198 results about "Principal component analysis algorithm" patented technology

A system and method for acoustic detection of coronary artery disease (CAD) are provided. The system includes a transducer for acoustically detecting heart signals of a patient and a computer system which executes detection software for processing the detected heart signals to identify the presence of CAD from the heart signals. The software allows for the automatic definition of a diastolic “window” of the acoustic signal for analysis, and automatically edits the sampled acoustic signal to eliminate unwanted artifacts and / or noise in the acoustic signal. The edited signal is then processed by a plurality of signal processing algorithms, including spectral analysis algorithms, time-frequency algorithms, global feature algorithms, kurtosis algorithms, mutual information algorithms, negenthropy algorithms, and principal component analysis algorithms, to generate a disease vector. The disease vector is then classified to determine whether CAD is present in the patient. Classification can be accomplished using linear discriminant analysis or a support vector machine.

The invention discloses an indoor passive positioning method based on channel state information and a support vector machine. The method comprises the following steps: firstly preprocessing the acquired channel state information data, performing de-noising and smoothness through the adoption of a density-based spatial clustering of applications with noise and a weight-based moving average algorithm, and then using the principal component analysis algorithm to extract the features. The data after the preprocessing and feature-extracting can reflect the signal change more accurately and the dimension is greatly reduced. The passive positioning adopts two-stage positioning. In the training stage, the large positioning space is divided into sub-regions, the support vector machine classification and regression model is established for each sub-region so as to acquire a statistic model for accurately representing the nonlinear relationship between the position and the signal. The two-stage positioning firstly determines the sub-regions through the classification of the support vector machine, and the precision position is determined in the sub-region through the regression of the support vector machine. The method disclosed by the invention has the beneficial effects that the passive positioning can be performed in the absence of the active participation of the target, and the indoor positioning precision is improved to sub-meter level.

The invention discloses a face recognition method based on PCA image reconstruction and LDA and belongs to the technical field of computer image processing and pattern recognition. The face recognition method is based on a principal component analysis algorithm; an intra-class covariance matrix serves as a generation matrix for acquiring face feature subspaces of individuals; then an image to be recognized maps the feature subspaces to extract features; the image is reconstructed according to feature values; a residual image is subjected to the linear discriminant analysis; and finally, face recognition is realized by a minimum distance classification recognition algorithm. Compared with the prior feature subspace method, the face recognition method can better extract the face features of different people, and the face recognition rate is increased effectively. In addition, when a face database is required to be expanded, only the newly added faces are required for feature face training; not all the face feature subspaces are retrained; and the face recognition method also has good extendibility.

The invention discloses a bearing life degradation performance evaluation index construction method based on principle component analysis. First, time domain feature extraction, frequency domain feature extraction and time domain and frequency domain feature extraction are conducted on bearing degradation process full-service-life vibration data collected in each time, so that the degradation tread feature of a bearing is comprehensively extracted. Then, weighing fusion is conducted on the original feature through a principle component analysis algorithm to achieve feature simplification. A simplified feature index has the advantages of representing state information of a rolling bearing and further effectively eliminating original multi-dimensional feature inter-information redundancy, so that a rolling bearing service life degradation performance evaluation index based on feature space weighing fusion is effectively constructed, the shortcomings that a traditional evaluation index is insensitive to initial failures and is low in universality are overcome, and the degradation tendency of the bearing can be well presented.

The invention provides an adaptive radiation source modulation identification method based on time-frequency analysis. The method comprises the steps of I, carrying out time-frequency analysis on a received radiation source signal by use of time frequency distribution, converting the radiation signal from a time-domain signal to a time-frequency two-dimensional image; II, reducing computation complexity and characteristic dimension by use of an image processing technology, and improving the proportion of signal characteristic information in the image through normalization, binaryzation, image thinning and image preprocessing operations; III, carrying out image shape characteristic extraction on the preprocessed image in combination with a second-order and four-order moment estimation method by use of an adaptive principal component analysis algorithm; and IV, identifying a modulation mode of the radiation source signal by use of an LIBSVM (Library for Support Vector Machine) classifier. According to the adaptive radiation source modulation identification method based on time-frequency analysis, the characteristic missing of a low signal to noise ratio signal can be effectively avoided, and characteristic redundancy of a high low signal to noise ratio can be also avoided, and the modulation identification rate is not affected at the same time.

The invention provides an invasion detection method based on channel state information and a support vector machine. No special hardware facility is needed, an existing wireless network is fully used, and a common business router is used to realize security monitoring function. The coverage scope is wide, and privacy exposure can be prevented. The invasion detection method includes the steps of after obtaining CSI original data, conducting clustering and de noising for the subcarrier data in a channel by using a density-based clustering algorithm DBSCAN, smoothing the denoised data by using weight-based sliding average algorithm, and extracting characteristic values for data by using major constituent analyzing algorithm after data pre-processing. Data subjected to pretreatment and feature extraction can more accurately reflect the main change of signals and greatly reduce number of dimensions. The invasion detection precision is improved and the calculating complexity is reduced. The method uses an SVM classification algorithm to obtain a statistics model of non-linear dependence relation between an invasion state and a signal fingerprint, thereby achieving the purpose of invasion detection.

The invention provides an unmanned aerial vehicle (UAV) classification method and device based on dual frequency radar signal time-frequency distribution. The method comprises steps that S1, short-time Fourier transform is utilized to process time domain data of each UAV acquired by a dual band radar system, and time-frequency maps of two bands of each UAV are acquired; S2, characteristic extraction of the time frequency maps of two bands of each UAV is carried out through a principal component analysis algorithm; and S3, for each UAV, the characteristics of the two bands of the UAV are fusedto acquire fusion characteristics, each fusion characteristic is taken as a sample to input to a support vector machine to classify the UAV. The method is advantaged in that the dual band radar systemis utilized to emit electromagnetic waves in different bands to the UAV, characteristic extraction of the micro Doppler information of UAV echo is carried out, then dual band characteristic fusion analysis is carried out, so different UAV categories are acquired, and UAV classification accuracy is improved.

The invention discloses a neural network face recognition system based on a memristor. The neural network face recognition system comprises a face capture module, a preprocessing module, an input module, a memristor neural network module, an output module and a weight updating module. The face capture module is used for capturing a face picture in the picture; the preprocessing module is used forcarrying out dimension reduction processing on the face picture; the input module is used for converting the picture subjected to dimension reduction into an electric signal; the memristor neural network module is used for storing network weights, carrying out matrix vector multiplication operation on the electric signals and transmitting an operation result to the output module; the output moduletransmits the operation result to a weight updating module for weight updating, and transitting the updated weight to a memristor neural network module, and the output module reads an identificationresult of the network; the memristor neural network module is composed of a memristor array. The structure scale of the memristor neural network is reduced by utilizing a principal component analysisalgorithm, so that the operation speed is increased, the operation energy consumption is reduced, and the hardware cost is reduced.

The invention discloses a fault detection method based on a distributed nonlinear dynamic relation model, and aims to establish distributed nonlinear dynamic relation models for respective measured variables, and to implement fault detection based on the distributed models. The main core of the method comprises establishing nonlinear dynamic relation models for respective measured variables by using a RBF neural network, and taking account of the self-correlation of the measured variables embodied at different sampling moments and the cross-correlation of the measured variables and other variables embodied at different sampling moment. Compared with a traditional method, the method of the invention, by using the RBF neural network algorithm, constructs nonlinear dynamic relation models embodied in different sampling moments for respective measured variables, and reflects the advantages and characteristics of distributed modeling. Secondly, the method of the present invention takes an error as an object to be monitored, and is very helpful for the subsequent establishment of a fault detection model by using a principal component analysis algorithm. Therefore, the method of the present invention is more suitable for the fault detection of dynamic processes.

The invention discloses a mobile phone glass cover plate defect detecting method based on principal component analysis. The defect detecting method includes the six steps of contrast enhancement of a mobile phone glass cover plate image, image binaryzation, outlier rejection, defect recognition based on PCA, edge detection and detection result output, the contrast of defects and a background is enhanced by preprocessing the mobile phone glass cover plate image, influence of dust, electronic noise and the like can be avoided, the defects are recognized by adopting a principal component analysis algorithm, the result of defect types is output, and edge detecting and marking are performed on the defects by adopting an edge extraction algorithm. The method remarkably improves detection efficiency and overcomes the defects that manual visual detection consumes time, is tedious and highly depends on detection experience.

The invention discloses a user behavior abnormality detection method under a Hadoop cluster. According to the method, user behavior data is acquired and analyzed and is recorded in a form of logs, feature vectors are formed according to behavior features of users, a feature vector set is processed by utilizing a parallel principal component analysis algorithm, behavior modes of the users are acquired efficiently, and by comparing the behavior modes of the users with historical modes, the abnormal behavior problem generated when the users access a HDFS is found and simultaneously, the hidden security threat under the Hadoop cluster is found, so that an effect of ensuring security of the HDFS is achieved. According to the user behavior abnormality detection method disclosed by the invention, not only is effective monitoring established for the data access behaviors of the users, abnormal behaviors are timely found and data security of the Hadoop cluster is ensured, but also by the parallel principal component analysis algorithm, model training efficiency is improved, and the problem of low training efficiency of a conventional model is solved.

Disclosed are a shadow detection method and device. The method includes a step of obtaining a depth / disparity map and color / grayscale image from a two-lens camera or stereo camera; a step of detecting and acquiring plural foreground points; a step of projecting the acquired plural foreground points into a 3-dimensional coordinate system; a step of carrying out, in the 3-dimensional coordinate system, a clustering process with respect to the projected plural foreground points so as to divide the projected plural foreground points into one or more point clouds; a step of calculating density distribution of each of the one or more point clouds by adopting a principal component analysis algorithm so as to obtain one or more principal component values of the corresponding point cloud; and a step of determining, based on the one or more principal component values, whether the corresponding point cloud is a shadow.

The invention discloses a nondestructive testing method for Fushi apple quality based on a near infrared spectrum. The method comprises the following steps: 1) selecting a batch of intact Fushi apples as experiment samples; 2) collecting a near infrared spectrum of the apples; 3) utilizing a principal component analysis algorithm for performing dimension reduction treatment on near infrared spectrum data of the apple samples; 4) measuring the acidity and sugar degree values of the apple samples and recording; and 5) utilizing the pretreated near infrared spectrum data and acidity and sugar degree values, and combining with a neural network algorithm for establishing an apple acidity and sugar degree prediction model, thereby realizing the detection for the interior quality of the apples. According to the invention, the acidity and sugar degree values of the apples can be quickly forecast and the forecast value thereof can be directly displayed, the forecasting precision can meet the practical production requirement, the experiment system is simple in operation method, and the nondestructive testing method can be used for online detection and classification of the interior quality of the apples.

The invention provides a method for identifying the human face organs in a location image, a device and a video processing chip. The method comprises the following steps of: adopting a principal component analysis algorithm to establish a statistic model in shape of a human face organ and adopting a gray information searching method to carry out a preliminary location for the human face organs in the identified image; on the basis of the statistic model, adopting a human face edge information searching method to determine and adjust the contour point on the chin; changing a red, green and blue mode into a hue saturation mode for the color space of the identified image, and determining and adjusting the contour point in the lips through a chromatic value searching method on the basis of the statistic model; and determining the position of the human face organ and finishing the identification process of human face organs according to the contour points of the human face organs in the identified image. Based on the preliminary identification of human face organs through the prior art, the invention adopts an edge information method for independently processing the contour point of the chin and adopts a color space to process the mouth area, thereby identifying and positioning human face organs more precisely.

The invention discloses a dynamic process monitoring method based on nonlinear autocorrelation rejection of variables, and aims to reject nonlinear autocorrelation of each measured variable of sampledata and establish a nonlinear dynamic process monitoring model on the basis. The method comprises steps as follows: firstly, a nonlinear regression model between the sample data and multiple delay measurement data of the sample data with a kernel partial least-squares-algorithm; secondly, a model error is taken as a new monitored object, the process monitoring model is established with a PCA (principal component analysis) algorithm, and fault monitoring is implemented. Compared with a traditional method, the method has the advantages that the error is taken as the monitored object, the error's capacity of reflecting the abnormal change condition of nonlinear autocorrelation characteristics is utilized, furthermore, by means of absence of time sequence autocorrelation of error data, convenience is provided for following establishment of the process monitoring model based on the PCA algorithm. Therefore, the method is more suitable for the dynamic process monitoring.

A method for compressing and synthesizing a bidirectional texture function realizes compression and synthesis of the bidirectional texture function data by principal component analysis and color rectangular texture blocks. The method comprises the following steps; firstly the bidirectional texture function data are grouped according to different viewpoints and compressed by the principal component analysis algorithm; and then data blocks are immediately extracted from the compressed bidirectional texture function data by an error calculation method, one color rectangular texture block is formed by every four data blocks, and multi-dimension texture synthesis is carried out by the color rectangular texture blocks; and at last, the compressed and synthesized bidirectional texture function data are drawn on a 3D model surface. The method can compress large-scale bidirectional texture function data and synthesize new multi-dimension texture images according to the existing texture images. The method has the advantages of fast synthesis and precious effect, solves the consistency problem of the multi-dimension texture synthesis, expands means of the multi-dimension texture sources, and meets the realistic drawing requirements in a virtual environment.

The invention discloses a dual-source image feature extraction and fusion identification method based on a convolution neural network, which comprises the following steps: utilizing the characteristics of the convolution neural network with migration learning, training the convolution neural network model parameters through a large number of visible light databases; The trained model is used to automatically extract the hidden features of visible and thermal infrared target images, and the maximum desampling method is used to reduce the feature dimension. Combining Fisher discriminant method and principal component analysis algorithm, the dimension reduction and fusion of multi-source image features are carried out. Support vector machine classifier is used to classify and recognize the fusion features of the target image. A method for classify and identifying multi-source image target in unmanned aerial vehicle (UAV) platform features that image hidden features are extracted by convolution neural network, and Fisher discriminant method and principal component analysis algorithm are combined for dimension reduction and fusion of features, which provides a new and effective way forclassifying and identifying multi-source image targets based on feature level.

The invention provides a rolling bearing fault prediction method based on wavelet principal component analysis. The method mainly comprises the following steps: extracting a wavelet packet transform coefficient of a vibration acceleration signal of the rolling bearing, and calculating a multiresolution similarity coefficient entropy between the wavelet packet transform coefficient and the vibration acceleration signal of the rolling bearing; and applying a wavelet principal component analysis algorithm to carry out feature fusion processing on the multiresolution similarity coefficient entropy, obtaining the fusion feature measure of the vibration acceleration signal of the rolling bearing of a multiresolution state fusion space, and identifying a fusion interval of the rolling bearings under a normal state, a fault state and a hidden danger state. The rolling bearing fault prediction method adopts the PCA (Principal Component Analysis) algorithm to carry out hidden danger identification on hidden dangers and carries out positioning classification on the vibration signals of a motor bearing subjected to the hidden danger of heterology by a thought that multi-dimensional state principle components are extracted by the PCA algorithm on the basis of the separability of the vibration features of the hidden danger of the heterology of the bearing, and experiments prove the effectiveness of a bearing hidden danger monitoring and positioning method based on the multiresolution state fusion space.

The invention discloses a signal processing algorithm of a non-contact type vital sign monitoring system. The signal processing algorithm comprises the following steps of step 1, respectively carrying out heartbeat signal and breathing signal separation on the received signals of a path I and a path Q; step 2, extracting the main components of heartbeat and breathing by a main component analyzing algorithm; and step 3, respectively carrying out smooth conversion and Fourier conversion on the heartbeat and breathing components obtained in step 2, so as to obtain the waveforms and the frequencies of heartbeat and breathing. The signal processing algorithm has the advantages that the effectivity and the reliability are realized, the vital sign signals can be monitored in a real-time way, the monitoring precision is high, and the real-time performance is high.

The invention discloses a data dimension reduction method based on a parallel principal component analysis (PCA) algorithm. The method comprises the steps of: S1, constructing a sample data matrix D<nxm> by high-dimensional data of which dimensions are to be reduced; S2, calculating a covariance matrix C<mxm> of the sample data matrix D<nxm>; S3, calculating m feature values of the covariance matrix C<mxm> and m corresponding feature vectors; S4, determining the number k of principal components according to the feature values and the feature vectors; and S5, utilizing the feature vectors, which correspond to the top-k greater feature values, to construct a transformation matrix, and utilizing the transformation matrix to calculate a principal component matrix, wherein the principal component matrix is data of which the dimensions are reduced. According to the method, the problem that according to a traditional stand-alone principal component analysis algorithm, the data cannot be loaded into a memory at once because a data size is too large is overcome, I / O operations are reduced, and the processing efficiency of data dimension reduction is improved.

The invention discloses a satellite gyrounit fault diagnosis method based on a principal component analysis algorithm, relates to a fault diagnosis for a satellite gyrounit, and particularly relates to a fault diagnosis method based on a Butterworth lowpass filter and the principal component analysis algorithm in order to overcome the shortcoming in fault diagnosis of the conventional principal component analysis algorithm and solve the problem of misreport. The method comprises the steps of 1, performing filtering preprocessing on output angular speed data Xp of the satellite gyrounit through the Butterworth lowpass filter; 2, constructing a PCA (principal component analysis) math model according to the preprocessed output angular speed data X of the satellite gyrounit; 3, detecting process data in a residual subspace by adopting a squared prediction error (SPE) statistical magnitude according to parameters of the PCA math model obtained in the second step; 4, diagnosing a fault position according to a process variable contribution figure after a fault is detected. The satellite gyrounit fault diagnosis method is applied to the field of fault diagnosis in a satellite gyrounit running process with observable process data.

The invention belongs to the field of picture processing, and provides a blood vessel bifurcation extracting method and system of a lung CT picture. The spherical characteristic of the blood vessel bifurcation on a geometrical structure and the accumulation performance of tensor voting are fully used, spherical tensor voting is carried out on pixel points, belonging to a blood vessel, on the reinforced lung CT picture, the partial maximum value of a spherical tensor significance coefficient is solved to obtain an alternative blood vessel bifurcation, then the tail end points of the blood vessel are removed through a main component analysis algorithm, and an accurate lung blood vessel bifurcation is obtained. The method and system do not need earlier stage operation of the picture, has high robustness on noise, provides important basis for later picture registration and lung movement, is favorable for later lung CT picture analysis, and assists in clinical diagnosis and treatment.

The invention discloses an area-of-interest detection method based on full convolution neural network and low-rank sparse decomposition. The method comprises steps of 1) carrying out super-pixel clustering on an original image, extracting color, texture and edge characteristics of each super-pixel and forming a characteristic matrix based on the characteristics; 2) in an MSRA database, based on agradient descent method, learning to obtain a characteristic conversion matrix; 3) in the MSRA database, by use of the full convolution neural network, learning to obtain high-level semantic prior knowledge; 4) by use of the characteristic conversion matrix and the high-level semantic prior knowledge, converting the characteristic matrix; and 5) by use of the robust principal component analysis method, carrying out low-rank sparse decomposition on the converted matrix, and according to sparse noise obtained through the decomposition, calculating a saliency map. According to the invention, themethod is used in an image preprocessing process, and can be widely applied in visual working field like visual tracking, image classification, image segmentation and target re-positioning.

The invention provides a multi-response parameter optimization method based on a radial basis function neural network prediction model and improved WPCA (weighted principal component analysis). According to the method, a non-linear prediction model of a production process is built by adopting a radial basis function neural network, capacity prediction indexes of the neural network model are introduced, a WPCA algorithm is adjusted, response with high prediction capacity receives priority in improvement in multi-response parameter design, and the optimization effect of technological parameters is improved. The WPCA generally adopts linear regression to establish a relation model between a response variable and a controllable factor variable in the multi-response parameter optimization design, however, the fitting degree of a linear regression model is not high for a complicated non-linear production process, and modeling requirements for parameter design cannot be met. The method is applied to the multi-response parameter optimization design of a thermal polymerization process of an aluminum-metallized polypropylene film capacitor, so that a satisfying comprehensive optimization effect of two responses of capacitance and loss tangent value of the capacitor is realized.

The invention discloses an online medicine-adding control method and system for wastewater treatment. The method comprises the following steps: obtaining current incoming water monitoring index data,and inputting the current incoming water monitoring index data into an online medicine-adding control model to obtain current optimum medicine-adding quantity; wherein the online medicine-adding modelis constructed through adopting a principal component analysis algorithm, a genetic algorithm and a neural network model algorithm; by the principal component analysis algorithm, the dimensionality of a training sample is reduced, the speed of a BP neural network model is improved, and the speed of medicine-adding quantity calculation is improved; by the genetic algorithm, the connection weight value and threshold value of the BP neural network model are optimized, the prediction accuracy of the BP neural network model is improved and is difficult to fall into local optimum; therefore, by adopting the online medicine-adding control model provided by the invention, the defect that the medicine-adding quantity is difficult to accurately determine by a medicine-adding mode in the wastewatertreatment process is overcome, and the rapid and accurate online medicine-adding control process is realized.

The invention provides a method for identifying a medical insurance fraud based on a principal component analysis algorithm. The method comprises the following steps: acquiring medical insurance basic data and generating a medical insurance structural data set; performing standard treatment on various data, thereby generating a standard matrix; calculating a covariance matrix of the standard matrix, solving a characteristic equation of the covariance matrix of the sample and confirming principal components; converting the standard index variable into principal component scores; respectively calculating the average value and the standard difference of the principal component scores; calculating an abnormal threshold value under each principal component dimension according to the chebyshev law; taking each principal component as a coordinate, drawing a two-dimensional space scatter diagram, representing a practical medical insurance account by each scatter point and regarding a medical insurance reimbursement account with the value which is more than the abnormal threshold value in step S5 as an abnormal account. According to the invention, the medical insurance data is cleaned and settled; a principal component analysis method is adopted for performing feature dimension reduction on the variable related to the fraud; the abnormal threshold value is calculated according to a statistic method; and the high risk in medical insurance fraud can be identified.

The invention discloses a building rubbish on-line sorting system and method. The building rubbish on-line sorting system comprises a material conveying device, a material detection device, an industrial personal computer, a sorting device and a plurality of recycling bins. The material conveying device is used for conveying materials to a detected area in a dispersed mode. The material detectiondevice collects pictures and spectral information of the materials in the detected area and further conducts image processing on the pictures to obtain geometrical characteristics of the materials, and the spectral information is processed through a principal component analysis algorithm and a machine learning algorithm to obtain the material types of the materials and is further fed back into theindustrial personal computer. The industrial personal computer is connected with the sorting device so as to control the sorting device to move according to the geometrical characteristics and the material types, and the materials are sorted into the corresponding recycling bins. By means of the building rubbish on-line sorting system and method, the materials made of different kinds of stuff canbe sorted into different places, no human interference is required on the whole production line, efficiency is high and accuracy is reliable.

The invention discloses a remote fault diagnosis and push method and device for a robot. The method comprises the following steps: acquiring a robot state signal and an operation environment state signal, and pre-processing the robot state signal and the operation environment state signal to obtain pre-processed signals; performing feature extraction on the signals based on time domain analysis and frequency domain analysis to obtain state feature data of the signals; fusing the state feature data of the multiple signals based on a principal component analysis algorithm to obtain operate statefeatures of the robot; substituting the operation state features of the robot into a fault diagnosis model stored in a fault identification model base, performing logic programming to judge that therobot has faults, obtaining a fault type of the robot, and pushing the fault type to a management user terminal to perform real-time display of fault warning. According to the embodiment of the invention, the remote fault diagnosis and push method can remotely diagnose whether the robot has faults or not, and pushes the faults to corresponding management staff, so that fault diagnosis efficiency of the robot is improved.

The invention relates to a method and a device for establishing a wheel / rail force load identification characteristic data model, wherein, the method comprises the following steps of: extracting time-frequency characteristics of track-vehicle system detection data by using an adaptive signal decomposition algorithm based on a variable parameter domain and a short-time Gaussian chirplet; fusing thetime-frequency characteristics of track-vehicle system detection data by using a multi-node collaborative conjugate sparse principal component analysis algorithm; and establishing a wheel / rail forceload identification characteristic data model by taking multi-node characteristic fusion data as a sample data set and using an L1 / 2-SparsePCA-ELM neural network machine learning algorithm. The technical scheme uses the data modeling method to identify the wheel / rail force in real time, then evaluates the safety status of track-vehicle system interaction, and is an important supplement to the existing track quality evaluation standard based on track geometry state detection, which is conducive to the comprehensive analysis of the track-vehicle system status and guidance of high-speed railway track maintenance work.

A principal component analysis algorithm based compression method for data on a road traffic time series. The method comprises: firstly, based on historical data on a road traffic time series and in combination with a principal component analysis method, acquiring a projection matrix of road traffic historical data; next, based on real-time data on the road traffic time series, obtaining a road traffic data matrix; then, based on the projection matrix, acquiring a principal component of the road traffic data matrix, so as to implement compression of the real-time data on the road traffic time series; and finally, based on the principal component and the projection matrix of the road traffic data matrix, obtaining a reconstructed data matrix, so as to implement reconstruction of the real-time data on the road traffic time series. Experimental results show that the method has excellent performance on the aspect of compression of data on the road traffic time series.