362results about How to "Reduce overfitting" patented technology

The invention discloses a rolled metal surface defect automatic labeling method based on a multi-task self-adaptive model, which comprises the following steps: S1, collecting a rolled metal surface image, preprocessing and enhancing the rolled metal surface image, performing data labeling, and constructing a rolled metal surface defect data set; s2, constructing a multi-task adaptive model, and training the multi-task adaptive model by using the data set to obtain an automatic defect labeling model; and S3, inputting a to-be-labeled metal surface image into the defect automatic labeling model, and outputting a surface defect labeling result by executing the target detection task. According to the method, a multi-task learning framework is combined, joint training of target detection and instance segmentation tasks of metal surface defects in the rolling technology is achieved, the target detection task is focused in the reasoning stage, the labeling efficiency is improved, the labor cost is remarkably reduced, and the labeling efficiency is improved. The method can be widely applied to industrial defect detection and data automatic labeling scenes of rolling process metal, and is suitable for defect image automatic labeling tasks in a complex industrial environment.

The invention discloses an EEG-fNIRS deep fusion cognitive state classification method based on feature decoupling, and the method comprises the steps: firstly synchronously collecting EEG data and fNIRS data of a tested mental calculation task, and carrying out the preprocessing of the two collected data, so as to obtain EEG data and hemoglobin concentration change data; secondly, electroencephalogram signal data and hemoglobin concentration change data are taken, data enhancement is carried out through a time window, then the data are sent to a deep fusion network, and public features and private features of EEG and primary features of fNIRS are obtained; and an EEG classification result, an fNIRS classification result and an output fusion classification result are obtained. And finally, designing decoupling loss and cross entropy loss based on a label smoothing strategy, and carrying out training. According to the method, the overfitting problem in fNIRS modal deep learning is effectively relieved, and cognitive state classification can be accurately carried out.

The invention belongs to the field of fault diagnosis, and particularly relates to a main shaft bending common characteristic diagnosis system based on a stacking integration method, and the system comprises the following modules: a sensor module which is used for collecting a vibration signal of a main shaft in real time; the data acquisition and processing module is used for converting an analog signal into a digital signal and preprocessing the signal acquired by the sensor; the feature extraction module is used for extracting features from the preprocessed vibration signals; the machine learning diagnosis module analyzes and classifies the extracted features by using stack integration based on a plurality of trained machine learning models, and judges the health state of the main shaft; and the output module is used for outputting the diagnosis result and sending out an alarm signal when abnormality is detected. Real-time monitoring and automatic diagnosis of the main shaft bending fault are achieved, compared with a traditional method depending on expert experience and manual detection, the fault detection efficiency and accuracy can be greatly improved, and misjudgment and missed judgment risks caused by human factors are reduced.

The invention discloses a pig farm disease risk attribution method and system based on a knowledge graph. The method comprises the following steps: constructing a pig farm biosafety knowledge graph according to a pig industry biosafety application and a pig industry biosafety field knowledge base; secondly, inputting all pig farm events of the pig farm in N days, risks of the pig farm and risk probabilities; factors of pig farm diseases are searched through the pig farm biosafety knowledge map, and the disease text attribution of the pig farm is given according to constraints such as the contribution degree of pig farm problems. The pig farms with medium and high risks are rectified and predicted according to attributed risk items, and the pig farms with low risks or the pig farms without diseases after early warning can be summarized as excellent management experience. Through analysis and reasoning of the atlas, more logical analysis can be provided for pig farm risk attribution tasks, the pig farms are helped to deeply understand characteristics and prevention and control requirements of various diseases, and a scientific basis is provided for formulating targeted prevention and control strategies.

The invention relates to the technical field of artificial intelligence, and discloses a network security situation awareness method based on artificial intelligence, which comprises the following steps: acquiring network security data; performing feature extraction on the network security data by adopting a full-connection neural network based on dynamic group evolution optimization to obtain a first feature; performing feature dimension reduction on the first feature by using a self-encoding neural network based on contrast entropy to obtain a second feature; constructing a classifier model by using a fractional order neural network based on complexity perception, and training the classifier model by using the second feature; and performing network security situation awareness classification by using the trained classifier model. According to the method, the capability of distinguishing the complex network security situation is improved, a more accurate classification effect can be realized, and the applicability of the model in the network security data and the classification accuracy are enhanced.

The invention discloses an image denoising method based on double-branch feature fusion, and the method specifically comprises the steps: 1, carrying out the collection of an image, and carrying out the preprocessing of the collected image; 2, a denoising network with double-branch feature fusion is built, the network comprises a generation network and a discrimination network, and authenticity discrimination is carried out on a denoised image generated by the generation network and a real image through the discrimination network; step 3, adversarial training is carried out on the generative network and the discriminant network to obtain a trained dual-branch feature fusion denoising network; and step 4, performing denoising processing on the image by using the trained generative network. The method is remarkably superior to a traditional denoising method in denoising quality, model generalization ability and operation time, different types of noise can be effectively removed, and more image details and structure information are reserved.

The invention relates to the field of chlorophyll a concentration prediction, and discloses a 1D-DSCAM-based subsurface chlorophyll a concentration inversion method, which comprises the following steps of: obtaining sea surface temperature and sea surface chlorophyll a concentration data of satellite remote sensing and subsurface chlorophyll a concentration, subsurface temperature and salinity profile data of a biogeochemical buoy; carrying out preprocessing, constructing a data set, and inputting a training set into a 1D-DSCAM-GRU deep learning model to carry out model training; the model comprises an Embedding module, a 1D-DSCAM module and a multi-layer GRU module; and preprocessing sea surface temperature and sea surface chlorophyll a concentration data obtained by satellite remote sensing of a sea area to be predicted and subsurface layer temperature and salinity profile data obtained by a biogeochemical buoy, and inputting the data into the qualified model for subsurface layer chlorophyll a concentration prediction. According to the method disclosed by the invention, the prediction precision can be improved.

The invention relates to the technical field of lithium batteries, and discloses a lithium battery life decline quantitative prediction and health state intelligent evaluation method, which comprises the following steps: S1, multi-source data acquisition: arranging strain gauge sensors at battery electrode plates and a shell to acquire stress-strain data, integrating a gas sensor in a battery module to acquire gas data, and calculating the data of the stress-strain data; battery data and environment data are acquired by a high-precision synchronous acquisition system through nanosecond clock signals and are processed and stored, and the data quality is monitored at the same time; s2, data preprocessing: generating an extended data set through data transformation and combination and GAN-based data, and processing abnormal data by using a method based on statistics and machine learning; according to the invention, through multi-source data fusion acquisition, a stress-strain sensor and a plurality of gas sensors are added to monitor the internal mechanical change and gas component concentration of the battery, so that the problem that the accuracy and practicability of the model are greatly reduced due to a traditional method based on an electrochemical model is improved.

The invention provides a relation extraction model construction method and system based on semantic enhancement and knowledge distillation, relates to the technical field of relation extraction, and aims to solve the problem that in the prior art, when continuous learning is processed, a model is often difficult to keep memorizing an old relation when a new relation is learned, and forgetting is caused. Model overfitting is easily caused by excessive dependence on a small amount of data, so that a new relationship cannot be effectively generalized. The method comprises the steps of obtaining a relationship prototype of a training sample and a typical sample, performing semantic enhancement on the relationship prototype to obtain an expanded enhanced data set and an enhanced memory set, and performing feature extraction, relationship classification prediction and relationship extraction based on a student model and a multi-teacher model. According to the method, the overfitting problem caused by scarcity of small sample data is effectively relieved through a semantic enhancement technology, old relation forgetting caused by new relation learning in the continuous learning process is reduced through a multi-teacher knowledge distillation method, and the classification accuracy and generalization ability of the model facing different relations are improved.

The invention discloses an urban underground space structure health detection method and detection system, and the method comprises the steps: S1, collecting the multi-modal data of an underground space structure through the deployment of diversified sensors; s2, preprocessing the collected data, wherein the preprocessing comprises filtering denoising and feature extraction; s3, performing pattern recognition and health state evaluation on the processed data by using a deep learning model; s4, dynamically adjusting a threshold value of health assessment according to the environmental parameters, and predicting a future health trend; and S5, the evaluation result is displayed through a visual platform, a corresponding health examination report is generated, the health condition of the underground space structure can be monitored in real time by integrating multiple sensors and utilizing a data processing technology, potential problems can be found in time, an early warning threshold value can be dynamically adjusted according to real-time monitoring data through a dynamic threshold value adjusting mechanism, and the early warning efficiency is improved. The problems of false alarm and missing alarm caused by a fixed threshold value are avoided, and the accuracy and reliability of early warning are improved. The method is suitable for various underground space structures.

The invention provides a ViT distillation training method, system and device and a readable storage medium, and the method comprises the steps: selecting a long-tail data set, and carrying out the preprocessing of the long-tail data set; training the convolutional neural network to generate a teacher model; generating a data sample based on the long-tail data set, inputting the data sample into a teacher model, and generating a soft label; constructing a student model based on a visual Transform, and configuring a knowledge distillation function and a loss function for the student model; generating a data sample based on the long-tail data set, inputting the data sample into a student model, generating a classification result, and optimizing student model parameters by using a loss function based on the classification result and the soft label to train the student model; and after training is completed, performing model verification and output on the student model. According to the method, the technologies of long-tail data set preprocessing, teacher network training, knowledge distillation, loss reweighting and the like are combined, so that the classification performance and generalization ability of the ViT model on the long-tail data set are effectively improved.

The invention provides an advanced control system for a direct-current submerged arc furnace. The advanced control system comprises a sensor unit, a data acquisition unit, a central processing unit, a man-machine interaction unit, a model prediction control unit, a numerical control output unit and an execution mechanism unit, the central processing unit sends direct current submerged arc furnace production process data to the model prediction control unit for model prediction control data calculation, and then the model prediction control unit returns calculation result information to the central processing unit; the central processing unit processes the received calculation result information and adjusts a production process control strategy, and a control signal is transmitted to the execution mechanism unit through the numerical control output unit; the method can achieve the soft measurement of the insertion depth of the positive and negative electrodes, achieves the automatic adjustment of the electrodes of the direct-current submerged arc furnace through a generalized prediction control algorithm, stabilizes the production, improves the economic benefits, and improves the overall intelligent level of the direct-current submerged arc furnace.

The invention discloses an oral cancer multi-task diagnosis system based on a Raman spectrum, and belongs to the technical field of biomedical detection. Comprising an acquisition module and an oral cancer multi-task diagnosis module, and the acquisition module is used for acquiring Raman spectrum data; the oral cancer multi-task diagnosis module is used for processing the Raman spectrum data through a trained oral cancer multi-task diagnosis model so as to generate and output a tumor staging result, a lymph node staging result and a histological staging result; the oral cancer multi-task diagnosis model comprises a bidirectional Mama branch and a residual attention fusion branch, local feature information in the Raman spectrum data is extracted by using the residual attention fusion branch, and multi-scale features are generated by fusing global feature information extracted by the bidirectional Mama branch. Feature information in Raman spectrum data can be effectively extracted, the accuracy of oral cancer diagnosis is improved, and multi-task diagnosis of oral cancer is realized; the problems that multi-task common diagnosis cannot be realized and the diagnosis accuracy needs to be improved in the prior art are solved.

The invention discloses a semi-supervised seismic wave impedance inversion method and system based on Swin Transform iterative gradient correction, and belongs to the technical field of earthquakes in geophysical exploration. According to the method, the impedance gradient is iteratively corrected in the model space through the deep neural network, and complex nonlinear prior information is captured; and each anisotropic structure direction regularization item is established, and constraint is carried out along the direction of the stratum structure, so that the space consistency of the prediction result is ensured. Meanwhile, the USTNet network designed by the invention combines the advantages of the local feature capture capability of the convolutional neural network and the global modeling capability of Transform, and reduces the non-uniqueness of solutions. The network is used for inverting synthetic data and actual data, and the effectiveness of the method is verified.

The invention relates to the technical field of battery management, and discloses a lithium battery thermal runaway feature recognition algorithm based on a BP neural network, and the algorithm comprises the following steps: data collection and preprocessing: employing a temperature sensor and a gas sensor to monitor the temperature and gas information of a lithium battery system in real time; the temperature change of the temperature sensor before and after the thermal runaway and the temperature change slope at the moment of the thermal runaway event are used as main characteristic parameters; for a gas sensor, gas concentration change before and after thermal runaway and gas concentration change slope are considered as main characteristic parameters; and taking the collected temperature, gas data and characteristic parameters as a data set. According to the invention, the temperature and gas changes of the lithium battery are monitored in real time through the temperature sensor and the gas sensor, the characteristics of thermal runaway can be quickly identified, and early warning can be timely given out, so that the real-time monitoring capability of the system is improved, and the accuracy and timeliness of thermal runaway prediction are enhanced.

The invention provides a Chinese electronic medical record named entity recognition method based on a multi-head attention mechanism and a pulse neural network (SNN), and the method comprises the steps: obtaining a dynamic word vector through a pre-training BERT model, obtaining a text code through a bidirectional long short-term memory network (BiLSTM) model, and screening the features of the code through the SNN network for a named entity in an electronic medical record; the multi-head attention mechanism calculates the dependency relationship between characters, enhances the relevance to the context, and finally outputs a final sequence through a conditional random field (CRF) module to obtain an entity recognition result. According to the method, the time cost is remarkably reduced, the efficiency and precision of identifying the named entities in the medical text are improved, and a more efficient and accurate technical solution is provided for the field of medical information processing.

The invention relates to the technical field of breach image processing, in particular to a breach rapid situation awareness method and system based on multi-source data. The method comprises the following steps: acquiring dam monitoring station data; performing hydrological sensor network setting on the dam monitoring station data to obtain dam hydrological sensor network data; performing ultrasonic water level change trend analysis on the dam hydrological sensor network data to generate an ultrasonic water level change trend chart; performing breach node sensing on the dam hydrological sensor network data according to the ultrasonic water level change trend chart to generate breach node data and breach prediction node data; and according to the breach node data, macroscopic disaster risk analysis is carried out on the breach area, and a breach area macroscopic disaster risk schematic diagram is generated. According to the invention, through real-time data integration, comprehensive risk assessment, unmanned aerial vehicle monitoring, dynamic prediction model and multi-party cooperative response, the accuracy and response speed of breach rapid situation awareness are improved.

The invention discloses a dam effect quantity prediction method and system, a medium and equipment, and relates to the technical field of safety engineering, dam monitoring data is acquired, influence factors of a dam effect quantity are determined, and effect quantity influence factor samples are constructed; performing dimension reduction processing on the influence factor sample through a principal component analysis method to obtain an influence factor matrix after decoupling and dimension reduction; coupling the long and short-term memory neural network with an artificial fish swarm algorithm, constructing a dam effect quantity prediction model, performing parameter setting on the dam effect quantity prediction model, and optimizing the initial weight and the number of neurons of the long and short-term memory neural network by using the artificial fish swarm algorithm; training the optimized long and short term memory neural network until convergence, and obtaining a trained dam effect quantity prediction model; and performing prediction and anti-normalization processing to obtain an effect quantity prediction result. And the accuracy, the stability and the efficiency of dam effect quantity prediction are improved.

The invention relates to an artificial intelligence model building method and device, equipment and a storage medium. The method comprises the following steps: preprocessing input original data to obtain a privacy protection data set; performing feature extraction through a multi-scale convolutional neural network connected in parallel in the target model to obtain a target multi-level local feature; dividing the multi-level local feature into two parts to obtain a first part of local feature and a second part of local feature, and inputting the second part of local feature into a bidirectional long-short-term memory network and a bidirectional gating loop unit network which are connected in parallel in a target model for feature extraction to obtain a global semantic feature; splicing and weighting the first part of local features and the global semantic features to obtain fusion feature representation; and performing model parameter optimization on the target model based on the fusion feature representation to obtain a trained artificial intelligence model. According to the invention, the training speed and convergence stability of the artificial intelligence model are improved.

The invention relates to an underwater acoustic target recognition system and method based on multi-modal depth feature fusion, and belongs to the field of underwater acoustic target recognition. The method comprises the following steps: firstly, performing preprocessing on an obtained underwater acoustic target original audio and associated metadata, and constructing a multi-modal data set; and then the constructed multi-modal sample is input into an identification model, the model extracts deep representation of each modal through a multi-branch feature coding network, depth alignment and complementary aggregation of different modal features are realized by using a cross-modal cross attention mechanism guided by potential query, and a target identification result is output based on a decision network of mixed experts. According to the method, experimental verification is carried out on two disclosed underwater acoustic data sets, the experimental result verifies the effectiveness of the multi-modal deep fusion and hybrid expert adaptive decision strategy adopted by the method, and through mining the complementary advantages of acoustic features and semantic priori, the multi-modal deep fusion and hybrid expert adaptive decision strategy is obtained. And the robustness and generalization ability of the underwater acoustic target recognition system in the strong-noise and multi-working-condition environment are remarkably improved.

The invention discloses a charger communication test method and system, a storage medium and a program product, and the method comprises the steps: obtaining a state parameter of a to-be-tested charger; collecting a communication message of the charger to be tested, and constructing a message time sequence queue; generating a message interval sequence reflecting the time sequence characteristics of the message; calculating a discrete degree coefficient of a message time sequence according to the message interval sequence; recording record data at an abnormal moment when the communication abnormality occurs, and inputting the record data into a preset state transition matrix to obtain a corresponding relationship; calculating probability distribution of communication abnormity under different state parameters to obtain a correlation model; generating a message sequence for testing based on the association model, wherein the time sequence characteristic of the message sequence for testing is the same as the state parameter and the dispersion degree coefficient corresponding to the high-probability abnormal scene of which the probability is greater than a preset probability threshold in the association model; and sending the test message sequence to the to-be-tested charger to detect response data of the to-be-tested charger. And the accuracy and efficiency of the communication test of the charger are improved.

The invention discloses a pedestrian crossing tolerance time prediction method under a signal control condition. The method comprises the following steps: calculating a risk score of each pedestrian sample in a pedestrian crossing data set through a random survival forest model; carrying out tolerance grouping on the pedestrian samples by using a K-means clustering algorithm in combination with the risk score, and generating a data set of a plurality of tolerance; dividing the data set of each tolerance into a training set and a test set; a Stacking integration model is constructed; performing hyper-parameter tuning on each primary model through a Bayesian hyper-parameter optimization algorithm; inputting the training set into each primary model for training according to five-fold cross validation, and outputting a prediction result of each primary model; integrating the prediction result of each primary model and the true value of the tolerance time into a new training set and a new test set, inputting the new training set and the new test set into the meta-model, and outputting a corresponding tolerance time prediction value; and realizing tolerance time prediction of a single pedestrian through the trained Stacking integrated model. According to the invention, the accuracy of pedestrian crossing tolerance time prediction can be improved.

The invention relates to a lightweight electrical equipment multi-target detection method and system, and the method comprises the steps: collecting multi-angle images of different types of electrical equipment, obtaining an experiment data set, constructing a Ghostnet lightweight network structure, introducing a CBAM attention mechanism and an SPPFCSPC convolution structure to improve a YOLOv8 model, and obtaining a multi-target detection model based on the improved YOLOv8 model. And training, verifying and evaluating the multi-target detection model based on the improved YOLOv8 model to obtain a multi-target detection optimal model for electrical equipment multi-target detection. According to the method and the system, the model complexity and the calculation amount can be reduced while the detection precision is ensured, and an effective solution is provided for multi-target identification detection of lightweight electrical equipment.

The invention discloses a current transformer error online monitoring method and device based on machine learning, and relates to the technical field of power system monitoring. According to the method, on the basis of historical operation data of the current transformer, the time sequence prediction capacity of the Prophet model and the advantages of SVR, RFR and Lasso regression algorithms are fused, seasonal fluctuation is effectively processed, the future trend is captured, prediction of the current value is achieved, the deficiency of primary side current data is made up, the predicted value serves as a reference standard value for online monitoring of the current transformer, and the current transformer online monitoring accuracy is improved. On-line monitoring of the error of the current transformer is realized. Experiments prove that the method can more effectively adapt to different data features and changes, and a more accurate prediction result is provided.

The invention provides an end-to-end automatic driving system and method. The system comprises an environment sensing module, a feature extraction module, a data fusion module, a decision generation model module and a decision analysis module. The environment sensing module comprises a plurality of types of sensors, and each sensor collects environment information data of different modes. The feature extraction module extracts features of each modal environment information data. And the data fusion module fuses the features of the modal environment information data to obtain fused features. The decision generation model module determines a driving decision by using a trained decision generation model based on the fusion features. And the decision analysis module converts the driving decision output by the decision generation model module into a specific vehicle control instruction. The method has higher adaptability to complex and dynamic driving scenes, reduces the cost of data annotation during training, and improves the generalization ability of the model in various driving scenes.

The invention discloses a remote sensing image change detection method based on DCT (Discrete Cosine Transformation) perception. The method comprises the following steps of: extracting and fusing double-time-phase characteristics by utilizing a four-layer double-branch network; the adaptive spectral feature extraction module generates a DCT filter, and extracts and synthesizes frequency domain features; the confidence coefficient weighted fusion device calculates the confidence coefficient through softmax, the confidence coefficient is used for the feature map after normalization in combination with the attention weight, and the image change is detected through the point convolution integral class. According to the method, an innovative network and an attention mechanism are adopted, extraction of detail features of the remote sensing image is enhanced, and the interpretability of the model is improved; the Siamese network structure ensures the consistency of dual-time-phase image feature extraction; the adaptive spectrum module selects a key frequency according to requirements, and optimizes the parameters of a DCT (Discrete Cosine Transform) filter; the confidence coefficient weighted fusion device improves the recognition capability of a change area through dynamic feature selection and weighting.

The invention is suitable for the technical field of medical image segmentation, and provides a semi-supervised segmentation method based on multi-modal teacher-student consistency learning, which comprises the following steps: constructing a teacher-double student model framework which comprises a multi-modal teacher model and two independent single-modal student models with different learning conditions, the two student models use segmentation networks of the same structure to respectively process images of two different modalities; in the data input and training stage, data with labels train a student model through real labels, and data without labels are trained through false labels generated by a teacher model; in the training process, the consistency prediction between the teacher model and the student model and between the two student models is ensured. According to the method, the problem of overfitting caused by insufficient annotation data is effectively relieved. In an actual medical scene with limited annotation data, the method can achieve a reliable segmentation result only by means of fewer labels, so as to relieve the pressure of medical conditions.

The invention relates to the technical field of network traffic, and provides a smart grid Dos attack detection method and system, a medium, equipment and a program. The smart grid Dos attack detection method comprises the following steps: merging three features corresponding to the same network flow data to construct a fusion feature set; taking the fused feature set and the corresponding numerical tag set as input, and respectively adopting a support vector machine and a decision tree to obtain a first output set and a second output set; inputting the first output set, the second output set and the corresponding numerical label set into a classifier to obtain the weight of a support vector machine and the weight of a decision tree so as to construct a multi-level decision model; and constructing a comprehensive multi-dimensional objective function by considering various performance indexes, and optimizing hyper-parameters of the multi-level decision model to obtain an optimized multi-level decision model for Dos attack detection of network traffic. According to the method, the DoS attack behavior in the smart power grid can be quickly identified, and safe operation of the smart power grid is ensured.

The invention provides a small sample image classification method fusing global and adaptive local information, and relates to the technical field of image processing, and the method comprises the steps: S1, obtaining small sample images to establish a data set, and dividing the data set into a support set and a query set; s2, inputting the data set into a global sensing network DCCP-Net for pre-training; s3, inputting the support set and the query set into a task aware network QSA-TAN for meta-training; s4, inputting a pre-training result and a meta-training result into a global-local knowledge collaborative network GLKC for joint training to output probability distribution differences of the small sample images, and classifying the small sample images according to the probability distribution differences to obtain a classification result; according to the method, through collaborative optimization of knowledge in the global data space and the local task space, image characteristics and category information to which the images belong can be better highlighted, generalization of a small sample image classification model is remarkably improved, and therefore the intra-class change problem of the images is effectively solved.

The embodiment of the invention provides a time sequence prediction method and device, equipment and a medium, and the method specifically comprises the steps: decomposing historical time sequence data, so as to obtain k components corresponding to the historical time sequence data; determining an error evaluation index of the ith component corresponding to the historical time sequence data under the jth neural network; selecting the neural network with the minimum error evaluation index from the M neural networks as a target neural network corresponding to the ith component; according to the data of the ith component in the historical time sequence data, performing optimization processing on the hyper-parameter of the target neural network; determining component prediction results respectively corresponding to k components of the to-be-predicted time sequence data by using the trained target neural network; and reconstructing the component prediction results corresponding to the k components to obtain a prediction result corresponding to the to-be-predicted time sequence data. According to the embodiment of the invention, the accuracy and reliability of time sequence prediction can be improved.