64 results about "Weak model" patented technology

The invention discloses an industrial defect detection method based on self-supervised fine tuning, and solves the problems of scarcity of industrial scene defect samples and weak model generalization ability. The method comprises the steps that a data set is divided and preprocessed, and the data robustness is improved through size scaling, random luminosity transformation, geometric enhancement and the like; extracting a foreground mask by using a saliency model, synthesizing a Perlin Noise and DTD texture fused pseudo-abnormal image, carrying out self-supervised fine tuning on the ImageNet pre-trained WideResNet-50, and enhancing the industrial data feature extraction capability; a model containing a visual trunk, feature aggregation mapping, noise feature adaptation and a discriminator is established, local neighborhood features are fused through Unfold operation, Gaussian noise is superposed to generate pseudo-abnormal features, and an abnormal score is output by the discriminator after multi-scale fusion. And the training adopts binary cross entropy and focus loss optimization parameters. The innovation points of the method are that self-supervised fine tuning adapts to industrial data distribution, feature aggregation improves fine-grained detection, and multi-scale fusion considers different defects.

The invention discloses a water supply system scheduling optimization method and device, electronic equipment and a storage medium, and relates to the field of intelligent scheduling of water supply systems. The method comprises the steps of collecting and preprocessing multi-dimensional data to obtain a data set; the pressure data is analyzed from the space-time dimension, and the unfavorable points and the pressure requirements thereof are accurately identified; the method comprises the following steps: constructing a water volume prediction model by adopting a time sequence model, and constructing a total water head difference prediction model by adopting a Light GBM gradient boosting tree in combination with a MultiOutputRegressor multi-output regression framework; constructing a minimum total water production cost objective function based on a prediction result, and outputting an optimal scheduling scheme by combining water volume and pressure constraint iterative optimization; and establishing a model updating mechanism to ensure dynamic adaptation of the strategy. According to the method, the problems of insufficient pressure guarantee, extensive cost control and weak model practicability and generalization ability are solved, the inherent contradiction that a traditional mechanism model is high in complexity and a pure data driving pressure prediction model is poor in generalization and lacks physical significance is overcome, and safe, stable and efficient intelligent technical support is provided for a water supply system.

The invention provides a geological environment monitoring method and system, and the method comprises the steps: constructing an air-space-earth-depth four-dimensional cooperative monitoring network, collecting multi-modal geological environment data, carrying out the preprocessing of the data, and obtaining multi-modal feature data and abnormal data points after dynamic load balance distribution of calculation resources; fusing the data to generate a comprehensive geological environment feature map containing space-time correlation features, abnormal hot spot distribution and a geologic body three-dimensional reconstruction model; and based on the map, performing geological risk prediction by using a geological risk prediction model to obtain a prediction probability. A monitoring network is constructed to integrate multi-modal geological data, resource allocation is optimized by combining edge calculation dynamic load balancing, geological risk prediction is realized by using a prediction model, the problems of single data, poor dynamic adaptability and weak model generalization ability of a traditional method are solved, the prediction precision is improved, the response time is shortened, and the prediction efficiency is improved. And a high-risk area is visually displayed through a three-dimensional risk thermodynamic diagram.

The invention discloses a fault-tolerant reconstruction method for sensing faults of a spatial dexterous hand, and relates to the field of robot control. In order to solve the defects that in the prior art, a dexterous hand on-orbit sensor is weak in fault-tolerant capability, insufficient in model generalization capability and incapable of being adaptive to complex space environment changes, the technical scheme provided by the invention comprises the following steps: constructing a conditional variation auto-encoder model, inputting a joint position and a rope length sequence, and constructing a model; time sequence features are extracted, and a reconstructed joint position mean value and a prediction variance are output; the reconstructed joint position mean value is compared with actual joint data, a dynamic confidence interval is generated, and whether the data breaks down or not is judged; for the data judged to be normal, information entropy is calculated based on the prediction variance, and data samples with the information entropy exceeding a set threshold value are screened and written into a dynamic buffer area; and when the buffer data is fully loaded, incremental learning is triggered, and model parameters are updated by utilizing the buffer data. The method is suitable for real-time detection and fault-tolerant reconstruction of sensor faults in the long-term on-orbit operation process of the space dexterous hand.

The invention relates to the technical field of shield tunnel construction intelligent control, in particular to a shield tunneling machine tunneling speed intelligent prediction method based on multi-algorithm collaborative optimization. The method comprises the following steps: firstly, acquiring operation parameters and geological and environmental parameters of the shield tunneling machine in real time through a multi-source sensor and an industrial bus; carrying out localized cleaning, normalization and feature extraction on the data by utilizing an edge computing device; further, integrating particle swarm optimization, a genetic algorithm, a sparrow search algorithm and a starvation game search algorithm, and realizing global automatic optimization of the hyper-parameters of the bidirectional long-short-term memory network through a parallel independent optimization and result aggregation strategy; and finally, predicting the tunneling speed in real time by using the optimized model. According to the method, the problems of low prediction precision, weak model generalization capability, dependence on manpower on hyper-parameter optimization and the like caused by insufficient multi-source heterogeneous data processing capability are effectively solved, the prediction accuracy, the adaptability and the engineering practical value are improved, and reliable support is provided for safe and efficient propulsion of shield construction.

The invention discloses a deep learning-based rock hardness intelligent identification model and method, a DF (data fusion)-improved CNN (Convolutional Neural Network) model provided by the invention still keeps 97.92% stable accuracy under continuous cutting and different working conditions, and the problems of tedious artificial feature extraction engineering and weak model generalization in traditional rock hardness identification are solved. And the problems of insufficient characterization capability and weak model generalization performance of a traditional method are solved. Different from a feature screening process dominated by expert experience in a traditional mode, the method provided by the invention realizes feature adaptive extraction through a multi-channel convolutional neural network, carries out overlapped sampling data enhancement on an original vibration signal, constructs a time-frequency entropy multi-domain fusion graph through short-time Fourier transform, and obtains a time-frequency entropy multi-domain fusion graph; experimental verification shows that the classification accuracy of the multi-channel structure is greatly improved compared with that of a single-channel model, and time-frequency entropy multi-domain fusion is more accurate than that of a single-domain recognition model.

The invention relates to the technical field of aerospace manufacturing and intelligent welding control, and solves the technical problems of insufficient single-mode perception, limited quality prediction precision, lagging parameter regulation and control and weak model adaptive capacity in traditional welding process monitoring. In particular to a spaceflight structure friction stir welding multi-modal sensing and intelligent quality control method, which combines multi-modal real-time sensing and reinforcement learning by introducing a body intelligent body thought, so that a system can autonomously understand a welding state, predict a quality trend and actively adjust process parameters in a welding process; and adaptive optimization control with quality as constraint is realized. According to the method, the limitation of traditional off-line modeling and manual adjustment is broken through, an intelligent control framework with self-learning and self-evolution capacity is provided for spaceflight-level friction stir welding, and a new technical approach is provided for achieving high-consistency and high-reliability welding seams. The intelligent level of friction stir welding of the spaceflight structural part and the welding seam consistency level are remarkably improved.

The invention discloses a dynamic assembly line object grabbing method and system based on multi-modal information fusion, and belongs to the technical field of industrial robot grabbing control, and the method comprises the steps: synchronously collecting the visual information of a dynamic assembly line, the joint motion information of a robot and the touch information of a dexterous finger end, and obtaining a standardized multi-source information data set; constructing a coarse grabbing action generation sub-model and a fine grabbing action generation sub-model; the coarse grabbing action generation sub-model obtains a coarse grabbing action sequence through a de-noising diffusion implicit model; according to the coarse grabbing action sequence and the touch information, a fine grabbing action sequence is obtained through a fine grabbing action generation sub-model; adopting a self-supervised learning mode to train the two models in stages; and based on the trained model, a final fine grabbing action sequence is generated, and the robot is driven to execute dynamic grabbing. The core problems that in the prior art, the approaching precision is insufficient, the grabbing stability is poor, the model adaptability is weak, and a control closed loop is lacked are effectively solved.

The embodiment of the invention provides a big data index management method and system based on a weak model, and the method comprises the steps: selecting a data source based on a business scene, writing an SQL statement, recognizing dimension information and index information based on the SQL statement, and writing the dimension information and the index information into an index table, correspondingly writing the SQL statement and the index table into a distributed column database; receiving a query request of the user-defined index, querying SQL statements based on parameter key information of the query request in combination with basic information of the user-defined index, assembling the SQL statements, generating user-defined SQL statements, and performing statement optimization on the user-defined SQL statements; index codes are created according to the dimension information and the index information, detail indexes are generated based on the index codes, synchronization parameters of the detail indexes are set, and the detail indexes comprise basic indexes, derivative indexes and composite indexes.

The invention relates to the field of genome prediction, and particularly discloses a GEP-GANGP-based genome prediction system and method, and the method comprises the steps: generating synthetic phenotype data by using GEP-cGAN; processing real and synthetic data through a feature extraction network; calculating interaction characteristics of the genotype and the environment by adopting a cross-modal attention mechanism; a transfer learning module is used for realizing knowledge transfer and feature enhancement; synchronously outputting a phenotype prediction value, an effect weight and an interpretability characteristic through a multi-task decoder; the system comprises a GEP-cGAN module, a biological multi-head attention mechanism BMAM module, a cross-modal interaction feature extraction module, a transfer learning TrG2P module and a multi-task decoder module. According to the method, the problems of low prediction precision, poor generalization ability and weak model interpretability in a small sample scene in forest tree breeding are solved, the accuracy and practicability of genome prediction are remarkably improved, and reliable technical support is provided for forest tree genetic improvement.

The invention relates to a model based on multi-modal cross attention and uncertainty integral gradient and application, and relates to the technical field of bioinformatics. The model is obtained based on multi-modal cross attention and uncertainty integral gradient construction, the problems that in an existing drug sensitivity prediction method, multi-modal data fusion is difficult, the model generalization ability is weak, and interpretability is lacked can be solved, the model captures correlation between modals through a cross attention mechanism, and the prediction accuracy of the drug sensitivity is improved. According to the method, uncertainty estimation is realized by combining Monte Carlo Dropout (MC Dropout), feature importance is analyzed by using an integral gradient algorithm, and finally, unification of high-precision prediction and biological interpretability is realized.

The invention relates to the technical field of highway engineering, in particular to a method for predicting the dynamic resilience modulus of coarse-grained soil based on static index and dynamic modulus conversion, which specifically comprises the following steps of: testing various different types of roadbed soil samples to be tested; obtaining dynamic and static rebound modulus values of each subgrade soil sample under the conditions of optimal moisture content and set compaction degree and a relation between the dynamic rebound modulus value and CBR; establishing a dynamic and static modulus conversion equation; performing linearization processing on the dynamic and static modulus conversion equation, constructing a matrix equation based on a linearization processing result, solving the matrix equation by adopting a least square method to obtain regression constants k1, k2, k3 and k4, and substituting the regression constants into the dynamic and static modulus conversion equation; and constructing an equation by using the coarse-grained soil dynamic rebound modulus and the CBR value, and solving an equation regression constant to obtain a coarse-grained soil dynamic rebound modulus prediction formula. The problems that in the prior art, an existing conversion relation mechanism is insufficient in consideration, sample representativeness is limited, and the model generalization ability is weak are solved.

The invention provides a construction method and application of a well site real-time data pre-training model, and belongs to the crossing field of petroleum engineering and artificial intelligence. Carrying out discrete marking; the method comprises the following steps: constructing a self-supervised time sequence model based on a Transform architecture, and fusing position coding and an embedding mechanism; designing a self-supervised task, and setting multiple task targets such as mask prediction, comparative learning, sequence judgment and future state prediction; performing end-to-end training on the model through a joint loss function, and extracting deep time sequence features; deploying the pre-training model in an actual well site environment, and executing tasks such as anomaly detection, working condition recognition and trend prediction; and carrying out model optimization and iteration. The invention further provides application of the method in the field of intelligent monitoring of oil and gas drilling engineering. According to the method, the problems of high dependence on manual labeling, weak model generalization ability, insufficient real-time adaptability and the like of a traditional method are solved, and the method has good engineering deployment and cross-well-site popularization value.

The invention discloses an urban territorial space dynamic monitoring and intelligent analysis system and method, and relates to the technical field of territorial space dynamic monitoring. The system comprises an initialization module used for constructing a lightweight learning framework, a dynamic modeling module used for establishing a dynamic network of geographical environment factors and a monitoring target based on parameters and a characteristic mode set output by the initialization module, and an optimization feedback module used for establishing a monitoring result feedback channel, and comparing the actual monitoring data set with a model prediction data set output based on the dynamic modeling module to calculate a prediction error. Through cooperative work of the three, the problem of weak model generalization ability caused by spatial heterogeneity of geographic phenomena is solved, and efficient and accurate monitoring and analysis of urban territorial spaces in different regions are realized.

The invention discloses an unmanned aerial vehicle interference detection method based on a wide-deep attention residual network and random perturbation enhancement, and belongs to the technical field of unmanned aerial vehicle communication security. The method comprises the following steps: firstly, extracting basic features of unmanned aerial vehicle received signals, and expanding feature dimensions through polynomial interaction; and then feature importance is calculated by using a set uniformity index and dynamic classification is carried out. In a network construction stage, a wide-deep fusion architecture is designed: a deep path establishes an independent SE-ResNet sub-network for each feature level, and high-dimensional nonlinear features are extracted by using residual connection and a channel attention mechanism; a wide-layer path directly processes full-amount characteristics and memorizes a low-order linear rule. And finally, fusing and classifying the two paths of features. In the reasoning stage, an enhancement mechanism during testing is introduced, and environmental accidental errors are eliminated by injecting tiny random noise into real-time signals and performing multi-step average prediction. The method effectively solves the problems of insufficient feature physical association mining and weak model generalization ability in the prior art, and significantly improves the accuracy and robustness of unmanned aerial vehicle interference detection in a complex electromagnetic environment.

The invention discloses a machine learning-based pregnant woman preeclampsia subtype classification probability prediction method and system. According to the method, the one-carbon metabolic cycle core metabolite is combined with multi-source clinical indexes for the first time, and cross-modal deep feature fusion is realized by constructing derivative composite features with clear biological significance; in the model level, a double-layer self-adaptive model integration strategy is adopted, firstly, multiple base classifiers are subjected to self-adaptive weighted fusion based on cross validation performance, then further optimization is carried out through stacking generalization, and a final prediction model with high precision and high robustness is formed; besides, the system integrates SHAP interpretability analysis and risk positioning visualization based on historical distribution, not only outputs the classification probability, but also provides visual clinical decision-making assistance. According to the method, the problems of insufficient subtype identification ability, weak model generalization ability and poor result interpretability in the prior art are effectively solved, and reliable technical support is provided for health management and clinical decision-making in the gestation period.

Disclosed are techniques that generate a structural representation of a plurality of documents, the structural representation including a plurality of nodes and a plurality of edges, with the plurality of nodes being representations of the plurality of documents and the plurality of edges representing a feature in common between nodes of the plurality of nodes, with each node holding a vector of confidence values for weak models on a current optimization step and a weighted prediction for each of the weak models, generate a local ensemble model from the structural representation of the plurality of documents combined with the weighted prediction of the weak models, with the generated local ensemble model having a higher predictive power than any weak model individually, and generate a label for each node based on the local ensemble model.

The invention discloses a model training method and a data detection method. The model training method comprises the following steps: determining a model prediction value of a target sample in training samples by using a neural network model; under the condition that the difference between the true value of the target sample and the model prediction value of the target sample is smaller than a preset threshold value, determining a loss function of the neural network according to a mean square error and a logarithmic hyperbolic cosine between the true value and the model prediction value; under the condition that the difference is not smaller than a preset threshold value, determining a loss function of the neural network according to an absolute error and a logarithmic hyperbolic cosine between the true value and the model prediction value; and training the neural network model based on the loss function. According to the method and the device, the technical problems of weak model generalization ability and high over-fitting risk caused by lack of a loss function for considering the robustness and gradient smoothness of the abnormal data when the structured data is processed by a related method are solved.

The invention discloses a rock permeability prediction method, a medium, equipment and a product, and relates to the technical field of permeability prediction.The method comprises the steps that multi-dimensional logging characteristics are obtained; constructing a Pearson's correlation coefficient matrix thermodynamic diagram to quantify the linear coupling strength of the multi-dimensional logging characteristics and the permeability, obtaining the correlation coefficient of the logging characteristics and the permeability of each dimension, screening the logging characteristics from the multi-dimensional logging characteristics according to the correlation coefficient, and constructing a characteristic matrix; building a parallel permeability model and a vertical permeability model based on a table prior data fitting network, training the parallel permeability model by using the feature matrix, splicing the parallel permeability predicted by the trained parallel permeability model as a new feature to the feature matrix, and training the vertical permeability model by using the spliced feature matrix; and predicting the rock permeability by using the trained parallel permeability model and vertical permeability model. The method can solve the problems that an existing method has high requirements for data quantity and quality and is weak in model generalization ability.

The application discloses a hydraulic servo valve opening degree intelligent early warning method based on digital twin fault data enhancement and multi-source data preprocessing, relates to the technical field of metallurgical hydraulic system state monitoring and predictive maintenance, and comprises the following steps: S1, a data fusion and cooperative preprocessing mechanism is established based on field real-time data and digital twin simulation data; S2, a servo valve opening degree error prediction model and an adaptive updating mechanism thereof are established based on an enhanced multi-source data set; S3, a fusion prediction and real-time early warning link is established based on real-time data and multi-source historical trends; and S4, a model optimization and digital twin correction closed loop is established based on early warning results and actual operation feedback. The application effectively solves the problems of fault data scarcity, large field noise, weak model generalization ability and early warning lag, and realizes high-sensitivity and high-credibility intelligent early warning of early-stage faults of a hydraulic servo valve position out-of-tolerance.

The application discloses a retrieval enhancement generation method and system based on knowledge boundary perception, an electronic device and a storage medium, and belongs to the technical field of natural language processing. The method comprises the following steps: generating a high-quality supervised track by using a teacher model, and learning the ability of gap planning and answers by instruction fine-tuning of a weak model; paired samples reflecting overconfidence and over-conservatism are constructed, and a DPO algorithm is used for confidence calibration; gap planning is generated by a student model during actual prediction, and it is accurately determined whether each knowledge point needs retrieval according to cognitive information labels, and accurate retrieval is triggered only for the knowledge points with knowledge gaps. The application can be widely applied to open domain question answering, dialogue systems and knowledge-intensive tasks by explicitly identifying knowledge boundaries, dynamically adjusting thresholds and fine-grained on-demand retrieval, while ensuring the accuracy of answers, significantly reducing the consumption of computing resources and response delay.

The invention discloses a pathogenic microorganism infection immune feature recognition system and method, belongs to the technical field of immune feature recognition, and aims to solve the problems of incomplete feature extraction, weak model generalization ability, poor adaptability to novel pathogenic microorganisms and the like due to the fact that a traditional immune feature recognition technology mostly adopts single-dimensional features or a traditional machine learning model. In order to solve the problems of low recognition accuracy, high false positive rate and difficulty in meeting actual requirements of clinical diagnosis and epidemic situation monitoring in the prior art, the pathogenic microorganism infection immune feature recognition system comprises a data acquisition module, a data preprocessing module, a multi-dimensional immune feature extraction module, a fusion deep learning recognition model module and a result output and verification module, according to the method, through a deep learning architecture fused by Transform, CNN and LSTM, the long-distance dependency relationship, local features and time sequence features among the features are captured at the same time, optimization strategies such as transfer learning and regularization are combined, and the recognition accuracy and generalization ability of the model can be significantly improved.

The application discloses a strong-weak large model cycle fine-tuning training method for fact-rich conversation content generation, and belongs to the technical field of artificial intelligence. The method comprises the following steps: performing one-time cleaning and cold start on an original fact-rich data set by using a strong large model; mixing the cleaned data and the original data according to a probability to form a mixed data set, and performing initial fine-tuning on a weak large model; performing performance evaluation on the fine-tuned model and calculating a performance value; when the performance value exceeds a dynamic threshold, triggering the weak large model to generate a new answer and updating a historical answer queue; performing multi-round training cycle fine-tuning based on the updated data set, and traversing multiple groups of parameter configurations; and finally saving a performance-optimal model. Through the strong-weak model decoupling cooperation and the cycle self-enhancement mechanism, the application effectively reduces the training cost, avoids overfitting, and improves the performance and generalization ability of the model in the fact-rich conversation generation.

A battery state-of-charge prediction method, apparatus and device based on an ensemble learning model, and a medium, the method comprising: constructing a sample data set according to charge and discharge characteristic data and a real state-of-charge of a battery, and dividing a part in the sample data set as a training sample set into a plurality of mutually exclusive subsets; taking one subset as a verification set and the remaining subsets as a training set in turn, performing cross verification training on the base learner, and obtaining a first predicted charge state output by the base learner according to the verification set; training a meta-learner by using the first predicted charge state in combination with the training sample set; through the charge state prediction model constructed by integrating the trained base learner and meta learner, the charge state of the to-be-detected battery is predicted, so that the technical problems of insufficient prediction precision of the charge state of the battery, weak model generalization ability and poor adaptability to complex battery behaviors in related technologies are solved; the precision and stability of battery charge state prediction are remarkably improved, and the generalization ability of the model and the adaptability to the complex battery state are enhanced.

The embodiment of the invention discloses a high school physics law dynamic teaching system, method and device and a medium, and the system calls an API interface of Geo Ge bra, and generates a dynamic model according to input physical parameters; a slider, a check box and a gesture operation control are integrated through the interaction control module; physical quantity values and change curves are displayed in real time through a data visualization module, and theoretical formula derivation results are superposed; and a student operation log is recorded through a learning analysis module, and an error analysis report is generated. Through functional reconstruction and educational transformation of GeoGebra, the GeoGebra is converted into an exploration type teaching platform suitable for the cognitive level of senior high school students, the mathematical modeling ability of the GeoGebra is deeply coupled with physics teaching requirements, a dynamic modeling, interactive control and data visualization integrated teaching system is constructed, and the teaching efficiency is improved. The problems of insufficient abstract concept visualization, low student participation degree, weak model construction capability and the like in traditional physics teaching are solved.

The invention relates to the technical field of data processing, in particular to a multidimensional data analysis method based on machine learning, which comprises the steps of data preprocessing, feature extraction and fusion, model construction and optimization, and data analysis and decision making. During preprocessing, an improved isolated forest algorithm is used for removing abnormal values, a Bayesian network is used for supplementing missing values, minimum-maximum scaling and logarithm transformation are combined with normalization data, in feature extraction fusion, a high-order singular value decomposition tensor is combined with an attention mechanism for weighting fusion components, and during model construction optimization, a DDQN architecture is improved, and the probability that the model is optimized is lowered. Parameters are updated in combination with empirical regression and a strategy gradient algorithm, finally processed data are input, and decision suggestions are generated by using a multi-objective decision and a Pareto frontier analysis method; the objective of the invention is to solve the problems of insufficient processing precision of abnormal values and missing values during preprocessing of multi-dimensional data, incapability of dynamically capturing key information by feature extraction and fusion, weak model generalization ability and difficulty in processing multi-target conflicts.

The application discloses a kind of mine drill multiple operating condition operating state data processing and evaluation method, belong to coal mine equipment condition monitoring and artificial intelligence technical field.Solve the technical problem of coupling of operating condition and state characteristics, data distribution confusion, signal baseline drift, weak model generalization ability and easy misjudgment of evaluation under multiple operating conditions of drill.The technical scheme is: data set is constructed by collecting multidimensional sensor signals;Temporal characteristics are extracted by inflation convolutional encoder, and are divided into operating condition and state latent variables according to physical priori, and operating condition residues are washed out by using gradient reversal layer;Replace operating condition latent variables with all-zero priori statistical prototype, and reconstruct homogeneous stationary signal by double-flow decoder;The model is optimized and trained by using multi-objective joint loss;After inputting the real-time signal homogenization, the final prediction entropy realizes high uncertainty early warning.The application can effectively purify state characteristics, eliminate baseline drift, realize cross-operating condition distribution alignment, and significantly improve the accuracy of drill condition monitoring.

This invention discloses an engineering decision-making method and system based on BIM and a hybrid large-scale model, belonging to the fields of construction engineering management and artificial intelligence technology. The method includes: acquiring engineering data output from the BIM model and time-series data from the construction site, performing fusion and spatiotemporal alignment preprocessing to form a standardized time-series dataset; extracting time-series feature vectors from the dataset using a recurrent neural network module; constructing a decision tree module, using the time-series feature vectors as input, embedding schedule and cost constraint rules with an information gain penalty term during training to generate interpretable decision paths; and dynamically adjusting the decision tree module according to the change type and updating the parameters of the recurrent neural network module when engineering changes are detected. This invention solves the problems of disconnect between BIM data and intelligent decision-making models, weak decision interpretability, and difficulty in adapting models to dynamic engineering changes, improving the response efficiency and decision reliability of engineering management, and achieving precise and intelligent engineering management. This technical solution has been evaluated and certified by the Ministry of Housing and Urban-Rural Development.

The invention discloses a cross-domain event extraction method and system oriented to a low-resource scene, and belongs to the technical field of natural language processing. In order to solve the problems of weak model migration ability and poor generalization performance caused by lack of target domain annotation data, the technical scheme of automatically generating a pseudo annotation for a target domain text, constructing semantic mapping of a source domain and a target domain event type, training a cross-domain event extraction model and performing cross-domain feature alignment is mainly adopted. According to the method, automatic extraction of the event trigger word and the event argument can be realized under the condition that the target domain labeling data is scarce or missing, and the cross-domain migration capability and generalization performance of the model are remarkably improved.

The application discloses a kind of open disaster scene-oriented quick point cloud instance segmentation method, belong to three-dimensional data processing technical field;The method includes: after voxelization to point cloud, feature and central position prediction are extracted via sub-manifold sparse convolution network;Subsequently, both are input to the core of center position auxiliary sparse point cloud decoder;The decoder is through the synergistic effect of self-attention layer, asymmetric attention layer and learnable embedding scoring screening layer, efficiently interacts and sparse decoding to feature and position information, outputs final embedding;Finally, based on the center point of this predicted instance, class and mask.The application effectively utilizes the sparsity of point cloud through a unique decoder structure, solves the key problems that the prior art faces in the open disaster scene, such as weak model generalization ability, slow training convergence and insufficient inference speed, significantly improves the accuracy and efficiency of segmentation.