61 results about "Resampling" patented technology

The invention discloses a FastSLAM algorithm based on an improved resampling method and particle selection. The algorithm comprises the following steps: (1) a robot predicts robot poses according to input control data and updates the robot poses and road signs according to measuring data combined with a measuring model of an external sensor of the robot; (2) the robot poses are predicted by the calculation of a particle filter, and a particle resampling criterion is amended according to an effective particle number, particle weighting covariance and particle-measuring residual consistency; (3) new particles are generated by using an index grad method and a crossover operator; (4) the robot is positioned, and a map is created according to the generated new particles. The invention improves the particle resampling criterion and a new particle-generating method in the FastSLAM algorithm, thereby obviously improving the estimation consistency of the FastSLAM algorithm to the robot poses and simultaneously improving the precision of the robot positioning and map creation.

The invention discloses a model-cluster-analysis-thought-based variable selection method suitable for a support vector machine. According to the method, sub-datasets are acquired from a full-spectrum data matrix by carrying out Monte Carlo sampling, an SVM (support vector machine) sub-model is established for each sub-dataset and is predicted and classified, then Mann-WhitneyU examination is utilized for carrying out statistic analysis on the prediction accurate rate of all the sub-models, and available vectors remarkably acting on the model prediction capability are sorted out; a once modeling result is not taken as the basis, but data information is maximally and effectively utilized by carrying out replaceable resampling, the internal relation among the vectors in the datasets are adequately investigated, and the statistical distributions of different results are analyzed, so that the method has high universality and stability.

The invention discloses a fault diagnosis method for a rotary machine based on angle resampling and an ROC (Receiver Operating Characteristic)-SVM (Support Vector Machine) and belongs to the field offault diagnosis of mechanical equipment. The method comprises the steps of eliminating fluctuation of a rotation speed by use of an angle resampling technology; performing characteristic value extraction from the dimensions of a time domain and a time-frequency domain; and implementing characteristic selection and fault diagnosis of the rotary machine by use of the ROC-SVM. According to the faultdiagnosis method for the rotary machine based on angle resampling and the ROC-SVM, the change of the number of sampling points of a vibration signal in unit time caused by the fluctuation of the rotation speed can be effectively eliminated by use of the angle resampling method, thereby improving the quality of the subsequent extraction characteristic value; the time domain and the time-frequency domain are combined to achieve wider characteristic extraction and obtain sufficient vibration signal information; the characteristic selection and fault diagnosis are performed by use of the ROC-SVM,the best characteristics are selected to prevent poor characteristics from reducing the effect of a fault classifier; the accuracy and effectiveness of the bearing fault diagnosis can be improved, thediagnosis speed can be accelerated, and a new concept is provided for solving the problem of the bearing fault diagnosis.

The invention discloses the designing of current-statistical-model-based probability hypothesis density particle filter and the current-statistical-model-based probability hypothesis density particle filter. An observed value of the filter is connected with the first input end of an updating circuit. The first input end of a prediction circuit is connected with the first output end of a state estimation circuit, and the output end of the prediction circuit is connected with the second input end of the updating circuit. The output end of the updating circuit is connected with the input end of the resampling circuit. The first output end of the resampling circuit is connected with the second input end of the prediction circuit, and the second output end of the resampling circuit is connected with the input end of the state estimation circuit. By the invention, a hardware circuit realization scheme for the current-statistical-model-based probability hypothesis density particle filter is designed based on the theory of the current-statistical-model-based probability hypothesis density particle filter, and simulation results show that the tracking performance of the designing of the current-statistical-model-based probability hypothesis density particle filter and the current-statistical-model-based probability hypothesis density particle filter is similar to that of theoretical analysis and can be used for tracking problems about maneuvering multi-target movement in a clutter environment.

The invention discloses an AAKR model uncertainty calculation method and system based on resampling, and the method comprises the steps: dividing a historical state data set of a sensor into a training data set and a testing data set, carrying out the denoising on the training data set through a wavelet denoising method, calculating a noise variance, improving the data precision, randomly selecting and replacing the historical state data of the sensor to obtain a new training data set sample so as to optimize the AAKR model architecture and the change among the plurality of model prediction values to obtain the model prediction variance of the plurality of model prediction values, and calculating the mean square error between the prediction values and the test values by utilizing Bootstrapresampling training data. Model deviation is calculated by combining prototype model variance, 95% uncertainty value is formed, modeling calculation of a noise estimation value by an empirical distribution model is not needed, the resampling process is simplified, the calculation efficiency is improved, confidence interval deviation is reduced by combining a Jackknife method, the reliability is guaranteed, and the estimation efficiency is improved on the basis of keeping convergence performance.

The system receives exemplary time-series sensor signals comprising ground truth versions of signals generated by a monitored system associated with a target use case and a synchronization objective, which specifies a desired tradeoff between synchronization compute cost and synchronization accuracy for the target use case. The system performance-tests multiple synchronization techniques by introducing randomized lag times into the exemplary time-series sensor signals to produce time-shifted time-series sensor signals, and then uses each of the multiple synchronization techniques to synchronize the time-shifted time-series sensor signals across a range of different numbers of time-series sensor signals, and a range of different numbers of observations for each time-series sensor signal. The system uses the synchronization objective to evaluate results of the performance-testing in terms of compute cost and synchronization accuracy. Finally, the system selects one of the multiple synchronization techniques for the target use case based on the evaluation.

The invention discloses an intelligent fault diagnosis method for a rotating machine at a time-varying rotating speed. Angular domain resampling is carried out on a time-domain vibration signal of a rolling bearing and a non-stable time-domain signal is converted into a stable angular domain signal, to so that the impact on the analysis of the vibration signal from the change of the rotating speedis eliminated; gabor transformation capable of well describing transient characteristics of drastically changing signals is used for a rotating speed estimation method based on time-frequency spectrum ridge line fitting; a keyless phase order tracking method without installing a rotating speed sensor is suitable for occasions where the rotating speed sensor cannot be installed. An LSTM model capable of adaptively extracting time sequence features without expert experience and domain knowledge is adopted, and due to the existence of a BN generalization layer, convergence of the model can be accelerated, over-fitting is prevented, the generalization ability of the model is improved, and adaptive intelligent fault diagnosis and identification of the rotating machine under the variable-rotating-speed working condition are realized.

The invention discloses a track cycle prediction method based on resampling, and mainly solves the problems that in track prediction in the prior art, the prediction duration is short, and the prediction error is too large due to the fact that the motion state of a target changes in the future. According to the implementation scheme, the method comprises the steps of simulating a historical track and a future track of a maneuvering target; filtering, resampling and normalizing the historical track data of the target in sequence; constructing a neural network model composed of a Bi-LSTM layer, a Dropout layer, a Dense layer and an activation layer, and training the neural network model by using the preprocessed track data; generating part of historical track data by using a loop strategy, and calculating the historical track data by using the trained neural network model parameters; and carrying out smooth filtering on a calculation result to obtain a final predicted track. According to the method, the prediction error is small, the prediction time is long, when the target motion state changes in the future, the accurate prediction track can still be obtained, and the method can be used for target tracking.

The invention discloses a cross-version deep defect prediction method capable of relieving a class overlapping problem. The method comprises 1, a deep semantic learning-oriented overall framework in cross-version software defect prediction; 2, a semantic feature learning model based on a convolutional neural network; 3, a hybrid nearest neighbor cleaning strategy for deep semantic learning. According to the method, a hybrid nearest neighbor cleaning strategy is adopted to relieve the class overlapping problem existing in semantic features learned by deep learning. Specifically, for an abstractsyntax tree corresponding to a source code, a convolutional neural network is adopted to learn deep semantic features, and then a hybrid nearest neighbor cleaning strategy is adopted to perform resampling and data cleaning on a labeled data set. By adopting the hybrid nearest neighbor cleaning strategy, the class imbalance problem and the class overlapping problem can be processed, and the statistical analysis result of the data shows that the strategy can improve the performance of software defect prediction based on deep semantic learning.

The invention discloses a wind power probability prediction method based on hierarchical integration. According to the method, a subspace set is constructed through resampling and a partial least square method, a plurality of local areas are obtained on each subspace through GMM clustering, a corresponding local GPR model is established, and a Bayesian reasoning strategy and a finite mixing mechanism are used for fusing the local models to establish a first-layer integrated model. And a genetic algorithm is adopted to select a suitable first-layer integration model for selective adaptive integration, so that a selective hierarchical integration Gaussian process regression probability prediction model can be obtained. In order to solve the problem of performance deterioration caused by change of wind power data characteristics, an adaptive updating strategy is introduced, so that the prediction model has adaptive updating capability. According to the method, the selective hierarchical ensemble learning framework is used for ultra-short-term wind power prediction, compared with a traditional ensemble learning prediction method, the method has higher prediction precision and stability, and the generated prediction interval can provide effective reference for power dispatching.

The invention provides a wave compensation prediction method based on a random forest algorithm and an Adam neural network. The method comprises the following steps: S1, collecting historical data ofa ship, and carrying out the normalization of the data, and obtaining an original data set; s2, extracting samples by utilizing a Bootstrap resampling method to form a plurality of training set samples; s3, calculating feature importance according to the random forest, and performing attribute screening; s4, constructing a neural network model, and training the neural network model by using an Adam algorithm according to the screened attribute samples as input quantities and the ship prediction values as output quantities; and S5, inputting the real-time monitoring data into the Adam neural network model to obtain a predicted value.

The invention provides an Internet of Things equipment working condition data analysis method and device and computer equipment, and the method comprises the steps: receiving original working condition data, carrying out the resampling of the original working condition data at a preset frequency, and generating the working condition data; splitting the working condition data in the preset time period into at least one working condition data segment with the same data change trend in the segment by utilizing a preset algorithm; performing operational analysis on the at least one working condition data segment by utilizing a preset physical model to obtain an analysis result, the preset physical model comprising an operational rule when the target equipment runs; writing the generated working condition data of the preset frequency into a cache in an offline state; and regularly executing the step of splitting the working condition data in the preset time period into at least one workingcondition data segment with the same change trend by utilizing a preset algorithm and subsequent steps. According to the invention, the accuracy of positioning the working state of the target equipment can be improved, and the accuracy of analyzing the working state of the target equipment is improved.

The invention discloses a self-service capacity expansion method based on a correlation coefficient criterion. The method comprises the following steps: S1, initializing parameters, and setting a correlation coefficient threshold value rho epsilon, a positive integer M, a self-service sample capacity B and an equal fraction m of a histogram function; S2, copying a new sample, generating a positiveinteger R-U (0, M) obeying uniform distribution randomly, calculating a remainder p = mod (R, n), wherein n is the sample capacity; S3, calculating the mean value of the copied samples, if the ith new sample is the ith new sample, repeating S2 to obtain a group of copied samples, calculating the mean value of the copied samples; S4, obtaining self-service resampling samples, and repeating the steps S2 and S3 to obtain the similarity rho (f (x))) of the self-service resampling samples to calculate the resampling samples and the original samples; f (x *)) if rho (f (x); and if f (x *) is greater than or equal to rho epsilon, outputting a resampling sample, otherwise, repeating the steps S1 to S4 until a similarity condition is satisfied. The resampling sample obtained by the method fully utilizes the information of the given sample, and the obtained resampling sample is closer to the real situation.

The invention discloses a garbage classification model training method and a garbage classification method and device, and the method comprises the steps: carrying out the resampling of a pre-labeled garbage image sample set, and obtaining a balanced sample set; based on the junk image sample set and the balanced sample set, training a preset initial smooth perception model until the initial smooth perception model reaches a first preset convergence condition, and obtaining a target smooth perception model; performing iterative training on a preset first garbage classification model based on the garbage image sample set to obtain a first loss value, and performing smoothing processing on the first loss value by using the target smoothing perception model to obtain a second loss value; and finally, according to the second loss value, model parameters of the first garbage classification model are updated until the first garbage classification model reaches a second preset convergence condition, a target garbage classification model is obtained, and the accuracy of a classification result of the model is improved.

The invention relates to an unbalanced data resampling method for ecological environment evaluation, which comprises the following steps of: 1, acquiring ecological environment data, and standardizing the ecological environment data of each instance; 2, calculating the data density of each instance and the quantity difference d between the majority class and the minority class; 3, calculating the distribution unbalance degree DI and the number unbalance degree IR of the minority class and the majority class; 4, performing oversampling on the minority class; and 5, undersampling the majority classes. The method has the advantages that the data density of each instance in the data set is used for measuring the uniformity degree of distribution, and oversampling and undersampling are carried out according to the unbalance degree of data distribution, so that the purpose of balancing data is achieved; the ecological environment data is further balanced, the quality of the data set is improved, and when the ecological environment quality evaluation is carried out, the accuracy is higher, and the true positive rate is increased.

The invention provides a rock-soil sampling layout system based on artificial intelligence analysis. The system comprises a rapid pre-sampling engineering vehicle and a re-sampling static penetrometer, wherein the rapid pre-sampling engineering vehicle comprises a power cone penetrometer, a data processing engine and a visual interface; the power cone penetrometer carries out penetration samplingon a predetermined rock-soil target area, and an optical fiber sensor sends data obtained by each time of penetration sampling to the data processing engine; after the data processing engine analyzesthe sampling data through an AI data extension engine, recommended resampling layout points in the predetermined rock-soil target area range are displayed on the visual interface; and the re-samplingstatic penetrometer performs rock-soil sampling on the recommended resampling layout points. The invention also discloses a rock-soil sampling layout method realized based on the system and a computerreadable storage medium for realizing the corresponding method. By means of the technical scheme, the rock-soil sampling data can be more representative.

The invention discloses an adaptive resampling method based on window function design. The adaptive resampling method comprises the following steps: firstly, establishing a universal resampling modelbased on a Farrow structure; establishing a fractional delay filter design model based on a window function method, and obtaining a plurality of groups of fractional delay filters by utilizing the model; constructing a fractional delay filter matrix on the basis of the obtained multiple groups of fractional delay filters; solving a coefficient matrix C of the Farrow structure sub-filter by using aleast square method; and substituting the obtained Farrow structure sub-filter coefficient matrix C into the resampling model established in the step 1 to complete the establishment of a universal Farrow structure resampling model. The method is better in flexibility and smaller in calculated amount, a proper model can be obtained by adjusting parameters for multiple times, sampling at any pointis achieved, and the function of arbitrary conversion between sampling rates is achieved.

The invention relates to a hybrid framework-based unbalanced classification method, system and equipment and storage medium. According to the method, an unbalanced network anomaly detection data set is used for verifying a hybrid resampling integrated model. The number of majority classes is reduced by providing the combination of resampling methods, so that the processing speed is increased. The unbalanced dataset is processed at the data level, and the dataset is converted into equilibrium distribution using a resampling technique. An integrated model comprising 12 different classifiers is established, and compared with 5 classifiers in previous work, more choices are provided. The slightly balanced data obtained after the processing is classified by using an integrated model, so that a novel combination of undersampling and oversampling is provided to balance the imbalance among different data categories, and the processing speed is increased with less memory overhead.