56 results about "Gaussian network model" patented technology

A method of generating graphical models for providing security analysis in computer networks that in one embodiment includes the steps of generating a type abstract graph independent of particular networks that models abstract dependency relationships among attributes and exploits; generating network-specific attack graphs by combining the type abstract graph with specific network information; monitoring an intruder alert; and generating a real-time attack graph by correlating the intruder alert with the network-specific attack graph. The real-time attack graph can be generated using reachability checking, bridging, and exploit prediction based on consequence alerts and may further include the step of calculating the likelihood of queries using a Bayesian network model. The method may also include the steps of inferring unobserved attacks that may have been missed by intrusion detection sensors, and projecting on which hosts and using what exploits additional intruder attacks may occur. The method may further include the step of comparing alternate actions by computation, wherein the alternate actions include the step of patching some vulnerabilities, and wherein the specific network information includes network topology. The specific network information may also include firewall rules.

The invention relates to an evaluation model for evaluating the operation risk of the electrical network on the basis of a Bayesian network modeling method, and belongs to the technical field of scheduling automation of the power system. The method comprises that (1) a Bayesian formula based network model is established; and (2) the risk early-warning evaluation index and model are calculated. The evaluation model has the advantages that the electrical network risk early-warning evaluation model is based on the Bayesian network, the Bayesian network model has unique bidirectional inference technology, the operation risk of the electrical network is evaluated, influence of components on the operation risk of the electric network can be analyzed, weak links of the system are obtained, requirements for risk early-warning evaluation and analysis in practical electrical network service are met, and basis is provided for improvement of the system.

The invention discloses a Bayesian network compensation method for thermal error of a numerical control machine tool, which comprises the following steps: (1) a Bayesian network thermal error prediction model is constructed according to measured sample data; and (2) the real-time compensation of the thermal error of the machine tool is realized according to the prediction result of the Bayesian network model. The compensation system of the invention has a simple structure and reliable application; and the adopted Bayesian network modeling method, on one hand, uses the language of a graph theory to intuitively express the causal dependency relation among various factors which produce the thermal error, on the other hand, analyzes and utilizes the inherent correlation among the factors according to the principle of probability theory to reduce the calculation complexity of inferential prediction, and has the characteristics of intuitive expression, high modeling accuracy and self-adaptation.

The invention relates to a Bayesian network model based public transit environment dynamic change forecasting method. The Bayesian network model based public transit environment dynamic change forecasting method comprises the following steps of screening out various factors affecting public transit passenger flow fluctuation or travel time change; abstracting random jamming conditions of exterior environments and passenger flow or travelling time decision variables into nodes of a Bayesian network, determining a station set and the value range of the station set, and performing discretization on the historical information data of the station set and the value range of the station set; analyzing the influence relation between exterior environment jamming input nodes and passenger flow or travelling time decision nodes and establishing a Bayesian network structural diagram for public transit dynamic environment forecasting; determining a conditional probability table between determinant conditions and the decision nodes; computing the posterior probability when certain public transit passenger flow or travelling time occurs, and accordingly, achieving forecasting of public transit environment dynamic change. Combined with public transit incident detection under the environment of an Internet of vehicles, the Bayesian network model based public transit environment dynamic change forecasting method achieves a dynamic passenger flow time and space change forecasting function and provides data support for daily public transit operation and management.

The invention discloses a brain function magnetic resonance image classification method based on network centrality. The method comprises the following steps of: preprocessing a brain function magnetic resonance image, performing brain region segmentation, and extracting an average time sequence of each brain region; calculating a partial correlation coefficient between each average time sequence, and obtaining a partial correlation coefficient matrix; performing binarization on the partial correlation coefficient matrix to obtain a brain network model; calculating the network centrality of each node in the network; and classifying the brain function magnetic resonance image by utilizing an adaptive improvement classifier, and checking the adaptive improvement classifier by employing a leave-one-out cross validation testing method. The brain function network is established by utilizing the brain function magnetic resonance image, the brain function magnetic resonance image is classified by utilizing the network topology information, and the brain function magnetic resonance image can be accurately classified.

The present invention relates to accident probability evaluation, particularly provides a Bayesian network model based risk evaluation method for a road transportation accident, and aims to solve the problem that the influence of superposition and counteraction among root node factors on a final evaluation result is not considered in an existing method. For the purpose, the method comprises: acquiring factors related to the road transportation accident; constructing a Bayesian network model of the road transportation accident according to the factors; and estimating a risk probability of the road transportation accident according to the Bayesian network model. The method is characterized in that the Bayesian network model has a three-layer network structure, and the method evaluates the probability of an intermediate node according to the probability of a root node and evaluates the risk probability of the road transportation accident according to the probability of the intermediate node. According to the method, superposition and / or counteraction effects among the factors of the root node are considered when the Bayesian network model is constructed, so that the accuracy of final evaluation can be greatly improved.

The invention discloses a wireless cognitive network knowledge base constructing method based on a Bayesian network. A Bayesian network model is constructed by using perceived network performance parameters, so that the dependence among the parameters is obtained, and the conditional probability dependence is transformed into knowledge in a wireless cognitive network to be stored in a knowledge base for guiding a process of intelligent decision making. The process of constructing the knowledge base mainly comprises the steps of: perceiving the network performance parameters, discretizing and analyzing the parameters, constructing the Bayesian network model and transforming the knowledge to construct the knowledge base. The network performance parameters are collected through a perceiving process, and data is discretized and analyzed. Through a structure learning and parameter learning process, the Bayesian network model is constructed by using historical data. The conditional dependence among the parameters in the Bayesian network is transformed into the knowledge so that the knowledge base is constructed and updated.

The embodiment of the invention provides an evaluation method of the debris flow disaster risk based on a Bayesian network model. The evaluation method is characterized by comprising the following steps of: determining an evaluation unit; obtaining an evaluation index of the debris flow disaster risk of the evaluation unit by processing an evaluation parameter of the debris flow disaster risk of the evaluation unit; creating a training sample set according to historical data of the debris flow disaster risk in the evaluation unit and the evaluation index; creating the Bayesian network model according to the training sample set; and evaluating the debris flow disaster risk in an area to be evaluated by adopting the Bayesian network model. According to the evaluation method provided by the invention, the Bayesian network model is created by combining the historical data of the debris flow disaster risk in the evaluation unit and the evaluation index, the debris flow disaster risk in the area to be evaluated is evaluated by adopting the Bayesian network model, the accuracy of a debris flow disaster risk evaluation result is greatly increased, and the debris flow disaster risk in the area to be evaluated can be accurately evaluated by adopting the evaluation method.

The invention relates to a method for dynamically generating bus timetables on the basis of Bayesian network models. The method includes screening microscopic and macroscopic factors affecting dynamic generation of the bus timetables; building the double-layer microscopic and macroscopic Bayesian network models for dynamically generating the bus timetables, and in other words, building the Bayesian network models for forecasting dynamic variation of bus environments and the Bayesian network models for dynamically generating the bus timetables; predicting transport capacity and transport volume occurrence probabilities of various routes under the condition of random disturbance and reasons for unbalance of the transport capacity and the transport volumes of the various routes; combining scheduling policies with one another and generating possible timetable schemes around the target for timely evacuating passengers; computing various indexes for evaluating the quality of the timetables from the points of governments, enterprises and the passengers and evaluating the quality of the timetables. The method has the advantages that a function of dynamically adjusting the timetables according to variation of the bus environments can be implemented, and accordingly technical support can be provided for daily bus operation management.

The invention discloses a wind power climbing event probability prediction method and system based on a Bayesian network, and the method comprises the steps: mining the dependency relationship betweena wind power climbing event and related meteorological influence factors such as wind speed, wind direction, temperature, air pressure, humidity, and the like, and building a Bayesian network topological structure with the highest fitting degree with sample data; quantitatively describing a conditional dependency relationship between the climbing event and each meteorological factor, estimating the value of each conditional probability in a conditional probability table at each node of the Bayesian network, and forming a Bayesian network model for predicting the wind power climbing event together with a Bayesian network topological structure; deducing the probability of occurrence of each state of the climbing event according to the numerical weather forecast information of the mastered prediction time; the value of the corresponding conditional probability at each node is adaptively adjusted, so that the inferred conditional probability result of each state of the climbing event is optimized, and the compromise between the reliability and the sensitivity of the prediction result is realized.

The invention provides a monocular vehicle 3D target pose estimation method, system and terminal and a storage medium. The method includes: S01, establishing a basic network model; S02, acquiring a captured image, acquiring a labeled contour of an object to be labeled from the captured image, acquiring coordinate data of each vertex of the labeled contour, and taking plane coordinate data of eachvertex of the labeled contour as an input of a training basic network model to train the basic network model; and S03, inputting an image to be detected into the basic network model, and outputting the 3D labeled result of the captured image. The method establishes a fast detection based CNN network model, two-dimensional vehicle target detection and three-dimensional vehicle pose estimation are carried out, and the pose of the target is estimated while the target is detected. The SSD algorithm is extended to cover the whole 3D pose space, and only the synthetic model data is trained; and onlythe traditional vision sensor is needed, the hardware cost is very low, and the performance-price ratio is high.

The invention provides an aerial photograph image difference detection method based on a dual network, comprising four parts of image acquisition and processing, building a dual network model, training the dual network model and using the model. As the depth learning is not required to design feature manually, the problem that it is difficult to select an effective feature descriptor in image segmentation and difference detection can be avoided. The depth learning method can also overcome the shortcomings of low robustness to illumination in RGB image difference detection task. At the same time, the invention adopt the object detection instead of the traditional segmentation method, which can more distinguish the single object, and the position coordinate of the object is more accurate andeasier to express. By calculating the ROI of the prediction box after detection to determine the relevant objects of the two images, it can also reduce the requirements of registration accuracy. Importantly, the invention adds the semantic information of the object, has the object class information, has stronger anti-interference ability, and can better analyze the difference types.

Risk assessment method of submarine pipeline leakage accident based on fuzzy Bayesian network. The existing Reason model risk accident assessment method cannot carry out the risk assessment of bottom pipeline leakage accidents based on many uncertain factors of marine climate. The present invention is realized through the following steps: establish a Bayesian network model according to the characteristics of the Reason model and the leakage accident data of the submarine pipeline, establish an expert system, and determine the method for determining the expert weight; use triangular fuzzy numbers to determine the expert weight expressed in the fuzzy language determined in step 1 Quantify the method to determine the logical relationship between events; convert the fuzzy number into a probability value; define the logical relationship between events in GeNIe2.0 software, analyze the Bayesian network model, and obtain different degrees of accidents probability, so as to determine the risk level of accidents with different leakage degrees. The invention can more accurately evaluate the occurrence probability and level of the leakage accident risk of the submarine pipeline.

The invention discloses a blowout accident risk analysis method, and particularly relates to the field of energy exploitation, and the method comprises the following specific analysis steps: S1, analyzing main cause of well kick-blowout accidents and influence factors of the well kick- blowout accidents; s2, establishing well kick- blowout accident tree; s3, converting the accident tree into a Bayesian network model by using GeNIe; s4, performing quantitative analysis on the Bayesian network model; s5, finding out a key cause factor; and S6, proposing a measure for controlling the key cause factor. According to the invention, accident causes are analyzed; the analyzed cause of the blowout accident is compiled; a reasonable blowout out-of-control accident tree and a blowout out-of-control Bayesian network model are obtained, the blowout out-of-control Bayesian network model is subjected to quantitative analysis and three importance analysis, key cause factors causing blowout accidents are found out, and a key role is played in controlling the blowout accidents.

The invention relates to a BNM-based (Bayesian network model-based) fishery forecasting method. The method includes the steps of discretizing historical ocean information datasets of fishery environment; establishing a table of conditional probability between Bayesian network structure charts and Bayesian network nodes; calculating a posterior probability distribution formula of fishery by the Bayesian network structure chart obtained by optimal learning algorithm; and forecasting the fishery by the obtained posterior probability distribution formula. The function of rapidly forecasting for the fishery can be realized by the BNM-based fishery forecasting method.

The invention discloses a speaker recognition method based on a twin network model and a KNN (K-nearest neighbor)algorithm. The speaker recognition method based on the twin network model and the KNNalgorithmcomprises the steps that S1, voice information of a speaker is collectedby using a microphone and taken as a data set to train an RNN network model; and S2, the speaker is identified by using atrained RNN network model to construct the twin network model and combining the KNN algorithm. By adopting the technical scheme of the speaker recognition method based on the twin network model and the KNNalgorithm, the data set of the speaker in a database is trained, it is ensured that input of each speech signal inputted into a twinnetwork can output the characteristics representing the speaker, distances between different output characteristic vectors are calculated by cosine distance and used in the KNN algorithm for determining whether the speech signals belong to the same speaker or not, the speaker can be identified with asmall amount of samples, the network does not need to be retrained as the number of speakers increases, the requirement of data sample sizeofa neural network isreduced, and meanwhileinstantaneity and accuracy of speaker recognition are effectively improved.

The invention discloses a convolution neural network visual analysis method based on difference comparison, which comprises the following steps of designing a basic network model by using Tensorflow,and modifying parameters to obtain a control model; training the two models and extracting the parameters of the model after training; inputting the obtained model parameters into the variance analysis system to display the variance; by observing the variance overview component of the variance analysis system, identifying the possible key variance points quickly; through the interactive exploration component provided by the system, analyzing the possible key differences in further detail, and drawing the conclusion. The method of the invention effectively realizes the difference visual analysis, and the user can find the key problem more efficiently in the process of actually modifying the neural network model by understanding the difference.

A method for detecting gas path fault of gas turbine based on Bayesian network model is provided. The method comprises the following steps: the data of gas path components of the gas turbine obtainedby real-time signal acquisition is generated into a data set, normal operating parameters are obtained from the data set, Abnormal parameters to be measured, Set and Test Set are trained, the optimaltraining set and optimal test set are obtained by pretreatment and clustering analysis, Then the optimized Bayesian network model for real-time detection of the current operating conditions of the gasturbine system is obtained by testing the initialized and parameter optimized Bayesian network with the optimized test suite, so as to detect system failures. The invention provides a specific Bayesian network model structure learning and parameter learning method, and further establishes a correlation model between measured parameters and gas turbine normal operating condition parameters, and realizes on-line fault detection of gas turbine gas path system.

The invention discloses a target detection method and device based on a BSSD ( Bidirection Single Shot Multibox Detector). The BSSD network model is constructed based on the SSD network model; firstly, a relatively low high-level feature layer is fused with a feature layer, used for detecting a minimum target, of previous SSD through linear interpolation, and then a relatively high low-level feature layer is spliced with the fused feature layer through a passthrough method to obtain a feature layer used for detecting a small target; according to the improved BSSD, feature information extractedby each network layer can be fully utilized; the problem that the SSD does not use high-level feature semantic information is effectively improved, and the setting mode of the SSD default box and thesampling proportion in the data augmentation strategy during model training are further modified so as to improve the detection capacity of the model for small targets. Therefore, compared with the SSD, the improved BSSD has a better detection effect on small target detection and has better robustness.

An automobile engine fault maintenance method based on Bayesian network models and multi-criteria decision analysis includes: using each fault category and each fault source under each fault category, which affect an automobile engine, to build Bayesian network models so as to obtain a first probability value of each fault category and a second probability value of each fault source; respectively determining standardized weight of each decision criterion through multi-criteria decision analysis according to the decision criteria; obtaining a first effect value of each fault category according to each decision criterion, the standardized weight and the first probability value, selecting the fault category with the highest first effect value, obtaining a second effect value of each fault source under the fault category with the highest first effect value, and selecting the fault source with the highest effect value in the second effect values to serve as the maintenance judging results.

The purpose of the invention is to provide a maintenance decision-making method for a diesel engine fuel oil system through cost analysis in combination with a Bayesian network model. The method comprises the steps of: firstly, establishing the Bayesian network model of the diesel engine fuel oil system, and carrying out fault diagnosis on the fuel oil system based on the model to obtain an occurrence probability of each fault; secondly, carrying out non-dimensional processing on factors which influence a maintenance operation cost of the fuel oil system by use of a standardization formula; thirdly, fusing a plurality of influence factors in maintenance operations by use of an RBF (Radial Basis Function) neural network, and evaluating a corresponding maintenance cost; and finally, comprehensively evaluating a fault occurrence probability and a maintenance cost through a multiplication formula, and ordering the maintenance operations according to a product decreasing rule to obtain an optimal maintenance strategy of the fuel oil system. According to the method, through the cost analysis in combination with the Bayesian network model, the fault probability and the maintenance cost are comprehensively evaluated, and a decision making is carried out on a maintenance strategy of the fuel oil system so that a decision-making result has reference value.

The invention discloses an image annotation method based on a convolutional neural network and binary coding features. The method comprises the following steps: constructing an Incepton V3 basic network model; intercepting a final pooling layer of the Incepton V3 network basic model; removing Logits and softmax functions of the Incepon V3 network basic model, and using a sigmoid function as an activation function of the last layer to obtain a modified first basic network model; adding two full connection layers on the first basic network model, and using a sigmoid function as an activation function of the last layer to obtain a multi-label classification network model; performing training learning on the training set by using a multi-label classification network model, and optimizing the weight of the multi-label classification network model; marking the feature vector set of the target image based on the trained multi-label classification network model to obtain multi-label probability output of the target image; and in combination with multi-label probability output, labeling the target image by adopting a TagProp algorithm. Multi-label labeling of images can be realized, the cost is low, and the efficiency is high.

The invention discloses a shoe sample and footprint key point detection method based on depth learning, which comprises the following steps: S1, obtaining shoe sample / footprint data climbing shoesample data pictures by using crawler technology, obtaining shoe sample pictures, marking key points by manual demarcation mode, and generating shoe sample data sets; 2, setting a network model; 3, calculating a loss function, and proposing a loss function based on that outline of the sole / footprint; S4, training the network model, adopting the transfer learning mode of partial network structureadjustment for training; 5, inputting the normalized image size into the train network model, and marking the coordinates of the output result on the original map. Using depth learning network to extract key point information makes it possible to mark footprints or footwear images by computer which greatly reduces the human cost.

The invention discloses a hopping sequence prediction system based on a graphical model. The system comprises a preprocessing module, a prediction module and a feedback adjusting module, wherein the preprocessing module is used for removing noise and bandwidth from an intercepted original hopping sequence; the prediction module is connected with the preprocessing module and used for reconstructing a phase space and constructing a prediction model; and the feedback adjusting module is connected with the preprocessing module and the prediction model and used for precision detection, feedback and model adjustment. According to the system, embedded dimension m and time delay phi are solved by adopting a Cao method and an autocorrelation method, and then the phase space is reconstructed; and the Markov boundary of query nodes is learnt on the basis of an improved Microsoft malware protection center (MMPC) algorithm, and then the prediction model is constructed. The embedded dimension m and the time delay phi are two key parameters for reconstructing the phase space, and the parameters acquired by using the autocorrelation method and the Cao method are stable and reliable; and a Bayesian network model is simplified through the Markov boundary, so that the prediction efficiency is high.

The invention discloses a vehicle multi-attribute detection method based on single-network multi-task learning. The method comprises the following steps of collecting and screening pictures; making adata set; carrying out network design on the basis of a Darknet deep learning framework, designing a network structure by adopting an end-to-end and one-stage non-cascade mode according to the characteristic of multiple attributes of the vehicle, and constructing a network model; carrying out model training, setting and adjusting model parameters, training a vehicle multi-attribute data set according to a designed network model, and carrying out data enhancement and multi-scale training during training; and carrying out model testing and model evaluation. The Darknet-based deep learning framework platform is designed according to the Darknet-based deep learning framework platform; a network model is built and is of an end-to-end one-stage non-cascade structure, the network improves the detection effect of multiple attributes of a vehicle by adopting the technologies of data enhancement, convolution kernel separation, multi-scale feature fusion and the like, and better real-time performance is achieved while the higher detection accuracy and recall ratio are achieved.

The invention discloses an industrial production process fault monitoring method based on a hierarchical non-Gaussian monitoring algorithm. The industrial production process fault monitoring method comprises the steps of collecting train data and to-be-detected data, calculating the cross-entropy between every two input variables, according to the cross-entropy, dividing all the input variables into various subblocks, building a non-Gaussian monitoring model in each subblock by utilizing a two-layer non-Gaussian monitoring algorithm to extract data of the non-Gaussian part in each subblock, and calculating control limits and a statistic amount of the data of the non-Gaussian parts; in each subblock, calculating data of the remaining Gaussian part to obtain the control limits and statisticamount of the Gaussian parts; conducting fault detection through the control limits and the statistic amount. The industrial production process fault monitoring method based on the hierarchical non-Gaussian monitoring algorithm is better than other traditional methods in fault detection of the non-Gaussian process, not only can the highly complex coupling relationship among variables be sufficiently considered, but also the non-Gaussian part of the data with unknown distribution characteristics can be extracted, and thus the fault detection in the chemical engineering process is more efficientand more accurate.

The invention provides a positioning error compensation method for a robot straight line shaft based on data driving, and belongs to the technical field of robot automatic assembly. According to the positioning error compensation method, a target ball is placed at the tail end of the robot straight line shaft, a plurality of mark points are arranged in a robot space, the robot is controlled to move the target ball to each mark point, and the nominal positions of all the mark points under a robot coordinate system are obtained to serve as input values of a training set; the actual positions ofall the mark points are measured, and difference values are obtained by comparing the nominal positions and the actual positions of all the mark points to serve as the spatial positioning errors of the mark points and serve as output values of the training set; a Gaussian process regression model is used for training, and a Gaussian error model after training is obtained; and the Gaussian error model is used for compensating the spatial positioning errors of a robot, and the compensated kinematic errors of the robot are obtained. According to the positioning error compensation method, the measuring process is simple and convenient, high-precision measuring results can be obtained, and therefore high-precision real-time online compensation for kinematic errors of an automatic hole making system is achieved.

The invention discloses a forwarding modeling method and device based on a spring network model, and the method specifically comprises the steps: setting a time sampling interval and a spatial sampling interval of forwarding modeling; carrying out the discretization of a to-be-simulated geologic model according to the spatial sampling interval, and obtaining a digitalized geologic model, wherein the digitalized geologic model is of a rectangular grid form or a cuboid grid form; determining the elastic parameters of all directions of the digitalized geologic model according to the spatial sampling interval; and updating the wave field value of the digitalized geologic model at each moment according to a seismic focus subwave function, the time sampling interval and the elastic parameters of all directions of the digitalized geologic model. The method overcomes the restrains of a conventional grid shape, can be suitable for square or cube grids, also can be suitable for rectangular or cuboid grids, can meet the requirements of actual production for a simulated grid, and also can preferably select the time and spatial sampling intervals according to the actual conditions, so as to improve the calculation efficiency.

The invention provides a power load peak forecasting method and device based on a Bayesian network model. The method comprises the steps of: taking the obtained meteorological data and the time data after numerical processing as clustering features, clustering the load data obtained, and then predicting the peak value of the electric load according to the pre-constructed Bayesian network model, wherein the Bayesian network model is constructed according to the clustering result, and the invention has the advantages of small deviation of the prediction result when the external influencing factors change greatly, fast learning speed, and large-scale sample processing. The present invention can effectively process incomplete data, combine with other technologies for causal analysis, effectively combine prior knowledge and data, effectively avoid transitional fitting of data, and still have high accuracy when environmental factors vary greatly.