549 results about "Kernel density estimation" patented technology

The present invention discloses method for moving target detection and tracking in a complex scene. The method comprises two steps of multiple moving target detection and multiple moving target tracking: in the multiple moving target detection, a background model based on self adapting nonparametric kernel density estimation is established with the aim at the monitoring of the complex scene, therefore the disturbance of the movement of tiny objects can be effectively suppressed, the target shadow is eliminated, and the multiple moving target is detected; in the multiple moving target tracking, the target model is established, the moving state of the target is confirmed through ''matching matrix'', and corresponding tracking strategy is adopted according to the different movement condition of the target. Target information is ''recovered'' through the probabilistic reasoning method, and the target screening degree of the target is analyzed with the aim at the problem that multiple targets screen mutually. The algorithm of the present invention can well realize the moving target tracking, obtains the trace of the moving target, and has good real time and ability of adapting to the environmental variation. The present invention has wide application range and high accuracy, therefore being a core method for intelligent vision monitoring with versatility.

The invention discloses a crowd density estimation method based on video analysis and a pedestrian volume statistical method based on the video analysis. The crowd density estimation method includes the flowing steps of (1) off-line training: manually counting crowd density data, extracting characteristics and conducting training; and (2) on-line estimating: extracting the characteristics and conducting regression prediction by utilizing trained model parameters. The pedestrian volume statistical method includes the step of setting up a robust relationship between a scene and a line-passing number of people by combing the crowd density and a micro-region pedestrian flow speed before a line is passed. Characteristics such as foregrounds, edges and gray scale co-occurrence matrixes are extracted based on a whole area to conduct crowd density estimation, problems of dense crowds, sheltering and the like can be well solved through mixing of the characteristics, and real-time crowd density estimation is achieved. In addition, on the basis of area crowd density estimation, pedestrian volume estimation is conducted through combination of the pedestrian flow speed based on an optical flow, detection and tracking of a large number of individuals under a complex environment are avoided, and two-way pedestrian volume counting of accurate robust under dense crowds is achieved.

The invention discloses a ship collision risk analysis method based on AIS (automatic identification system) data. Based on historical AIS data and standard ship selection and conversion, a density clustering algorithm is used for establishing a heat map of a ship collision risk to realize spatiotemporal visualization of the ship collision risk; based on real-time AIS data and the ship position field, the course direction field, and the navigational speed field, a regional ship collision risk assessment model is constructed, and a Gaussian kernel function kernel density estimation algorithm isused for proposing a dynamic ship collision risk visualization method to realize areal-time update of the regional ship collision risk. The ship collision risk analysis method is based on the AISdata,the spatiotemporal visualization of the ship collision risk is realized,the visual image after the complex abstract ship traffic flow multi-attribute information is effectively dug and fused is realized, so that the risk level of the environment of the location of the ship can be intuitively and conveniently obtained by a driver or an operator, thusself-alertness is improved, the reasonable control measures are taken, and the safe operation of the ship is ensured.

The invention discloses a residual error posterior-based abnormal value online detection and confidence degree assessment method. The method comprises the steps of collecting data, establishing time series data, performing linear fitting on the time series data to obtain a linear combination formula of data at a current moment and p pieces of previous data, and predicting a data value of subsequent time; comparing the predicted data value with an actually detected data value to obtain a predicted residual error series; determining a probability density function of the predicted residual error series by adopting a KDE (Kernel Density Estimation) method; performing posterior ratio check on the predicted residual error series, and judging whether the data at the current moment is an abnormal point or not; and by taking the time series data as an input, building an SOM state model, obtaining state series and state transition probability matrixes, defining an abnormal scoring function, and outputting an abnormal score. By comparing the probability that the data is the abnormal point with the probability that the data is a normal point, the abnormal value in the pollutant discharge concentration time series data is identified online, so that the accuracy and reliability of abnormal value judgment are improved.

The invention discloses a method for city building function classification based on a high-resolution remote sensing image. The method comprises the steps of A, extracting buildings in the high-resolution remote sensing image by adopting a CNN (Convolutional Neural Network) method, and acquiring a building extraction result; B, sorting and classifying POI (Point of Interest) data according to attribute information, respectively performing kernel density estimation on POIs of the commercial and service facility land, the public management and public service land and the residential land, and respectively acquiring kernel density maps of the land types; and C, calculating a kernel density average value of the single building by using the CNN based remote sensing image building information extraction result and the kernel density maps. The method is easy to implement and simple and convenient to operate, effectively solves a problem that semantic-level building function classification and recognition are difficult to realize by using a remote sensing information extraction technology, is high in precision of function classification for the city buildings, can provide dynamic data of city functional area classification for relevant departments quickly and accurately and serves for city management and reasonable planning.

Provided is a domestic service robot path programming method based on a walking trajectory. The invention discloses a service robot navigation method based on a human walking trajectory. Firstly, the method of describing the trajectory of pedestrian movement is discussed; a key point extraction and trajectory expression method based on kernel density estimation is proposed; the pedestrian movement trajectory classification is realized by the similarity measurement of the trajectory in space distance, motion direction and velocity; secondly, in order to avoid the expansion of a path search algorithm in the global space and improve the real-time performance of the algorithm, an improved RRT-Connect algorithm based on a trajectory gravitational function is proposed to realize the path search scheme which is oriented to the pedestrian walking trajectory; and finally, double-layer path programming based on a topology-grid environmental model is realized; and the navigation efficiency is raised on the basis of safety navigation.

The invention discloses a load curve clustering method based on improved spectral and multi-manifold clustering. The load curve clustering method comprises three steps of typical daily load curve extraction, load curve clustering and clustering effect evaluation. Firstly, load characteristic indexes of a user are extracted, and typical daily load curves of the user are calculated and extracted bycombining a non-parameter kernel density estimation method; canonical warping distance metering curve similarity is introduced into an improved spectral and multi-manifold clustering algorithm, localsimilarity is calculated by adopting a Gaussian kernel function, and a similarity matrix is calculated based on the local similarity; and various clustering effectiveness indexes are adopted for evaluating a clustering result and algorithm performance after clustering. The local similarity adopts load data of a plurality of users in Baoding area as calculating example samples for performing clustering analysis, and verifies the rationality and superiority of a typical daily load curve extraction method and the improved spectral and multi-manifold clustering algorithm disclosed in the invention.

Anomalous activities in a computer network are detected using adaptive behavioral profiles that are created by measuring at a plurality of points and over a period of time observables corresponding to behavioral indicators related to an activity. Normal kernel distributions are created about each point, and the behavioral profiles are created automatically by combining the distributions using the measured values and a Gaussian kernel density estimation process that estimates values between measurement points. Behavioral profiles are adapted periodically using data aging to de-emphasize older data in favor of current data. The process creates behavioral profiles without regard to the data distribution. An anomaly probability profile is created as a normalized inverse of the behavioral profile, and is used to determine the probability that a behavior indicator is indicative of a threat. The anomaly detection process has a low false positive rate.

The invention relates to a kernel density estimation-based load spectrum compilation method for a transmission shaft of a tracked vehicle. The method comprises the following steps of S1, collecting and preprocessing torque load sample data of the tracked vehicle; and S2, generating a two-dimensional load spectrum through first two-time rain flow counting, mean amplitude extremum inference, second rain flow counting, two-dimensional kernel density estimation, multi-condition synthesis and extrapolation in sequence. According to the method, the two-time rain flow counting is adopted; a first rain flow counting result is used for the mean amplitude extremum inference; a second rain flow counting result is used for the kernel density estimation; mean amplitude distribution can be well fitted; and an actually measured rain flow matrix can be reasonably extrapolated. The load spectrum compiled by adopting the method and the actually measured rain flow matrix have highly similar probability density distribution, reasonable extrapolation of the actually measured rain flow matrix is realized, and an expected effect is achieved.

Disclosed is a radar signal clustering method using frequency modulation characteristics and combination characteristics of signals including: a first step of assigning pulses of received radar signals to cells consisting of parameters including radio frequency (RF) and angle of arrival (AOA) of the pulses; a second step of calculating a pulse density distribution of each cell using a kernel density estimator; a third step of extracting a corresponding cell as a frequency fixed cluster if the calculated pulse density distribution is greater than a threshold of the frequency fixed cluster; a fourth step of making cell groups by merging remaining cells that are not extracted as the frequency fixed clusters; a fifth step of calculating a pulse density distribution of each cell group by using the kernel density estimator for each cell group; and a sixth step of comparing the calculated pulse density distribution for each cell group with each threshold according to a signal combination type of frequency agile clusters, thus to classify and extract each cell group according to the signal combination type.

The invention relates to a vertebra positioning method based on a convolutional neural network. Vertebra positioning is an effective way for judging spinal diseases, because tissues around the vertebra have complex structures and abundant textures, diagnosis is difficult, and the efficiency is low. A vertebra positioning method based on a convolutional neural network is provided based on the aboveproblem, and therefore vertebra positioning is more accurate. The vertebra positioning method comprises following steps: A, processing vertebra CT data to increase the contrast ratio; B, carrying outsegmentation and extraction on a key area in a vertebra CT picture by means of a Faster R-CNN model; C, predicting the location of the center of mass of each segmented vertebra by means of a kernel density estimation method; and D, carrying out three-dimensional reconstruction on the processed vertebra CT picture by means of the Mimics.

The invention discloses a regional average value kernel density estimation-based moving target detecting method in a dynamic scene. The method comprises the following steps of: firstly, initializing a background model; secondly, building a time and space background model for describing the dynamic complex scene by using a training sample in a background modelling process and considering the time sequence characteristics of pixel points in a video frame and the space characteristics in the adjacent regions of the pixel points; thirdly, continuously updating the background model by using the new video frame sample in a moving target detecting process; fourthly, adapting to the instantaneous background change by the regional kernel density estimating method and adapting to the continuous background change by using single Gauss background model, wherein the combination of the two models can fast and accurately adapt to the continuous change of the background and increases the executing efficiency of the method at the same time; and finally performing a foreground detecting method by providing an adjacent region information amount-based method so as to further remove noise points and inanition of a moving target in the background region in the detecting process and more completely extract the moving object in the foreground. The method can be widely applied to alarming the suspicious moving target in an intelligent monitoring system in an outdoor scene or a prohibited military zone and has wide market prospect and application value.

The invention relates to a health state assessment method for a rotary machine based on statistical methods such as kernel density estimation and K-L divergence calculation. The method comprises the following steps that 1) original vibration data of a monitored object is collected; 2) time-domain and frequency-domain features are extracted from the original vibration data; 3) the dimensions of the time-domain and frequency-domain features are reduced to obtain sensitive features; 4) a movable sliding window of the width k is used to dynamically select sample sets, and the statistical characteristics of the sample sets are calculated; 5) the probability density function of each sample set under each sensitive feature is calculated; 6) the K-L divergence between the probability density functions of two adjacent sample sets under the same sensitive feature is calculated; and 7) an integrated K-L divergence is calculated and serves as a health assessment index of the monitored object. The health state assessment method has the advantages that the statistical uncertainty of the samples is fully considered, and the accuracy and generalization performance of a health state assessment model are improved.

The invention discloses a dynamic probabilistic power flow (PPF) calculating method considering wind speed predication error temporal-spatial coherence. The method is to analyze the node voltage and dynamic probability distribution of branch power flow of a wind power station built power system, so as to enable operators to analyze a system state conveniently. The method comprises the steps of describing the input variable predication error process according to autocorrelation coefficient stationary process; directly fitting to obtain the predication error distribution on the basis of nonparametric kernel density estimation and according to historical predication error data; performing Nataf transformation technology on the basis of the iso-probability transformation theory to obtain an error sample of temporal-spatial coherence; performing dynamic PPF calculation by the monte carlo simulation method on the basis of latin hypercube sampling so as to obtain the node voltage amplitude value and the digital characteristics and dynamic probability distribution of the branch power flow. By adopting the method, the node voltage and the dynamic probability distribution of the branch power flow can be effectively analyzed; the method has the advantages of being accurate in result and convenient to realize.

The invention provides a kernel method-based collaborative filtering recommendation system and a kernel method-based collaborative filtering recommendation method. The corresponding system comprises a data preparation module which is used for standardizing the original data and carrying out corresponding preprocessing, generating a user-project rating matrix and a project distance matrix to output; a user interest modeling module which is used for constructing an interest model for a user on a project space according to the user-project rating matrix and the project distance matrix as well as a kernel density estimation technology; and a recommendation result generation module which is used for computing the similarities among the users according to the interest model, generating a neighbor set of a target user, and predicting a score of the project rated by the user according to a predetermined recommendation strategy and returning the recommendation result. Through the recommendation system and the recommendation method provided by the invention, the user interest model can be better presented, the user similarity in the practical application is estimated more accurately, the performance of the recommendation system can be promoted considerably, and more stable recommendation result can be obtained.

The invention relates to the field of data processing, in particular to a wind power modeling and performance evaluating method based on a confidence equivalent power curve band. The wind power modeling and performance evaluating method includes the steps that preliminary screening and rejection are conducted on an abnormal data sample; wind speed is divided into three regions, and a kernel density estimation method is used for obtaining wind speed and power probability distributions in each region through statistics to obtain the Copula function in each region; the maximum likelihood estimation method is adopted to obtain confidence equivalent power boundary model in the corresponding region; a piecewise cubic Hermite interpolating polynomial method is used for reconstructing missing datato complete original data sample cleaning; the average value of the confidence bandwidth ratio is used as the model performance evaluation index, a d-fold cross validation method is used for validating the upper and lower boundary models in different regions, when the index is basically stabilized at a certain constant value, the up and lower boundary models of the different regions are determined; a sliding time window method is adopted to updating data, the deviation degree of the confidence bandwidth ratio is used as a triggering condition, and the up and lower boundary models are updatedwhen the deviation degree exceeds a certain threshold value.

The invention relates to a terminal area prevailing traffic flow recognizing method based on track spectral clusters. The method includes the steps of firstly, analyzing a given airport pavement to obtain flight path data, and conducting dividing to obtain feature track points and normal track points; secondly, setting up a space rectangular coordinate system; thirdly, putting forward an occupation degree concept according to the distance between a track and the center of a space grid, and enabling the occupation degree concept to be used for representing the occupation degree of the track in the space grid; fourthly, setting up an inter-track overall similarity model on the basis of a track partial similarity model; fifthly, constructing an Laplacian similarity matrix, and then analyzing the clusters through the spectral cluster algorithm; sixthly, conducting prevailing traffic flow recognition on the clusters obtained in the fifth step through the nuclear density estimation method; seventhly, displaying the recognition result in a displaying and interaction module. The terminal area prevailing traffic flow recognizing method has the advantages that a prevailing traffic flow track and an abnormal track can be simultaneously obtained through the spectral cluster algorithm, therefore, related personnel are assisted in scientifically and reasonably planning a terminal area and improving airport entering and leaving air lines, and the capacity of the terminal area is improved.

The invention discloses a method for predicting the short-term wind power probability density based on the EWT quantile regression forest. The method comprises the steps of 1) decomposing an originalwind power sequence into a series of mutually different feature empirical modes by using the empirical wavelet transform (EWT); 2) recombining the empirical modes according to a frequency range to form high frequency, intermediate frequency and low frequency components; 3) select an input variable for each component by using the Pearson correlation coefficient; 4) establishing a quantile regression forest prediction model for each component, and obtaining regression prediction results of different quantile points; 5) superposing the prediction results of the components to obtain a wind power prediction value; and 6) obtaining the prediction of the wind power probability density by nuclear density estimation. The method provided by the invention effectively improves the prediction precisionof the wind power, obtains the prediction of the wind power probability density at any moment, and can well solve the wind power prediction problem in a power system.

The invention discloses a probabilistic power flow data acquisition method based on renewable energy uncertainty. The method comprises steps of S1 inputting a data file of electric parameters of a power grid and configuring random parameters of the power grid; S2, carrying out AC power flow calculation and probabilistic power flow calculation, and outputting MCS sampling scale, bus independent variable parameters and line independent variable parameters; S3, configuring a network equipment interval, including a bus voltage amplitude interval, a line load rate interval, and a transformer load rate interval; S4, utilizing the kernel density estimation to obtain the evaluation result of the utilization ratio of the electrical equipment; The invention can fully consider a plurality of uncertain factors of a power network, establishes a calculation program for estimating a power network power flow based on a probabilistic power flow method of Monte Carlo simulation, applies the program to apractical transmission system, analyzes a power network capacity level and a load rate, etc., and better guides the planning and evaluation of the power network.

The invention discloses a vehicle line-crossing detection method based on an intelligent video analysis technology. The method comprises the steps that a video monitoring image is preprocessed according to real-time light intensity; an improved three-frame difference algorithm is adopted to detect a vehicle in the video monitoring image; after the position of the vehicle is detected, a motion track of the vehicle is traced; the actual position of a target in a current frame is obtained through continuous iterative calculation, and the motion track of the vehicle is acquired; the result of thelast frame is used as an initial value of the next frame, and loop iteration is continued in this way; and multiple motion tracks of the vehicle are obtained, an appropriate motion track is selected to detect whether a warning line and the track intersect, and whether the vehicle crosses the line is determined. According to the method, the improved three-frame difference method is adopted to quickly detect the target vehicle, the motion tracks of the vehicle are traced through a mean shift algorithm based on kernel density estimation, then whether the vehicle has a line-crossing behavior is determined through the vehicle motion tracks, and the method is high in instantaneity, high in precision and wide in application range.

The invention provides a method and a device for generating a dynamic alarm threshold value of parameters of a refining process, and relates to the technical field of petroleum refining fault monitoring. The method comprises the following steps of obtaining historic data of the parameters of a refining process production device; determining initial historic training data of the parameters according to the length of a sliding window and the step length, and updating historic training data for once every other step length time from the initial historic training data; forming normalized historic training data; carrying out kernel density estimation on each parameter, and generating a probability density function of the parameters; determining a probability distribution function of each parameter; determining an alarm threshold value corresponding to each parameter according to the probability distribution function of each parameter and a preset alarm threshold confidence degree. According to the method for generating the dynamic alarm threshold value of the parameters of the refining process, provided by the invention, the dynamic alarm threshold value of the parameters can be obtained, and the problems that because a static model parameter threshold value method is adopted at present, and the alarm threshold value of the parameters is artificially regulated, false alarm or leakage alarm can be easily caused, and the sensitivity of abnormal monitoring can be easily reduced can be solved.

The invention relates to a continuous chemical process fault detection method. The continuous chemical process fault detection method comprises the following steps that (1) a linear regression model of a vector Xj and a vector Y is built, and a regression constraint function is introduced; (2) data compression is carried out through haar wavelet transformation to improve computational efficiency; (3) a regression constraint construction sparse pivot element model with the addition of 1-norm and 2-norm is built, and an optimal solution of a sparse pivot element is worked out through derivation of the SPCA algorithm; (4) the optimal threshold value of the T2 statistic and the optimal threshold value of the SPE statistic are estimated through the kernel density estimation (kde) method; (5) calculation of the T2 statistic and the SPE statistic is conducted on fault data, and the value of the T2 statistic and the value of the SPE statistic of the fault data are obtained in sequence; (6) whether a fault exists in the data is detected. According to the continuous chemical process fault detection method, the data size related to a pivot element after sparsity is reduced, so that the calculated quantity is reduced, the computation time is shortened, real-time performance of monitoring is improved, and accuracy and efficiency of fault detection can be improved.

The invention discloses an adaptive abnormal crowd behavior analysis method, which is used for analyzing crowd behaviors in a video image. The method comprises the following steps of performing streak line calculation on the video image; calculating a streak line flow; detecting abnormal behaviors; performing foreground detection on the video image of abnormal crowd behaviors; performing adaptive crowd density estimation comprising pixel-counting-based density estimation and texture-analysis-based density estimation, and finally dividing estimated density into four density levels, i.e. a low density level, a medium density level, a high density level and an ultrahigh density level, thereby finishing grading the abnormal crowd behaviors. According to the method, the concepts of streak line and streak line flow are introduced to analyze whether a crowd in the video image is abnormal or not; the method has the advantage of detection accuracy; the densities of crowds involved in the abnormal crowd behaviors in different density scenarios are estimated in an adaptive way, and the detected abnormal crowd behaviors are graded by using density estimation results as main characteristics; the method is used for accurately grading the abnormal behaviors (such as mass brawl) in crowded public places, and giving alarms.

The invention discloses an optimized configuration method of distributed power generation considering energy storage and considering photovoltaic randomness, comprising the following steps: firstly, based on kernel density estimation and K-means clustering method, according to a large number of historical photovoltaic output data, building a typical day scene generation method of distributed poweroutput; Furthermore, based on the theory of bilevel programming, a bilevel programming model of distributed generation is established, which combines the planning with the operation from the perspective of comprehensive social benefits and power company benefits, respectively, on the basis of the goal of the lowest annual total cost in the upper layer. Finally, the stochastic scenario generationis combined with the distributed generation bilevel programming model, and the optimal configuration scheme of distributed generation that meets the load demand is obtained by intelligent optimizationalgorithm. The bilevel programming model proposed by the invention can provide an optimal configuration scheme of distributed power generation including configuration node and capacity from differentbenefit parties, thereby effectively reducing the total annualized cost of the system.

Certain example embodiments described herein relate to systems and / or methods for event stream deviation detection. More particularly, certain example embodiments described herein relate to maintaining short and long-term statistics of an incoming stream of event data. In certain example embodiments, a deviation is calculated based at least in part on the long-term and short-term statistics. The deviation may then be compared to a threshold value. In certain example embodiments, the estimations required for the statistics are done with Kernel Density Estimators (KDEs).

The invention relates to an anomaly detection method based on multi-dimensional Epanechnikov kernel density estimation. By means of the method, abnormal data can be accurately detected. According to the technical scheme, the anomaly detection method based on multi-dimensional Epanechnikov kernel density estimation sequentially comprises the steps that (1) data at all distribution nodes are collected respectively, and then abnormal value diagnosis is conducted through a sampling method based on the kth closest distance; (2) a normal data sample is formed in a cluster head node sliding window, and a kernel density estimation model is established in the cluster head node sliding window according to the sample; (3) the kernel density estimation model is sent to all the distribution nodes, and each distribution node judges whether data arriving at the distribution node at the next moment are abnormal or not through the kernel density estimation model; (4) at each time interval T, each distribution node actively sends the normal data in the latest period of time to the cluster head node; (5) the step (1) is returned to.

The invention provides a detecting method and a detecting device for network attack and relates to the technical field of cloud computing. The detecting method for the network attack comprises the following steps: acquiring a current data flow in the network; based on a pre-established malicious act attack signature database, judging whether the behavior of the current data flow is abnormal or not; when the behavior of the current data flow is no, judging whether the behavior of the current data flow is normal or not by using a sliding window genetic algorithm frequent pattern mining model andan abnormal point detection model estimated based on nuclear density; when the behavior of the current data flow is no, extracting behavior characteristics of the current data flow, and adding the behavior characteristics into the malicious act attack signature database. According to the detecting method and the detecting device provided by the invention, by adopting a nested sliding window genetic algorithm frequent pattern mining model, the problems that a frequent mode, based on single-time scanning, of the current data flow is not high in mining accuracy, untimely processing of data is caused by high-speed growth of network data and the accuracy of a conventional intrusion detection technique is reduced due to complexity of a cloud computing environment network can be effectively solved.

The invention belongs to the technical field of power system dispatching automation, particularly relates to a robust scheduling method considering uncertainty of wind power and load prediction, whichcomprises the following steps of: describing the correlation between input load fluctuation and a prediction error of wind power output by using a correlation coefficient matrix method, and converting random variables with correlation into mutually independent random variable matrixes by using a Cholesky decomposition method; Constructing a probability density model of prediction errors of wind power output and load fluctuation by adopting non-parameter kernel density estimation; Introducing the direct-current power flow model into a power system dispatching model, and establishing an objective function and a constraint condition under the condition of uncertain factors by taking the minimum total dispatching operation cost of the system as an objective function of the model; adopting Benders decomposition method to solve a UC main problem of a robust SCUC problem, a network security check sub-problem of the UC main problem under a basic scene, and a network security check sub-problemunder an uncertain scene of new energy power generation and load.

The invention discloses a non-linear time-varying process fault monitoring method based on high efficiency recursion kernel principal component analysis and belongs to the fault detection and diagnosis technology field. The method comprises steps that data having non-linear and slow time-varying characteristics and containing faults is acquired from a Tennessee Eastman process simulator, a Gauss kernel function is utilized to project the acquired normal data to the high-dimensional characteristic space and is centralized, an initial offline monitoring model is established, and a kernel densityestimation function is employed to determine control limit; secondly, when new process data is acquired, through introducing a first-order interference theory method, a model is directly updated based on a characteristic value and a characteristic vector acquired in the offline model, the new data is projected to the updated kernel space and the residual error space to calculate T2 and SPE statistics; when the corresponding control limit is surpassed, occurrence of a monitoring fault is determined, otherwise, the whole process operates normally. The method is advantaged in that two problems are mainly solved, 1), a problem of relatively high false alarm rate generated during fault monitoring in the non-linear time-varying process of kernel principal component analysis is solved; and 2), aproblem of relatively high load existing in a recursion algorithm based on characteristic constant decomposition is solved.

The invention discloses a WSN (wireless sensor network) intra-network data fusion method based on kernel density estimation and non-parameter belief propagation, which comprises data acquisition and data fusion. The data acquisition is that monitoring unions which are respectively composed of no less than three sensor nodes for gathering the monitoring data are constructed in a monitoring region, each monitoring union is corresponding provided with a union header node for collecting the monitoring data, the sensor nodes in each monitoring union are respectively used for gathering the monitoring data of an object entering the monitoring region; and the data fusion is that the gathered monitoring data are subjected to KDE (kool desktop environment) processing by the sensor nodes in the monitoring unions respectively, the processed data are transmitted and collected to the union header nodes through NBP (name bind protocol) processing, the collected data are subjected to gauss mixing by the union header nodes, the data after gauss mixing are subjected to Gibbs sampling fusion, and the fused result is acted as a characteristic of the monitoring data. The accuracy of the monitoring data can be improved under a noisy or an uncertain environment, and the accurate fusion characterization of the monitoring data of the multi-node unions can be realized.