42 results about "Variable kernel density estimation" patented technology

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.

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.

The invention discloses a nonparametric kernel density estimation-based wind power prediction method which comprises the following steps: in a first step, historical wind power plant data in a preset time range is obtained, actually measured wind power data is divided into a plurality of subintervals, and the wind power plant data comprises actually measured wind speed and actually measured wind power; in a second step, statistics are run on actually measured wind power in each subinterval, and wind power probability density functions are established according to wind power distribution in each actually measured wind speed subinterval; in a third step, a confidence interval of the actually measured wind power is determined, wind power data outside the confidence interval is deleted, and data in the confidence interval is screened modeling data; in a fourth step, an S type function is used for fitting the modeling data and establishing a wind power prediction model which can be used for prediction.

The invention discloses a multiple-wind-power-plant correlation modeling method based on adaptive multi-variable nonparametric kernel density estimation, and belongs to the technical field of multidimensional variable correlation research. The method comprises the following steps that: S1: establishing a multi-variable nonparametric kernel density estimation model of a wind power plant; S2: constructing a bandwidth optimization model; and S3: constructing a bandwidth solving method of the multi-variable nonparametric kernel density estimation model of the wind power plant on the basis of ordinal optimization. By use of a method, a modeling process is practical and simple, correlation among a plurality of random variables can be quickly and efficiently modelled. Compared with a traditional parametric estimation method based on a copula function, the method is higher in accuracy and applicability, and a local adaptation problem of a traditional multi-variable nonparametric kernel density estimation method is favorably solved.

The invention discloses a steering engine reliability simulation sampling method based on Markova chain Monte Carlo, which comprises four stages: 1, Markova process simulation, namely selecting the initial state of a Markova chain, determining a random transition sampling probability density function, determining the next state of the Markova chain and constantly repeating to generate random sample points, of which the limit distribution is asymptotically optimal, of an importance sampling density function; 2, kernel density estimation, namely selecting a kernel density function, determining a window width parameter and a local bandwidth factor and generating a mixed importance sampling probability density function by using a self-adaptive width and kernel density estimation method according to Markova state points; 3, importance sampling, namely performing importance sampling according to the mixed importance sampling probability density function generated in the second stage; and 4,statistical calculation, performing failure probability estimation according to the important sample points generated in the third stage and calculating the failure probability of the system. The method effectively solves the problems of low simulation efficiency, low precision and mixed system.

The present invention provides a wind power fluctuation probability density modeling method based on nonparametric kernel density estimation. The method comprises the following steps: 1, extracting a fluctuation amount of wind power sample data by wavelet decomposition; 2, establishing a corresponding nonparametric kernel density estimation model based on a fluctuation amount sample, and then aiming at the model bandwidth selection problem, constructing a constrained bandwidth optimization model which uses a goodness-of-fit test as a constraint condition; and 3, solving the optimization model by using a constrained sequence optimization algorithm. According to the present invention, due to adoption of the wavelet decomposition method, a wind power fluctuation component can be more precisely extracted; moreover, a probability characteristic modeling method of the extracted fluctuation component is entirely driven by the sample data without performing prior subjective assumption on the probability density model, so that the method has higher modeling accuracy and applicability; and an improvement strategy aiming at the nonparametric kernel density estimation method also enables modeling accuracy and computing efficiency of the method to be effectively improved.

The invention relates to an expressway illegal parking detection method based on kernel density estimation and belongs to the field of image processing. The expressway illegal parking detection method comprises the following steps: firstly, carrying out background extraction by adopting a non-parameter kernel density model to obtain a background image; secondly, updating the background image by adopting a gradually-changed updating manner to obtain an updated background image; removing the background image from a currently acquired image to obtain a movement foreground; thirdly, calibrating positions of mass centers of a movable target vehicle; then tracking the target vehicle and measuring the distance between the mass centers; when the distance between the mass centers is gradually reduced, representing that the target vehicle enters a speed reduction process; after the target vehicle enters the speed reduction process, judging a movement state of the target vehicle; when the movement state is a static state, calculating illegal parking time; finally, determining whether the target vehicle is illegally parked or not according to the illegal parking time. The expressway illegal parking detection method provided by the invention can be used for monitoring a monitored scene in real time and alarming in time when the vehicle is illegally parked; the processing speed is rapid and the accuracy of alarming is improved; the expressway illegal parking detection method has the characteristics of good instantaneity, high robustness, high accuracy and the like.

The invention relates to a detection method for a construction change of a remote-sensing image based on a DSM and a kernel density estimation. The detection method comprises the following steps that a first image and a second image of the panchromatic remote-sensing image are preprocessed; angular points are extracted respectively; the center point of a candidate construction is searched, and the kernel density estimation is conducted by using a symmetrical Gaussian probability-density function; results of the kernel density estimation are overlapped, and two time phase kernel density estimation images are obtained; a difference algorithm is carried out on the kernel density estimation images, and a difference image Pdif is obtained; the difference image Pdif is labeled, and a change region CH is obtained; purification is carried out on the change region CH. The problems that a false detection rate and a missing detection rate of the construction change of the high spatial resolution remote-sensing image are high, the algorithm is complex are solved, and the method can be used in update of an urban geography database and fast recognition of illegal constructions.

The invention discloses a multichannel electroencephalogram data fusion and dimension descending method. The multichannel electroencephalogram data fusion and dimension descending method comprises the following steps of (1) reading in multichannel electroencephalogram data; (2) performing kernel density estimation on the electroencephalogram data by using a Parzen window to obtain an estimation value of the electroencephalogram data; (3) performing kernel transformation on the electroencephalogram data by using a polynomial kernel function, mapping the electroencephalogram data to corresponding kernel space to form kernel matrixes and fusing all the kernel matrixes corresponding to electroencephalogram of all channels into a synthetic kernel matrix by using different weight numbers; (4) calculating an eigenvalue and an eigenvector of the synthetic kernel matrix; and (5) performing entropy component analysis on the eigenvalue of the synthetic kernel matrix G and the eigenvector of the synthetic kernel matrix G by using a map of kernel entropy principal component analysis (KECA) to obtain low-dimension eigenvalue and eigenvector data and implement fusion and dimension descending of the multichannel electroencephalogram data. By the multichannel electroencephalogram data fusion and dimensional descending method, the electroencephalogram data of each channel are subjected to kernel function mapping, and effective fusion and dimension descending of the multichannel electroencephalogram data can be implemented through multi-kernel entropy component analysis.

The invention relates to an interpolation method applied to pseudo measurement generating and belongs to the fields of electricity system dispatching automation technologies and power grid simulation technologies. The method comprises the steps that firstly, load data collected by an electric quantity metering system are used as load measurement; an unknown predicted value is obtained by utilizing a certain algorithm; afterwards, the predicted value is utilized, historical data serve as a basis, and then an efficient equidistant-node interpolation method is combined; accordingly, the defects of a distribution network measurement device are made up for. According to the pseudo measurement generating method, the load data in the distribution network metering system are fully utilized, and the algorithm of the method is simple, and convergence performance is guaranteed; required accuracy can be always obtained based on an equidistant-node method as long as the distances between nodes are small enough; calculation is fast, the accuracy of a pseudo measurement load of a non-measurement point can reach or approach an actual measured value, the smoothness of the load data is maintained, and then the state estimation accuracy of a distribution network is improved.

The invention provides an anomaly detection method and device based on kernel density estimation. The method comprises the following steps: acquiring at least three feature vectors subjected to data processing in advance; determining a density estimate corresponding to each of the feature vectors; determining a probability density function of the at least three eigenvectors based on each of the density estimates; obtaining a probability of occurrence of each of the feature vectors according to the probability density function; determining an offset corresponding to each of the probabilities; normalizing each of the offset quantities to obtain a corresponding standard value; determining whether each of the feature vectors is abnormal according to each of the standard values and a preset threshold value. The scheme has wide adaptability.

The invention discloses a copy selection method based on kernel density estimation, belonging to the computer network technology field. The method comprises the following steps: divide copies in a network into old copies and new copies, for old copies, select a best old copy according to history data by utilizing a kernel density estimation policy; for new copies, select a best new copy accordingto current bandwidth condition of a node in which the new copies exist; calculate and compare the best new copy and the best old copy, therefore, select a best copy from multiple copies which have a same logical file name. The copy selection method based on kernel density estimation is suitable to a dynamic low-side network and is especially suitable to the condition with frequently changing network state. User access delay and bandwidth consumption can be reduced by the method and the network performance is increased.

The invention discloses a method for estimating lightning density distribution and annual average ground flash density (NG) values by kernel density estimation. The method comprises the following steps of: 1, constructing and maintaining a thunder and lightning space database and establishing a thunder and lightning real-time positioning monitor system by combining the geographic information system (GIS) technology to acquire the source of information on the occurrence of thunder and lightning, and displaying the positioning distribution of the lightning in real time in a space element mode; 2, acquiring the distribution of regional lightning density by a kernel density estimation process, and cutting graph regions which are selected by users and are used as condition layers to form a regional density distribution map and eliminate an edge effect; and 3, calculating a total regional area (kilometer) and annual total lightning time of the kernel density estimation to solve the NG values of the corresponding regions and form a weather product which combines the lightning density map and the numerical values, and releasing and sharing a thunder and lightning data analysis product by a sharing network finally. The fundamental aim of thunder and lightning service is fulfilled by the lightning density statistical method.

The invention discloses a solar collector output probability modeling method based on non-parametric kernel density estimation. The method comprises the following steps of: 1. obtaining hourly meteorological data measured in history and counting according to different years; 2. solving the output value and the bandwidth value of the unit area collector through a non-parametric kernel density estimation model of the output of the solar collector; 3. modifying the nonparametric kernel density estimation model. The method of the present invention can accurately reflect the random characteristicsand variation law of the output of the collector under different time scales, and can be used for planning and reliability research of the photothermal system under the condition of lacking historicaloperation data in the early stage.

The invention relates to a kernel density estimation-based license plate character segmentation method. At present, the conventional methods have the defect that binaryzation cannot be performed well under the extreme condition. The method comprises the following steps of: performing preprocessing of normalization, edge sharpening and noise removal on a license plate image; determining a character area of a license plate; finding a pixel value of which the probability of occurrence is maximum in the current kernel probability density curve and the kernel density half width of the point, and performing image binaryzation by utilizing the two parameters; and extracting a segmentation result of which the score is maximum within the width range to be used as a final character segmentation result. By the method, the distribution of character pixels can be determined accurately; and compared with the common iteration binaryzation method, the method has the advantage that the ambient interference resistant capacity is enhanced greatly.

The invention provides a method for correcting wind power data based on nonparametric kernel density estimation. Through analyzing a scatter diagram of wind speed and power, the confidence degree processing is performed on a power value within a small measured wind speed interval, and data beyond a given confidence interval is corrected to be within the confidence interval. The method provided by the invention does not need to consider the original data distribution situation and has very good inclusiveness of the distribution law of the data.

The invention relates to an electrical power system robust state estimation method based on self-adaptive kernel density estimation, comprising the following steps: (1) establishing a state estimation mathematical model based on a self-adaptive kernel density estimation theory; (2) acquiring self-adaptive bandwidth; and (3) according to the state estimation mathematical model in step (1) and the self-adaptive bandwidth obtained in step (2), performing electrical power system robust state estimation. Compared with the prior art, the invention ensures data measurement redundancy and system observability, eliminates residual error contamination and residual error flooding, and further, improves identification accuracy and convergence of the state estimation. The invention has advantages of being strong in robustness, high in computation accuracy, wide in applicability, great in flexibility, and the like.

The invention relates to a determination method for emergent environment event emergency disposal limit values and specifically discloses a nonparametric kernel density determination method for emergent environment event emergency disposal limit values. The method comprises following steps: step 1. toxicity data screening; step 2. Gauss kernel function window width selecting; step 3. Gauss kernel function calculating of toxicity data; step 4. nonparametric kernel density estimation and determination of the toxicity data of distribution of species sensitivity based on Gauss kernel function; step 5. model verifying; and step 6. environment concentration limit values determination. According to the method of the invention, the fitting method of distribution curves of species sensitivity is improved; due to the improvement, the distribution features of toxicity data can be well estimated unbiasedly and accurate environment concentration limit values can be obtained so as to realize standard determination in emergent environment event emergency disposal.

The invention discloses an improved density peak clustering-based social network community discovery method. The method comprises the following steps of: firstly calculating two indexes for each userin a network, wherein the two indexes comprise local densities and relative distances, the local densities are calculated by adoption of Gaussian kernel density estimation, and the relative distancesrepresent a distances between users and points which are greater than the users in the aspect of density and which are close to the users; selecting a point which has a large local density and relatively large relative distance as a community center on the basis of Gaussian distribution, and distributing the residual non-center points to communities of points which are greater than the non-centerpoints in the aspect of density and which are closest to the non-center points; and finally, measuring distance between every two communities on the basis of combination factors, wherein the communities, the combination factors of which are greater than a given threshold value, are combined into one community. Compared with the prior art, the method is capable of discovering spherical and non-spherical community structures in social networks at the same time, so that fewer parameters are needed under the premise of obtaining relatively high correctness and then the problem of clustering communities with any shapes is solved.

The invention relates to a kernel density estimation-based non-invasive power load identification method. The method comprises the following steps of: selecting a common household power load as a research object, acquiring power consumption data of the research object, carrying out sub-state division and extraction power distribution; generating a household working state set according to the power distribution, and calculating simulation power consumption data under each state; carrying out kernel density estimation to obtain probability distribution reference model of each state simulation data; identifying household working state transition points in the reference models, and dividing each household working state data segments; and for each data segment, searching a household working state which is closest to the probability distribution of the data segment, and comparing the household working state with the probability distribution so as to complete an identification task. According to the method, the main data features of power load power consumption can be effectively extracted, the main data distribution features are highlighted, and the influences of random power consumption data and abnormal fluctuation are weakened, so that effect can be well decomposed in the aspect of non-invasive identification, and the method is suitable for the changing and complicated working environment of the current household power grid.

The invention relates to a converter fault diagnosis method based on kernel density estimation. The method comprises the steps of: performing pre-processing of collected data through cubic B-spline wavelet analysis based on a mallat algorithm to obtain samples with fault features; employing a KDE fault classifier to perform offline training to select better parameters of the fault classifier and accurately dividing the normal conditions and each type of fault condition included in the training samples, and using the better parameters into a classifier network to obtain the optimal parameters;implanting the classifier network with the optimal parameters into online simulation to perform real-time online monitoring fault diagnosis of an actual circuit; and allowing the classifier network with completion of optimal parameters to distinguish known fault type samples and normal samples, complete the location of the known fault types of faults and identify the unknown faults for achievementof circuit protection in a condition of generation of unknown types of faults. The converter fault diagnosis method based on kernel density estimation can determine the health condition of the converter more accurately and more reliably, and also can improve the efficiency of the fault diagnosis of the converter.

The invention discloses a target tracking method based on kernel density estimation, wherein the method relates to the field of computer vision. The method comprises the following steps of firstly foradapting target form change in a vision tracking process, establishing a color distribution model for an H component of the target in an HSV space by means of kernel density estimation, so that tracking can be accurately finished on the condition that local shielding of the target occurs, and then performing target tracking by means of a CamShift algorithm. The target tracking method is an algorithm with relatively high robustness and is suitable for the environment which has high background change stability, single color and color distribution which is over-complicated. Compared with a gray-based template matching method, the target tracking method has higher accuracy.

The invention provides a probability forecasting method and system for wind power, and the method comprises the steps: extracting a wind speed sample from a pre-established wind speed probability density function model, inputting the wind speed sample into a pre-established wind speed and wind power conversion model, and obtaining the sample data of the wind power; performing kernel density estimation on the sample data of the wind power to obtain a kernel density estimation fitting probability density function of the wind power; and based on the kernel density estimation fitting probability density function, extracting a probability forecasting result from a preset confidence interval. The method comprises the following steps: extracting a wind speed sample from a pre-established wind speed probability density function model and inputting the wind speed sample into a pre-established wind speed-wind power conversion model; according to the method, the weather uncertainty information provided by the ensemble forecasting is fully utilized, the continuous probability density distribution function reflecting the wind power uncertainty can be provided, and the accuracy of the probability forecasting result is effectively improved by utilizing the set confidence interval.

The invention provides a capsule network rapid routing method based on kernel density estimation and mean shift. The method comprises the steps of giving a probability density function for a multi-clustering problem according to a kernel density estimation formula, modeling the dynamic routing process into an optimization problem, solving the optimization problem by utilizing a mean shift algorithm, and taking a solving result as the output of a capsule network to obtain a final classification result. Compared with the prior art, the dynamic routing time efficiency is greatly improved, at the same time, the identification performance of the capsule network is also improved.

The invention discloses a virtual sample generation method based on independent component analysis and kernel density estimation. At a system operation initial stage, the number of training samples isinsufficient, a kernel density estimation method is used, the overall probability density function of a sample is estimated through the probability density function of a small number of samples, whenthe attributes of an original sample have the correlation, the correlation between the attributes of the original sample is removed through an independent component analysis method, then kernel density estimation is conducted, and a virtual sample is generated according to the probability density function obtained through estimation. The problem that training samples are insufficient when the machine learning model is trained can be relieved, and the accuracy of the machine learning model is improved. Compared with other virtual sample generation methods, an independent component analysis method is introduced to solve the problem that all attributes of the sample have correlation, so that the application range of the method is widened.

The invention relates to a sign-in position prediction method based on personalized hierarchical kernel density estimation, and belongs to the technical field of data analysis. The method comprises the following steps: S1, establishing binary kernel density estimation based on a geographic space by utilizing extracted sign-in position data; S2, constructing kernel density estimation of the adaptive bandwidth, and selecting respective bandwidth for each data point; S3, constructing personalized hierarchical kernel density estimation; And S4, calculating a parameter value by using a gradient descent algorithm. According to the method, personalized sign-in prediction is provided for the user, meanwhile, the problem of data sparsity caused by too few sign-in data is solved, the method is closeto the actual life, and the prediction result is more accurate.

The invention discloses a virtual sample generation method based on kernel density estimation and a Copula function. The method includes: obtaining an original sample set and an original training set,constructing an initial classification model according to the original sample set and the training set, obtaining a probability density estimation function of the original sample set according to a kernel density estimation method and positive samples in the original sample set, obtaining Copula model parameters according to a maximum likelihood estimation method; and constructing a joint densityfunction of the positive samples according to the Copula model parameters, obtaining a virtual sample set through re-sampling by using the joint density function, and determining the generation number of the virtual sample set according to the difference between the data volume of the negative samples and the data volume of the positive samples in the original sample set. According to the technical scheme provided by the invention, different types of data distribution conditions of the original data set can be effectively improved, and the classification effect of various classifiers under the unbalanced sample condition can be improved, so that the generalization capability of the classifiers is improved.

The invention relates to a small target detection-orientated image threshold segmentation method adopting fast kernel density estimation. The method comprises the steps of: reading in an image, obtaining a greyscale matrix of the image in a computer, and setting a parameter gate value; taking pixel points with the same greyscale in the image as a set, and if the number Ni of the pixel points in the image is greater than the gate, then using a FRSDE (Fast Compression Set Density Estimator) for compressing; or else, then using an RSDE (Compression Set Density Estimator) for compressing; and establishing a relation matrix M to represent the interrelations among different greyscales on the image. The problem of extreme value evaluation for a target function is converted to be the problem of minimization sum evaluation for elements based on a matrix region, so that the optimal threshold is obtained. Compared with the prior art, the method disclosed by the invention has the advantages that the process is simple, the realization is easy, the robustness is good, high in the solution efficiency is high, and the like. Therefore, via the method disclosed by the invention, feasible scheme is provided for the problem of small target detection for a high-definition image; and simultaneously, an effective technical basis is provided for detection for a small target image in a complex background.

A method of acquiring a probability static state voltage stabilization margin based on Quasi monte carlo simulation and kernel density estimation is disclosed. The method comprises the following steps of according to power grid data, carrying out pretreatment to acquire an input random variable matrix; and then using a diffusion-based kernel density method (DKDM) to acquire a probability density of a voltage stabilization margin and cumulative probability distribution. In the invention, through introducing the Quasi monte carlo simulation, an input random variable sample is acquired so as to increase calculating efficiency of a simulation method; the diffusion-based kernel density method is used to accurately acquire a probability distribution function of the stabilization margin; and only a small sampling scale is needed and high calculating precision can be acquired.