement image information, reasonably reconstruct ecological information, provide high-quality data basis for subsequent classification, monitoring and protection work.

This invention relates to the field of combinatorial modeling technology, specifically to a method and system for impedance prediction using combinatorial modeling. The method includes the following steps: detecting impedance, current, and temperature parameters; comparing impedance measurements with thresholds to generate feature classifications; inputting similar parameter pairs into a Pearson algorithm to construct a correlation matrix; selecting high-scoring feature combinations based on the matrix; optimizing the structure using a K-means model; and assigning values ​​to the parameters with the highest scores after recombination, thus completing the initialization of the prediction sub-model parameters. In this invention, by systematically quantifying the correlation of multi-dimensional feature parameters such as impedance, current, and temperature, the coupling relationships between parameters are revealed. Based on data-driven dynamic recombination of the prediction model structure, key parameter combinations are automatically selected, and the optimal parameter set is used to accurately initialize the prediction sub-model. This eliminates the dependence on fixed equivalent circuit models, enhances model adaptability and prediction reliability, and provides a flexible and solid basis for characterizing the impedance characteristics of complex systems.

The application discloses a cross-well logging feature selection method and related device based on distribution difference perception, and belongs to the technical field of oil and gas well logging; the method identifies distribution difference significant logging features by constructing inter-well domain classification tasks, and weakens inter-well domain deviation influence based on a generalization performance oriented ablation strategy; the contribution of features to prediction results is quantified by using an explainability analysis method, and group intelligent optimization search is introduced to obtain an optimal feature combination balancing between prediction accuracy, feature stability and quantity. The application can effectively improve the cross-well generalization performance of a pore pressure prediction model, reduce feature redundancy, and is suitable for cross-well pore pressure prediction and related logging interpretation tasks.

A kind of motion target classification method, system, device and medium based on time-varying domain three characteristics joint belong to signal classification technical field, solve the problem that the existing classification method using eigenvalue extraction needs to learn a large amount of data in the analysis and processing of these characteristics, and there will be redundant features.This method first receives motion sound source signal, segments with equal time interval;Then Fourier transform is carried out to each small piece of segmented signal to extract line spectrum signal;After that, each small piece of line spectrum signal is transformed to obtain three kinds of characteristic values;Finally, three-dimensional feature space is established using the three characteristic values, and the distribution of two kinds of signals existing motion target signal and only background noise is analyzed to realize the classification of weak target signal and background noise.The present application is suitable for underwater moving target identification classification scene.

The application discloses a DDoS attack visual detection method based on multi-dimensional feature combination analysis, multi-dimensional traffic feature extraction, a lightweight attack detection model, DDoS attack knowledge graph construction, a graph-constrained attack behavior interpretation large model, attack visual detection display and the like to realize visual detection of network attacks. The application improves the visualization level and reduces the false alarm rate. The multi-dimensional feature combination method and the lightweight security detection model provide rich information for subsequent network attack knowledge graph establishment and professional interpretation based on a large language model. The constructed knowledge graph about network attacks can enhance the rigorousness of the thinking deduction of the large model and reduce the occurrence of model hallucinations. The enhanced large model can make more professional explanations and analyses on existing network attacks, thereby reducing the huge workload of manual attack analysis, and striving for a valuable time window for timely and accurate response to network attacks and deployment of defense.

The present disclosure can be applied in the technical field of image processing. Provided is a panoramic depth map generation method. The panoramic depth map generation method comprises: for at least two image groups which are obtained by means of grouping target images corresponding to different orientations in a target scene, performing feature combination on target image feature volumes of each image group, so as to obtain at least two target panoramic feature volumes, wherein the target images included in each image group cover a panoramic field of view of the target scene, and the target image feature volumes represent three-dimensional features of a target image; performing correlation processing on every two target panoramic features in the at least two target panoramic feature volumes, so as to obtain at least one target correlation volume; and on the basis of an initial depth map and the at least one target correlation volume, performing panoramic depth estimation, so as to obtain a target panoramic depth map for the target scene. Further provided in the present disclosure are a model training method, an electronic device and an unmanned aerial vehicle.

The application discloses an AI five-step deterministic reasoning method and system for open problems, and belongs to the technical field of artificial intelligence, rational AGI and high-trust decision-making. The application changes the traditional AI route of "probability generation, text continuation, semantic guessing" for open problems into a deterministic route of five-dimensional problem decomposition, rule library calling, boundary feature combination, axiomatic deterministic reasoning and traceable non-fabricated output. There is no probability, generation, fitting or dynamic learning throughout the process. The reasoning is strictly based on the fact library and the objective law library, which completely eliminates illusions and subjective speculation from the root. The application converts fuzzy open problems into a deterministic process executable by machines, realizes reproducible and verifiable rational decision-making, and has a substantial difference in route from the prior art, is highly creative, and can be widely applied to strong supervision and high-trust scenes such as finance, law and medical treatment.

The application provides a fault case intelligent matching method based on a large language model and Jaccard similarity, comprising: fault information extraction, extracting a service list, an alarm list and an association mapping from an original alarm event; service role abstraction, calling a large language model to infer a specific service name into a predefined service role; alarm type classification, calling a large language model to classify a specific alarm into a predefined alarm type; feature combination set construction, combining the service role and the alarm type into a "{service role.alarm type}" feature element to form a feature combination set; Jaccard similarity calculation, calculating the Jaccard similarity of the feature set of the current fault and the historical cases; result sorting and returning, returning the top N most similar cases after sorting according to the similarity; the application realizes semantic abstraction through LLM, utilizes Jaccard for efficient set operation, and introduces a cache reuse mechanism, so that the real-time performance, interpretability and low cost requirements are considered while the matching accuracy is ensured.

This invention provides a method for predicting pear sugar content based on interpretable multimodal feature fusion. It involves collecting spectral and image data of pear samples to construct a multimodal dataset, then building a multimodal feature-level fusion model based on a convolutional neural network architecture to achieve multimodal feature extraction and fusion, outputting a predicted sugar content value. Next, it calculates and analyzes the intermodal interaction strength to obtain feature combinations of image and spectral features with high interaction strength, and restores the spectral features back to the original input data to identify the key wavelength regions most important for sugar content prediction. This invention effectively integrates visible-near-infrared spectral data and fruit peel image data through a feature-level fusion strategy, overcoming the limitations of a single modality. It also introduces an interpretable method for calculating intermodal interaction strength, quantifying the feature interactions between modalities from a theoretical perspective, revealing the cross-modal correlation mechanism of the model prediction, and improving the accuracy, robustness, and transparency of non-destructive testing of pear sugar content.

Disclosed are an encoding / decoding method and device, relating to the field of multimedia. The method comprise: dividing 11B7T into three groups, i.e., 5B3T, 3B2T, and 3B2T; using a data pattern of 5 bits to set a feature combination in a specific bit field among 7 symbols to which 11 bits are mapped; using the feature combination of the specific bit field as an identifier to indicate an encoding mode; and at a decoding end, mapping 7 symbols into 11 bits on the basis of the encoding mode indicated by the feature combination of the specific bit field. Thus, a long sequence is divided into a plurality of short sequences, PAM3 encoding / decoding is performed on the plurality of short sequences, and an encoding mode is indicated on the basis of a feature combination of a specific bit field, reducing the complexity of PAM3 encoding / decoding for the long sequence, thereby reducing the circuit implementation complexity of PAM3 encoding / decoding of the long sequence and improving the PAM3 encoding / decoding performance.

The application relates to the technical field of medical artificial intelligence and prognosis prediction, and discloses a construction method of a prostate cancer patient prognosis risk prediction model. The method forms a diagnosis and treatment state transition topology by creating a hierarchical knowledge network and mining diagnosis and treatment path time sequence dependence, and accordingly depicts a patient whole-course time sequence data segment. A simulated annealing optimization feature selection cycle is synchronously run, and a deep prediction architecture containing a multi-scale time sequence attention mechanism is dynamically interactively trained, so that online adaptive screening of a feature subset is realized; a feature influence propagation graph is constructed according to the feature selection process, key feature clusters are identified, and guided correction of attention weights of the model is carried out accordingly. The above components are integrated to form a prediction model, and an online adaptive fine-tuning loop is established. The method can effectively optimize feature combination, improve the focusing ability and prediction robustness of the model on key prognosis factors, and support incremental adjustment according to new data.

This invention belongs to the field of lead-acid battery performance testing technology, specifically providing a method, testing method, and system for dividing a lead-acid battery performance testing dataset. The method includes: data acquisition, obtaining a preset number of feature combinations through a traditional random partitioning method; calculating the time-series dependency correlation value, data distribution characteristic value, and non-abnormal noise value of the feature combinations; then calculating the mean of these values ​​as evaluation indicators to obtain the optimal training and test sets; and finally training a neural network model for lead-acid battery SOH prediction and performance testing. This invention evaluates different feature combinations from multiple dimensions, making the resulting target test dataset and target training data more suitable for training the lead-acid battery SOH prediction model, thereby improving the prediction robustness and accuracy of the SOH prediction model.

The application discloses a kind of wetland environmental degradation identification processing methods based on remote sensing image monitoring, method includes: obtaining original remote sensing image including multi-temporal data and pre-processing, through different time nodes different wave band remote sensing image, capture dynamic wetland environmental change, truly reflect wetland condition;Based on texture feature and spectral feature combination image between each region texture boundary transition clear value to the remote sensing image after pre-processing is implemented regional growth segmentation, obtain local area image, feature extraction result is combined by texture feature and spectral feature, so that subsequent image segmentation processing can take into account details and overall structure;Finally, based on local area image by pre-trained wetland degradation detection model is analyzed, obtains the type of each local area image, provides accurate data basis for the detection and remedy of wetland degradation.