142results about How to "Improve generalization" patented technology

The invention discloses a pilot dangerous driving behavior monitoring method based on multi-modal data fusion, and belongs to the field of multi-modal data fusion monitoring, and the method comprises the steps: obtaining the physiological data of a pilot in the take-off and landing states through a multi-element sensor, judging the fatigue state of the pilot, and obtaining the dangerous driving behavior of the pilot. An eye tracker and a camera are used for conducting data collection on a pilot in the take-off and landing states, behavior data of the blinking frequency, the eyelid closure degree, the pupil change, the yawning frequency, the nodding frequency, the average moving speed of an operating lever and the moving direction of the pilot are obtained, and the fatigue state of the pilot is judged; the method comprises the following steps: constructing a multi-modal data fusion network based on Transform, and training the multi-modal data fusion network; the driving behavior of a pilot is monitored by using the trained network based on the Transform multi-modal data fusion; through a multi-channel interaction technology, the comprehensiveness and accuracy of information are improved, the influence of environment and individual differences is reduced, and the stability and robustness are enhanced.

The invention discloses a seismic source inversion data driven rock burst prediction method and system, and the method comprises the steps: carrying out the quantitative inversion calculation of the core parameters of a coal rock fracture seismic source mechanism through original sound wave data collected on site; a rock burst seismic source data prediction model is constructed based on an Informer algorithm, and the occurrence time and position of rock burst are accurately predicted. A rock burst risk evaluation model is constructed by combining the predicted occurrence time and position of a future fracture source and seismic source mechanism core parameters, the magnitude and property of rock burst are predicted, and a dual-drive rock burst intelligent model for seismic source data prediction and rock burst risk evaluation is established; and multi-dimensional intelligent prediction of sequential response, spatial distribution and evolution attributes in the rock burst catastrophe process is realized. According to the method, seismic source data prediction and risk evaluation are combined, the rock burst catastrophe risk can be predicted, and the rock burst evolution process and the disaster-causing mechanism can be visually represented.

The invention belongs to the field of zero sample target detection, and discloses a deep learning-oriented zero sample target detection method and system. The method comprises the following steps: establishing a zero sample target detection model comprising a feature extraction module, a cross-modal alignment module and a course learning module; extracting image features and text features through a feature extraction module; inputting the image features and the text features into a cross-modal alignment module; carrying out cross-modal similarity calculation on image feature embedding and text feature embedding; the obtained cross-modal similarity is imported into a course learning module; fusion of image features and text features is enhanced through a course learning module, and the difficulty of course learning is dynamically adjusted through knowledge distillation in the back propagation process; and obtaining a zero sample target detection model and a detection result. According to the method, the advantages of high performance and high generalization ability of a zero sample learning large model are reserved, mutual complementation and enhancement among cross-modal features are realized, and the detection performance and robustness of the model are improved.

The invention discloses an in-situ hyperspectral water quality parameter inversion method based on deep learning. According to the method, a parallel network model is constructed by combining a convolutional neural network (CNN) and a Transformer, local features are extracted by using the CNN, global information is captured by using the Transformer, and comprehensive analysis of complex features and waveband information in in-situ hyperspectral reflectivity data is realized. In the model training process, the initial layer of the network is pre-trained by adopting a self-supervised learning strategy of random shielding and noise injection, and then the weight of the network is finely adjusted by utilizing labeled data, so that the understanding of the model on a data feature mode is enhanced. According to the method, the unlabeled data is fully utilized, and the training precision and generalization ability of the model are improved. Compared with a traditional explicit linear or nonlinear model constructed by depending on prior knowledge, the method has the advantages that the recessive correlation between spectrums can be captured, end-to-end water quality parameter inversion is realized, and the result is remarkably improved in the aspects of precision and generalization ability.

The invention discloses a mechanical equipment lightweight intelligent fault diagnosis method based on a Kolmogorov-Arnold network, and the method comprises the steps: firstly obtaining original vibration monitoring data of mechanical equipment under different fault types, and constructing an equipment vibration monitoring fault data set; a KAN-based fault diagnosis model is constructed, and a learnable nonlinear function is adopted to replace a fixed linear weight in convolution; then, the output is converted into a one-dimensional feature vector through KAN convolution operation; inputting the one-dimensional feature vector into a KANLinear layer so as to realize the mapping of the high-dimensional feature to different fault types; then calculating training loss, taking the total loss as an optimization target of a training stage, and updating model parameters by using a gradient descent method; iteratively optimizing the fault diagnosis model; and finally, testing the fault diagnosis model by using the test set. According to the method, high diagnosis precision can be maintained, meanwhile, the model parameter quantity is remarkably reduced, the calculation cost is reduced, and therefore the overall calculation efficiency is improved.

The invention belongs to the field of tailing pond safety risk identification and accident analysis, and discloses a tailing dam accident path identification method based on a mixed causal knowledge graph and a related device, and the method comprises the steps: building a mixed causal logic model through employing an obtained tailing dam accident initial hazard source; solving a set of accident cause links of the tailing dam by utilizing the mixed causal logic model; in the set of the tailing dam accident cause links, the logic relation of the tailing dam accident and the accident cause links are used as nodes and edges of a knowledge graph respectively, and a Python + Neo4j method is used for constructing a tailing dam accident knowledge graph; and coding and classifying the tailing dam accident knowledge graph by using a BERT-GCN-SVM model to realize tailing dam accident path identification. The mine tailing dam accident path identification method is improved and optimized based on the thought of the knowledge graph, a set of mine tailing dam accident chain identification method is formed, and the mine tailing dam accident path identification method can be used as a new tool for mine tailing dam accident directional control and disaster situation deduction.

The invention provides a pipeline defect assessment method and system based on multi-source data fusion, and the method comprises the following steps: collecting the data of the operation state of a pipeline body through a related collection device, and marking the state and fault type of the pipeline, so as to construct a training set and a test set; and preprocessing the static data of the pipeline, wherein the preprocessing comprises data cleaning, standardization and feature extraction. And optimizing the extreme learning machine by using a bird flock algorithm, analyzing static data and evaluating pipeline defects. Preprocessing the image data of the pipeline body, wherein the preprocessing comprises image conversion, enhancement and denoising; and evaluating the circumferential weld defect based on the pre-trained VGG16 model. And performing decision-level fusion on a defect evaluation conclusion obtained by static data analysis and circumferential weld image analysis by using a D-S evidence reasoning algorithm to obtain a final evaluation result. The method effectively improves the accuracy and reliability of pipeline defect evaluation, and has wide application prospects and important practical value.

The invention relates to the technical field of data anomaly detection, and particularly discloses a multivariate time series anomaly detection method, system and product, and the method comprises the steps: building an anomaly detection framework which comprises a domain attention-based feature extractor, a domain attention-based feature predictor and a multi-domain anomaly discriminator which are connected in sequence; training the domain attention-based feature extractor and the domain attention-based feature predictor by using multivariate time series data of a source domain and a target domain to obtain a multivariate time series anomaly detection model; and performing multivariate time sequence anomaly detection by using the multivariate time sequence anomaly detection model. According to the method, a target domain with limited data and a plurality of source domains with rich data are aligned in distribution, so that cross reuse of data among different devices in a device group is realized; and the model is trained through a joint loss function constrained by domain attention to capture more comprehensive equipment state information.

The invention relates to the technical field of photovoltaic prediction of a power system, in particular to a photovoltaic power short-term prediction method, device and equipment based on multi-level reverse error correction and a medium, and the method comprises the steps: obtaining original data, and carrying out the data interpolation and data selection; variational mode decomposition is used to decompose the selected data, and kernel principal component analysis is carried out to reduce dimensions; dividing the data after dimension reduction into a plurality of sub-data sets with the same size and a test set as input data, selecting a plurality of training algorithms, constructing a prediction model based on model stacking Stacking, and inputting the input data into the prediction model for training; performing prediction by using the trained prediction model, and evaluating a prediction result; according to the method, historical data or environmental factors are analyzed, photovoltaic stations with similar characteristics or prediction results are clustered, photovoltaic stations of different clusters are predicted according to the clustering results, and a prediction model is subjected to reverse correction.

The invention relates to the technical field of computers, and provides a locomotive driver fatigue state detection method, which comprises the following steps of: S1, establishing a user image sample data set of a locomotive driver fatigue state and marking; s2, inputting a generative adversarial network model, and training a generator and a discriminator of the generative adversarial network model to perform data expansion; s3, after data expansion, inputting the data into a feature extraction model, a feature dimension reduction model and a classifier model for model training to obtain a trained user image sample model; and S4, based on the trained user image sample model, monitoring the fatigue level of the locomotive driver in real time, and giving out safety measures. According to the scheme, the training data set is effectively expanded based on the generative adversarial network, the training efficiency of the feature extraction model is remarkably improved, the convergence speed of the training process is accelerated, the model development period is shortened, the data reconstruction quality is improved through optimization of the feature dimension reduction model, and the classification precision is improved through improvement of the classifier.

The present invention relates to a scissor door hinge, in which a fifth axis is formed by cooperating with a multi-link support structure and a constraint shaft seat, so that a hinge arm is hinged to the constraint shaft seat through the multi-link support structure, and 5 axes are co-pointed, and the intersection point is located on an opening hinge. By arranging the fifth rotation axis at an end of the hinge arm away from the height shaft seat, the hinge arm is supported while adding a constraint force to the hinge arm, thereby avoiding the hinge arm from swinging, increasing the structural stability, and preventing the door from shaking when opening. When the door is opened or closed, the movement of the door in three directions is simplified to the coordinated movement of the five rotation axes, thereby increasing the stability of the door when opened, and realizing intelligent electric opening and closing of the scissor door. The whole movement process is simple and has little effect on the structure of the door, thereby retaining the aesthetics and practicality of the scissor door, and improving the smoothness and stability of the door movement. The door hinge has a simple structure, is easy to install, and has a low manufacturing cost, and is convenient for mass production and generalization of the hinge.

The invention discloses a scene text recognition method and system for multi-granularity training based on global vectors, and relates to the technical field of scene text recognition. The method comprises the following steps: acquiring a to-be-recognized image, and cutting the to-be-recognized image to obtain a part containing a text; the scene text recognition model is used for carrying out scene recognition on the part containing the text to obtain a recognition result, and training is carried out on the scene text recognition model stage by stage and comprises the steps that training samples are preprocessed, the preprocessed samples are encoded through an encoder, global vectors containing multi-granularity information are extracted, and the global vectors containing the multi-granularity information are obtained; and decoding the global vector by using a decoder to obtain an identification result. According to the method, the information of different granularities of the image can be better utilized, the information of the image can be fully learned from easy to difficult, and a better effect can also be achieved for complex texts and curved texts.

The invention discloses an underwater acoustic time-varying channel prediction method based on kernel recursive least squares, and relates to the technical field of ocean communication, and the method comprises the steps: obtaining channel sample data, and constructing a gain training sample and a time delay training sample; and constructing a multipath gain prediction model and a multipath time delay prediction model of any path p by using a kernel recursion least square method, and training corresponding prediction models. And utilizing the trained corresponding prediction model to predict and obtain a multipath gain prediction value and a multipath time delay prediction value of the prediction moment T according to the historical sample sequences of a plurality of continuous moments before the prediction moment T. And integrating the multipath gain prediction values and the multipath time delay prediction values of all paths at the prediction time T to obtain a channel prediction value of the underwater sound time-varying channel at the prediction time T. According to the method, real-time tracking prediction can be carried out on the underwater sound time-varying channel in a complex and changeable underwater environment, and performance improvement of an underwater sound communication system is facilitated.

The invention belongs to the field of log anomaly detection of a distributed system, and provides a log anomaly detection method based on Transform and time convolution, and the method comprises the following steps: S1, obtaining log data in a real environment; s2, log preprocessing: preprocessing the obtained log data set to obtain first log content; s3, dividing the log sequence: selecting a proper division strategy for the features of the first log content to obtain a second log content; s4, performing vectorization processing on the second log content to obtain a log vector fused with time, risk types and semantic features; s5, constructing a dual-channel anomaly detection model; and S6, traversing the log data set to train a model, and inputting the log vector into the trained anomaly detection model to perform log anomaly detection. According to the method, the problems of poor generalization ability and information loss in traditional log analysis are avoided, and the accuracy and efficiency of log anomaly detection are improved.

The invention discloses a method for predicting the concentration of dissolved gas in transformer oil based on an EVO-VMD-LSSVM. An energy valley optimization algorithm EVO, variable mode decomposition VMD and a least square support vector machine LSSVM are combined and used. The method comprises the following steps: decomposing an original dissolved gas concentration signal into a plurality of intrinsic mode functions IMFs with different frequency characteristics through a VMD method, and extracting multi-scale characteristics of the signal; a plurality of factors such as transformer load, operating oil temperature, environment temperature and operating time are introduced into the model, an LSSVM prediction model is established for IMF in combination with characteristic parameter correlation, LSSVM model parameters are optimized by using an EVO algorithm, and prediction precision is improved; reconstructing the prediction result of each IMF to obtain a final dissolved gas concentration prediction value; the prediction performance is obviously better than that of a traditional method, and evaluation indexes such as mean square error (MSE), mean absolute error (MAE) and decision coefficient (R) show that the method has high prediction precision and robustness.

The invention discloses a multi-agent hyperspectral band selection method based on a fusion state and a superior experience pool, and belongs to the field of remote sensing image processing. After a hyperspectral image is obtained, a sample set of the hyperspectral image is generated, and the sample set is divided into a training sample set and a test sample set; a multi-agent waveband selection model is constructed according to the training sample set, the state representation method comprises meta description statistics, an auto-encoder and a graph convolutional network, and the reward allocation strategy comprises an adaptive allocation strategy based on waveband importance; inputting the training sample set into a multi-agent waveband selection model for model training to obtain a trained multi-agent feature selection model, and optimizing a memory sample of an experience pool for model training by adopting a specific sample oriented experience pool optimization method based on a Gaussian mixture model; and inputting the test sample set into the trained multi-agent feature selection model for band selection, outputting a hyperspectral image band set, and classifying the hyperspectral image band set to obtain a classification result.

The invention discloses a double-backbone network insulator defect detection method based on YOLOv5. The method comprises the following steps: constructing a GELAN-Ghost double-backbone network as a new backbone network of a YOLOv5 model to extract feature information of an input image; a plug-and-play high-efficiency multi-scale attention module iEMA is designed; based on an attention module iEMA, specifically adding an iEMA module in a small target detection layer and a medium target detection layer of the model, and constructing an iFPN-PAN structure; according to the method, a novel DynamicHeadv3 detection head is constructed, and a deformable convolution and attention mechanism is utilized to extract richer features and enhance the perception ability of a model; and 4) taking a YOLOv5 model as a reference, introducing the GELAN-Ghost double-backbone network proposed in the step 1), the step 2) and the step 4), an iEMA attention mechanism and a DynamicHeadv3 detection head to improve the YOLOv5 model, and constructing the double-backbone network insulator defect detection method based on the YOLOv5.

The invention discloses a training method of a small sample image detection model and an image detection method, and belongs to the technical field of deep learning, and the base class training method comprises the steps: training an original image detection model through employing a support image set, a marked query image, and each target class in a class until the total loss is converged, obtaining a base class image detection model; support features and semantics of a target category are embedded and input into a variational auto-encoder, and compared with support features and reconstruction features of a specific category, potential vectors in a VAE intermediate process contain general features of more types of prototypes; potential vectors containing semantic information are sampled from the distribution, and feature aggregation is carried out on the potential vectors and query features, so that the estimated new class center is closer to a real class center; in addition, the reconstruction loss, the KL divergence, the classification loss, the regression loss and the semantic consistency loss in the VAE form the total loss, so that the training precision can be greatly improved.

The invention relates to an event-driven tactile sensing method and system based on a pulse graph neural network, and belongs to the technical field of tactile sensing and robot control. The method comprises the following steps: S1, extracting main features from data collected by an event-driven touch sensor; s2, encoding the main features extracted from the tactile sensor data; s3, using a spiking neural network to further learn feature embedding; s4, the encoded features are processed through a graph neural network; and S5, identifying the category or state of the target object by adopting a voting mechanism. According to the scheme, through the graph neural network, the graph based on the space structure of the tactile sensing unit can be automatically constructed, and local and global features are efficiently extracted; the pulse neural network is used for processing event-driven data, so that the energy consumption is remarkably reduced and the calculation efficiency is improved; the method has strong generalization ability, can adapt to different tactile sensor configurations and application scenes, and has excellent performance when processing real-time tactile perception tasks.

The invention discloses a palmprint recognition method and device, a storage medium and electronic equipment, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining at least one palm image sample of each known user in a plurality of known users; determining a palmprint area image and a first palmprint curve image of the corresponding palmprint according to each palm image sample; according to the palmprint area image and the first palmprint curve image, a feature extraction model is trained to obtain a palmprint recognition model, the feature extraction model comprises a first feature extraction module and a second feature extraction module which have the same network structure and share parameters, the first feature extraction module is used for extracting features of the palmprint area image, and the second feature extraction module is used for extracting features of the palmprint curve image; the second feature extraction module is used for extracting features of the first palmprint curve image; palmprint recognition is performed based on the palmprint recognition model, so that data enhancement training of the palmprint recognition model can be realized, and the palmprint recognition accuracy of the model is improved.

The invention discloses a wafer appearance detection method and device and computer equipment. The method comprises the following steps: labeling defect categories of defect samples of various wafer appearances to obtain a sample set with defect labeling categories of various wafer appearances, and training the sample set with defect labeling categories of various wafer appearances to generate a defect detection model of wafer appearances, and detecting wafer appearance defects according to the defect detection model of the wafer appearance. By means of the mode, the detection efficiency and accuracy of wafer appearance defect detection can be improved.

The invention discloses an optical glass chemical strengthening process prediction method, system, equipment and medium, and relates to the technical field of process simulation, and the method comprises the following steps: obtaining original experiment data of a target optical glass drop; calculating an estimated refractive index of the tempered glass according to the molten salt ion concentration ratio, and adding the estimated refractive index into the process parameter set # imgabs0 # to obtain a current process parameter set # imgabs1 #; based on a decision tree algorithm, performing regression modeling according to the current process parameter set # imgabs2 # and the performance parameter set # imgabs3 # to obtain an initial decision tree model; data in the current process parameter set # imgabs4 # and data in the performance parameter set # imgabs5 # are preprocessed, and model training data are obtained; training and testing the initial decision tree model according to the model training data to obtain a process performance prediction model; and calling the process performance prediction model. The method has the effect of reducing the design cost of the glass chemical strengthening process.

The invention relates to a coherent optical communication auxiliary fiber channel modeling method and system and electronic equipment, and the modeling method comprises the steps: obtaining a multi-span transmission signal sequence, and constructing an initial neural network model; training the initial neural network model by using the multi-span transmission signal sequence to obtain an optimized neural network model; importing the optimized neural network model, and cascading the linear physical model and an amplifier to obtain a single-span optical fiber link channel model; and iteratively transmitting a signal based on the single-span optical fiber link channel model according to a preset transmission scene to obtain a multi-span transmission result. According to the invention, optical communication transmission simulation modeling with low complexity and high prediction precision can be realized at the same time.

The invention provides a building height estimation method and device fusing a multi-view building roof contour, and relates to the technical field of building height estimation, and the method comprises the steps: carrying out the roof contour extraction of a satellite remote sensing image, and obtaining the building roof extraction results of different views; carrying out Hu invariant pitch matching on a building roof extraction result; determining unmatched roof targets in the building roof extraction result, and constructing an energy function based on the offsets corresponding to the multiple pairs of matched roof targets so as to determine the offsets corresponding to the unmatched roof targets; the roof target with the adhesion condition is corrected; and estimating height information corresponding to each building based on the offset corresponding to each roof target. The method can effectively reduce the labor cost required for estimating the height of the building, remarkably improves the reliability of the height of the building, and has better generalization capability.

The invention discloses a relay protection plug-in automatic simulation test method and system based on a universal backboard. The method comprises the steps that a relay protection plug-in to be tested is connected to the universal backboard; semantic information in the relay protection configuration file of the plug-in is analyzed and extracted, a corresponding test data model is dynamically constructed, and a matched operation scheme is generated; on the basis of the operation scheme, real-time simulation parameters are set, and a multi-dimensional excitation signal is applied to the plug-in through the universal backboard so as to simulate various operation conditions of the power system; and acquiring response data of the plug-in under the excitation signal through the universal backboard, and comparing the response data with an expected response to obtain a test result. According to the invention, plug-in level universal testing is realized, special debugging instruments of various models can be replaced by a single testing system, and the operation and maintenance efficiency and the spare part management level of the relay protection device are remarkably improved.

The invention discloses a multi-modal clustering method based on a hypergraph auto-encoder. Firstly, a modal data set is obtained, a hypergraph is formed based on the modal data set, and an adjacent matrix is obtained; then, encoder output and decoder output of each modal data set are obtained through an auto-encoder network; thirdly, fusing the potential representation corresponding to each modal data into a multi-modal data potential representation to obtain a corresponding similarity matrix; in conclusion, the network parameters and the multi-modal potential representation of the auto-encoder network are updated until the loss function converges, and finally, the multi-modal data are divided into k categories according to the updated multi-modal potential representation. According to the method, the potential similarity between the samples is learned through the structure priori of the data, and the effective structure priori can reduce the deviation of the sample characteristics. Characteristic prior and structure prior are integrated into a model by utilizing the characteristic representation capability of a pre-training model and the excellent modeling capability of a hypergraph. Therefore, the quality of the self-supervised learning algorithm is improved.

The invention discloses an interval particle neighborhood rough set three-way decision-making-based classification method, which comprises the following steps of 1, acquiring a data set, and converting the data set into a tetrad decision-making information system; 2, obtaining an interval particle set of the decision information system; 3, obtaining a threshold pair through a three-branch approximation method of the shadow set, and obtaining the average ambiguity of the interval grain shadow set according to the threshold pair; 4, obtaining a threshold pair corresponding to the minimum fuzzy loss between the average ambiguity of the interval grain set and the average ambiguity of the interval grain shadow set, and taking the threshold pair as an optimal threshold pair; and 5, according to the optimal threshold pair, identifying an interval particle neighborhood of the prediction object, and classifying the prediction object. According to the method, higher-quality decision information system processing and expression are achieved, and the accuracy and robustness of the model are further enhanced.

The application relates to the technical field of microbial agent, and discloses a plant root growth-promoting microbial agent screening method, which comprises the following steps: firstly, collecting rhizosphere microbial group data and corresponding crop growth phenotype data under different target crops, growth stages and soil environments; then, establishing a microbe-phenotype prediction model based on the data, obtaining two groups of key growth-promoting strains through characteristic importance analysis and microbial interaction network analysis; and finally, screening a microbial agent formula with the optimal expected growth-promoting effect according to the key strains and the prediction model. The application establishes a systematic plant root growth-promoting microbial agent screening scheme, can screen out efficient and synergistic dominant strain combinations by comprehensively considering the growth-promoting effects and interaction relationships of the strains, significantly improves the pertinence and effectiveness of the plant root growth-promoting microbial agent, and has a good application prospect in the field of microbial fertilizer development.

This invention provides a vehicle simulation speed correction method and system, relating to the field of vehicle intelligent dynamics modeling technology. The correction method includes the following steps: S1: Input and process simulation data output from the vehicle dynamics simulation model and corresponding real vehicle test data; S2: Construct a dynamic graph structure representing the interaction between state variables based on a preset physical coupling relationship of the vehicle powertrain; S3: Input the simulation data into a GCN and perform graph convolution operations under the constraints of the dynamic graph structure to extract graph embedding feature sequences representing the spatial dependencies between state variables; S4: Input the graph embedding feature sequences into a TCN and perform temporal convolution operations to learn the dynamic evolution law of state variables in the time dimension and output the correction amount of the vehicle simulation speed; S5: Output the result. Based on this, this invention solves the problem that existing correction methods have various limitations in practical applications.

The present application relates to a kind of integrated force arm structure, double rear axle suspension system and its vehicle, it includes: first force arm, the first force arm has first body, and from the first body towards one side extends to form two first arms spaced apart, each the first arm is rotatably connected with pin shaft, two the pin shaft on the first arm coaxially arranged and interval distribution;Second force arm, the second force arm has second body, and from the second body towards the side close to the first force arm extends to form two second arms spaced apart, two the second arm is arranged with two the first arm one by one, and each the second arm is rotatably connected to the pin shaft corresponding with the first arm;The first body extends to form first axle connecting arm towards away from the second body direction, the second body extends to form second axle connecting arm towards away from the first body direction.Two force arms integrated connection, simple structure, realize lightweight design.