58results about How to "Keep valid information" patented technology

The invention relates to an intelligent diagnosis system based on metering device failures. The intelligent diagnosis system comprises a data collecting and monitoring module, a meta event information generating module and a failure diagnosing and decision making module, wherein the data collecting and monitoring module is used for collecting and uploading related data according to collection frequency and transmission modes, the meta event information generating module is connected with the data collecting and monitoring module and used for comprehensively analyzing and preprocessing the related data collected through the data collecting and monitoring module, matching the collected related data according to the archival information of a user and generating meta event information through a meta event diagnosing model; the failure diagnosing and decision making module is connected with the meta event information generating module and used for carrying out relevance matching according to the meta event information provided by the meta event information generating module to achieve intelligent diagnosing of the metering device failures. Compared with the prior art, the intelligent diagnosis system has the advantages that information is effectively integrated, redundant information is removed, the diagnosing efficiency is improved, and the diagnosing accuracy is high.

The invention relates to a method for improving differential absorption spectrum on-line monitoring sensitivity, which comprises the following steps of: carrying out characteristic change extraction of the same characteristic on a gas standard absorption section and measured spectroscopic data, and carrying out computation of gas components and concentration on the basis of characteristic change data. The characteristic change of gas absorption spectrums is a point or a range of energy spectrum concentration in a standard absorption section frequency domain graph, and the characteristic change comprises a slowly changing point uL and a fast changing point vH. Band-pass processing with the passbands as uL and vH is carried out on the gas standard absorption section to obtain the characteristic change, and smoothing (low-pass) processing of the fast changing point vH is carried out on the measured spectroscopic data to obtain equivalent emergent light intensity I'(lambda). The invention can effectively eliminate the influences of various noises and interferences on on-line measurement, reserve the part having the greatest contribution to the signal-to-noise ratio and the detection sensitivity in signals, can find an optimal demarcation point of signal processing without using a trial and error method, and finally improve the on-line integrating precision and sensitivity of a differential absorption spectrometric method.

The embodiment of the invention provides a method and device for identifying a mammary gland image, which relates to the technical field of machine learning. The method comprises the following steps of acquiring the mammary gland image; determining a region of interest ROI of a breast lesion in the breast image and a glandular classification of the breast according to the breast image; determiningbreast lesion signs of the ROI according to the ROI; determining the classification of the breast image according to the breast lesion signs of the ROI and the glandular classification of the breast.

The invention relates to an automatic denoising method of electrocardiogram signals, and the method comprises the following steps: a. obtaining electrocardiogram signals from a human body, determiningthe length of network input samples, artificially destroying the sample data, obtaining destroyed electrocardiogram signal samples and constructing a training set and a test set; b. pre-training a denoising auto-encoder using untagged noise-containing electrocardiogram signal samples in the training set to obtain initialized network parameters; c. after the pre-training, tuning the overall network parameters using tagged noise-containing electrocardiogram signal samples in the training set, and making the network parameters meet the requirements of reconstruction error; and d. inputting the test set into a stack network structure of the lossless constraint denoising auto-encoder to obtain clean electrocardiogram signals. Through the treatment by the method in the invention, the clean electrocardiogram signals after denoising do not only effectively filter out noises, but also restore low-frequency characteristic waves of the electrocardiogram signals, so as to retain the effective information of the electrocardiogram signals.

The invention discloses a method for removing shadows and halos under weak light through an indoor polling robot. The method comprises the following steps: carrying out multiscale Retinex strengthening on a picture collected by the robot; carrying out limited times of iteration on pixel points of the strengthened image through mean shift filtration; eliminating the noise points; smoothing the image; removing the low-light shadow areas by utilizing an illumination normalizing method; and carrying out image segmentation through an OTSU threshold segmentation method so as to obtain a binary result. According to the method, the shadows of equipment in weak light environment can be effectively removed, the influences of the halos can be eliminated, and the effective information required by the later-period mode recognition can be kept.

The invention discloses a multivariate time series prediction method and system, a computer product and a storage medium, and the method comprises the steps: respectively extracting spatial and temporal feature vectors of long and short term historical data matrixes through employing two feature extraction codes, inputting a historical time series matrix into a spatial feature extraction encoder, generating a weighted attention spatial feature vector, and carrying out the prediction of the spatial feature vector; inputting the weighted spatial feature vector into a gating cycle unit to generate a spatial-temporal feature vector; inputting spatio-temporal feature vectors extracted from the long-term historical data matrix into an interactive attention module to generate weighted feature vectors; inputting the short-term historical data matrix into an autoregression layer, and generating a linear prediction result of the short-term historical time sequence data; and combining and inputting the weighted feature vector and the coding feature vector into a full connection layer to generate a neural network prediction result, and adding the neural network prediction result and an autoregression layer linear prediction result to obtain a final prediction result. According to the invention, accurate prediction of multivariate time series data is realized.

The invention provides a solar radio interference filtering method, a readable storage medium and electronic equipment. The method comprises the following steps: collecting intensity data in a set time period before and after a solar burst event, and screening out frequency channels where eligible radio signals are located according to radio flux value data of all frequency channels at a certain moment in a non-burst time period; preprocessing the acquired eligible radio station signal data, and establishing a data set; establishing a recurrent neural network, training the recurrent neural network by using the preprocessed radio station signal data, and predicting the signal value of a radio station in a solar burst area by using the trained recurrent neural network; and according to the linear additive property of the signal, subtracting the predicted radio station value from the original radio flux value of the burst area at the corresponding moment to obtain a value of a clean puresolar burst event. Radio station interference signals can be predicted more accurately, and more solar burst information can be reserved. The final solar burst observation effect of the method is obviously superior to that of a common image filtering method.

The invention provides a near infrared spectrum sugar degree detection method and system based on a BP artificial neural network; certain fruits of the same kind are selected to form a sample set, a modeling sample is selected, and the sample set is randomly divided into a correction set and a prediction set; original near infrared spectrums of all the correction sets and the prediction set samples are collected, equal interval division is carried out on the spectrums, and the absorbance of each interval is summed respectively; the content of the sugar degree in the sample is measured by usinga chemical analysis method; a near-infrared spectrum quantitative analysis model is established; a BP neural network is used for constructing a quantitative correction model between the sugar degreecontent of the correction set sample and the near-infrared characteristic spectrum; the near-infrared quantitative analysis model is used for determining the sugar degree content of the prediction setsample, the near-infrared spectrum information data of the prediction set sample is input into the correction model to obtain the sugar degree content of the prediction set sample.

An imaging control method, device, electronic device and computer readable storage medium, wherein the method is applied to an imaging apparatus comprising an array of pixel units composed of a plurality of photosensitive pixel units, each photosensitive pixel unit comprising a short exposure pixel, a medium exposure pixel and a long exposure pixel whose exposure time increases in turn. The methodcomprises determining a brightness level of ambient brightness; determining a target pixel from the short exposure pixel, the medium exposure pixel and the long exposure pixel of each photosensitivepixel unit according to the brightness level of the ambient brightness; controlling target pixels determined in each photosensitive pixel unit to output original pixel information; imaging according to the original pixel information output from each photosensitive pixel unit. The method can retain more effective information in the photographed image, improve the signal-to-noise ratio of the photographed image, and enhance the brightness of the photographed image, thereby improving the imaging effect and the imaging quality, and improving the photographing experience of the user.

The invention provides a medium- and long-term runoff ensemble forecast method and system based on a rainfall forecast. The method comprises the following steps: constructing runoff ensemble forecast data of a target region according to a forecast period, historical runoff data and forecast factors corresponding to the historical runoff data, wherein rainfall factors in the forecast factors are obtained by correcting rainfall forecast data; and inputting the runoff ensemble forecast data into a trained runoff ensemble forecast model to obtain a medium- and long-term runoff ensemble forecast result of the target region, wherein the trained runoff ensemble forecast model is obtained by training a plurality of data-driven models based on sample runoff ensemble forecast data. According to the invention, the medium- and long-term runoff forecast precision and reliability are improved, and the problem that a traditional medium- and long-term runoff forecast method can only generate a deterministic runoff forecast result is solved.

The invention discloses a logistics vehicle low-carbon route planning method based on a heuristic particle swarm algorithm, and the method comprises the following steps: (1), reading problem information, including the position coordinates and demand weight of a customer; (2) initializing algorithm parameters; (3) calculating the fitness of all individuals in the population, and determining an individual extreme value and a global extreme value; (4) performing variation on all individuals by adopting an individual multivariate variation strategy; (5) sequentially crossing the varied individuals with the individual extreme value and the global extreme value to generate new individuals; (6) updating the individual extreme value and the global extreme value; (7) carrying out local search on the individual extremum based on the heuristic information of priority unloading; (8) performing refined search on the global extremum based on the similarity of the population; and (9) judging whether a termination condition is met or not, if yes, terminating iteration, and outputting an individual with the optimal fitness which is the delivery service sequence of the trucks. The method has the advantages of high search speed, high search capability and low carbon emission in route planning.

The invention discloses a ground camera and sky mobile unmanned aerial vehicle linkage analysis method, which comprises the steps of comparing a detected target feature with a target feature issued bya client, and when the similarity is greater than a set threshold, determining that the detected target is a specific target; when the first ground camera locks the specific target, taking other ground cameras in a preset range of the specific target as first candidate cameras, and sending a control instruction to the sky mobile unmanned aerial vehicle; when no specific target exists in the monitoring range of the first ground camera, controlling the first candidate camera and the sky moving unmanned aerial vehicle to track the specific target; when the sky moving unmanned aerial vehicle locks the specific target, taking a ground camera in a preset range of the specific target as a second candidate camera; and when the definition of the specific target in the sky moving unmanned aerial vehicle is smaller than the preset definition, controlling the second candidate camera to track the specific target. Through the method, multi-camera tracking is realized for the target according to a space-time association automatic switching mode.

The invention discloses a human body finger vein identification method, which comprises the following steps that: reducing a target image into a set ratio, carrying out initial clustering on all pixels in the target image according to the gray values of the pixels; deleting a clustering value corresponding to a maximum center value and a minimum center value in the initial clustering, and carryingout secondary clustering on the gray values in residual clustering sets; deleting a category with the maximum center value in the secondary clustering, combining residual categories in each categoryinto a gray set, and updating the gray values in the gray set according to the maximum gray value and the minimum gray value in each category; extracting the pixel values which are greater than zero in characteristic pattern data, and independently forming a spatial point set by aiming at each characteristic pattern data; adopting an ICP algorithm to calculate the total rectification effect valueof the target image and the fingerprint image template; and when the minimum value in the total rectification effect value is smaller than a set threshold value, proving that a collector of a target image and a collector who corresponds to a fingerprint image template which generates a minimum rectification effect value are the same person.

The invention discloses a vehicle logo feature extraction and recognition method based on feature quantization of gradient direction division. The method comprises firstly preprocessing the vehicle logo image captured by the bayonet system, calculating the gradient magnitude and direction of each pixel, and storing the gradient information of all pixels in a corresponding gradient matrix; dividingk gradient directions in advance, counting the gradient magnitudes of all pixels around each pixel corresponding to k gradient directions, and accumulating the gradient magnitudes into k different gradient magnitude matrices; respectively extracting the LTP features of k gradient size matrices, and obtaining the pixel features of the original vehicle logo image by splicing the extracted k LTP features; through K-Means, classifying all the features in the sample to get offline codebook, and then using SVM to classify and recognize the logo image. The method of the invention puts forward a specific recognition scheme for the vehicle mark recognition in the bayonet image, and the recognition result has high accuracy rate, which can meet the requirements of the actual intelligent transportation system.

The invention relates to the field of digital image processing, and discloses a real-time rate control media stream encoding method and a system. The problem of non-synchronization of decoding and playback of video and audio data caused by coding frame skipping can be solved. The method comprises the following steps that: when a media stream is encoded, if the situation that the current encoded frame needs to be subjected to frame skipping is determined, then a particular frame is generated to be used as a coding result of the current coded frame to be inserted into the media stream, wherein all macroblocks in the particular frame are coded in accordance with a skipped macroblock mode.

The embodiment of the invention belongs to the technical field of information, and relates to a cloud native resource dynamic prediction method, which comprises the steps of performing relevancy sorting on acquired to-be-predicted resource data and performance index data based on a Pearson correlation coefficient to obtain a correlation relationship between the to-be-predicted resource data and the performance index data; defining a correlation threshold based on the correlation relationship; taking the performance index data greater than or equal to the relevancy threshold value as performance index time series data; performing transverse data expansion on the performance index time series data to obtain training data and test data; inputting the training data into a constructed time sequence neural network model for training to obtain a trained time sequence neural network prediction model; and inputting the test data into the time sequence neural network prediction model for prediction operation to obtain a resource prediction result. The invention further provides a cloud native resource dynamic prediction device, computer equipment and a storage medium. The prediction complexity can be reduced, and the prediction accuracy can be improved.

The invention discloses an aircraft trajectory prediction method based on a social long and short term memory network. The method comprises the following steps: modeling long-time sequence data of a plurality of aircrafts, capturing sequences and implicit relationships among the sequences, learning interaction behaviors among trajectories, and synchronously obtaining time-space characteristics and operation rules of all trajectories. The real-time sharing of dynamically updated implicit information is realized by establishing a long-short-term memory deep network for each aircraft and establishing pooling layer connection correlative networks between the networks, historical four-dimensional flight paths, speeds and angle information of a plurality of aircrafts in a certain airspace are taken as model input samples, model training is repeatedly performed, and model parameters are continuously updated, so that the prediction effect meets user expectation.

The invention discloses a target monitoring method based on multiple cameras. The target monitoring method comprises the following steps: acquiring target information acquired by each camera; when thetarget enters the multi-view overlapping area, the first camera sending handover prompt information; when the handover prompt information is received, obtaining coordinate information of the target through a homography matrix converted between views; according to the coordinate information of the target, judging whether the target is within the visual field range of a second camera or not; and when the target is in the visual field range of the second camera, the first camera stopping tracking the target, and the second camera tracking the target. Through the method, specific target monitoring based on multiple cameras can be realized, and cross-camera tracking of a specific target is realized.

The invention discloses an image fusion method based on algebraic multigrid and watershed segmentation. The method comprises the following steps that two or a plurality of images of the same scene butdifferent focuses are processed to obtain one synthetical image with rich information; an algebraic multigrid method is used for rebuilding a source image to obtain a rebuilt image; a mean value image is segmented into different regions by a watershed image segmentation method; the mean square error of the source image and the rebuilt image of the region is calculated according to the segmented image region; the clear degree is judged; a region definition decision image is generated; the picture clear and fuzzy boundary is obtained according to the region definition decision image; the imagefusion is performed according to the boundary. Compared with a multiresolution image fusion method, the image fusion method provided by the invention has the advantages that each target region of a fusion image is directly selected from a clear region in the source image; the image definition loss caused by image change is avoided.

The invention discloses an image noise reduction method based on a region growth label. The method comprises the following steps: S1, searching and determining a seed point; S2, carrying out regionalgrowth, marking seed points, carrying out stack-in growth, judging whether a stack top pointer is greater than -1 , if so, entering the next cycle process, carrying out stack-out on the grown seed points, carrying out statistics on the number of the seed points after stack-out, and judging whether eight field points of the seed points meet growth conditions; S3, continuing to traverse the image, judging whether the pixel points which do not grow meet the growth condition, continuously searching the field points meeting the growth condition in the eight field points and combining the field points, and labeling the seed points belonging to different regions until the image traversal is finished; and S4, setting the pixel value of the image region of which the total pixel value is lower thanthe threshold value to be 0 according to the set threshold value requirement on the total pixel value of one region, namely, taking the image region as noise, removing the noise, and setting the pixelvalues of other regions to be 255, namely, an effective region.

The invention discloses an MVB (multifunction vehicle bus) analysis device comprising an interface conversion module, an FPGA module, an STM32 microcontroller chip and a serial port communication module. The STM32 microcontroller chip is used for acquiring and encoding data to be transmitted and transmitting the data to be transmitted, to an MVB through the port conversion module. The FPGA module is further used for acquiring data frames which are transmitted in the MVB, and transforming the format. The STM32 microcontroller chip is further used for decoding the acquired data frames and transmitting the decoded data frames to an upper computer. The MVB analysis device has the advantages that monitoring response data frames of equipment on the MVB is achieved, multiple MVB devices can be automatically monitored in real time, faults of the MVB devices can be found quickly, and the efficiency in monitoring MVB devices can be greatly improved.

The invention discloses an intelligent identification method of an underground working face operation type based on gas concentration parameter. The method comprises the steps of firstly obtaining thegas concentration data of a working face from a mine safety monitoring system, and storing in a gas parameter database, processing the gas concentration data in the database by adaptive noise reduction filter, and then establishing the historical gas concentration data sample set on the basis of the classification of operation types and the extraction of sensitive characteristic parameters, and constructing the adaptive modular rough neural network prediction model, and identifying the corresponding operation types by using the constructed prediction model. The method of the invention has obvious noise reduction effect, can effectively remove noise while retaining the effective information in the gas concentration time series, can realize analysis and identification of the working type ofthe working face, and can better realize the depth mining and the secondary utilization of the gas concentration data.

The invention discloses a fault detection method for a train bearing box. The method comprises the following steps: acquiring a vibration signal and a noise signal of the train bearing box; performing time-frequency analysis on the vibration signal and the noise signal to respectively obtain a frequency spectrum trend term of the vibration signal and a frequency spectrum trend term of the noise signal; performing frequency spectrum division on the vibration signal by using the frequency spectrum trend term of the vibration signal, and performing frequency spectrum division on the noise signal by using the frequency spectrum trend term of the noise signal to respectively obtain a segmented frequency spectrum set of the vibration signal and a segmented frequency spectrum set of the noise signal; extracting an effective frequency band of the vibration signal from the segmented frequency spectrum set of the vibration signal, and extracting an effective frequency band of the noise signal from the segmented frequency spectrum set of the noise signal; and obtaining a fault detection result of the train bearing box according to the effective frequency band of the vibration signal and the effective frequency band of the noise signal. The fault detection of the bearing box can be realized based on the state data of the bearing box, and the safety of train operation is ensured.

The invention discloses an intrusion detection method based on a convolutional neural network and a lightweight gradient elevator, and relates to an intrusion detection technology in the field of network security, and the method comprises the steps: data preprocessing, wherein the data preprocessing process comprises data type conversion, an oversampling technology and image data conversion; feature selection: using a CNN model to select features; and intrusion detection classification: performing classification by using a Light GBM algorithm. According to the method, the difficulty of intrusion detection analysis caused by imbalance, high dimension and nonlinearity of data is overcome, so that the method shows better performance in multi-dimensional evaluation indexes such as accuracy, precision rate and missing report rate.

The invention relates to a method for detecting P and T characteristic waves of an electrocardiosignal. The method comprises the following steps of: a, acquiring the electrocardiosignal of a human body to perform pre-processing; b, carrying out smoothing processing on the signals under the scale of the to-be-detected sections of the P and T characteristic waves through stationary wavelet transform; c, carrying out electrocardio data P and T characteristic wave peak value point detection by utilizing a maximum and minimum value principle; d, correcting the positions of P and T characteristic wave peak points of the electrocardio data according to a time shift correction principle; and e, evaluating a detection result by calculating three indexes of the error rate e<r>, the sensitivity S<e>and the correct prediction degree P of P and T characteristic wave detection. The electrocardiosignal is processed and smoothed by the method in the invention; therefore, the noise is effectively filtered; furthermore, low-frequency characteristic waves of the electrocardiosignal are recovered; and effective information of the electrocardiosignal is reserved.

The invention relates to a license plate positioning method based on edge detection and mathematical morphology, which belongs to the technical field of image processing and comprises the following steps of: S1, graying a vehicle image by adopting a weighted average method; S2, carrying out binarization on the vehicle image; and S3, carrying out edge detection on the binarized vehicle image by adopting a Sobel edge detection operator, wherein the Sobel edge detection operator is used for obtaining a coarsely positioned license plate area, and introducing a local area into the Sobel edge detection operator to carry out mean value operation, and S4, carrying out accurate positioning on the coarsely positioned license plate area by adopting related operation of mathematical morphology; s5, labeling the accurately positioned connected region, and displaying a license plate region. According to the method, the limitation on the size of the license plate and the position of the license platein the image is little, so that the method does not depend on an accurate license plate frame shape, so that higher positioning effectiveness is ensured.

An electroencephalogram (EEG) feature extraction method based on dominant electrode combination and empirical mode decomposition (EMD) includes inputting n-channel EEG signal data, selecting dominant electrodes according to the principle that electrodes with record of EGG signals having classification performance higher than a certain threshold are called the dominant electrodes, otherwise, are called non-dominant electrodes, selecting dominant combinations, respectively extracting feature of EGG data of a training sample set and EGG data of a testing sample set corresponding to each dominant combination by means of the EMD to acquire training feature vector and testing feature vector of each dominant combination, respectively inputting the training feature vector, the testing feature vector, training sample set label, testing sample set label of each dominant combination input a naive Bayesian classifier to classify to obtain classification accuracy of each dominant combination, and finally surmising relation between different stimulations and activated brain regions when executing relevant motor imagery tasks according to the classification accuracy of each dominant combination.

The invention discloses an electrocardiogram signal denoising method based on optimization theory. The electrocardiogram signal denoising method includes the steps that received electrocardiogram signals are preprocessed, and a heartbeat matrix where a heartbeat cycle is stored is acquired; optimization decomposition is conducted on the heartbeat matrix on the basis of the optimization theory, and a denoised electrocardiogram signal matrix is acquired. According to the electrocardiogram signal denoising method based on the optimization theory, by preprocessing the received electrocardiogram signals, the heartbeat matrix where the heartbeat cycle is stored is acquired, and the heartbeat matrix that contains the electrocardiogram signals and electromyogram noise can be acquired; by conducting the optimization decomposition on the heartbeat matrix, the denoised electrocardiogram signal matrix is acquired, the electrocardiogram signals and the electromyogram noise in the heartbeat matrix can be separated, accordingly, the electrocardiogram signals from which the electromyogram noise is removed can be acquired, the electromyogram noise is effectively restrained, and meanwhile the efficient information of the electrocardiogram signals is retained. In addition, the invention further discloses an electrocardiogram signal denoising device based on the optimization theory. The electrocardiogram signal denoising device based on the optimization theory has the same beneficial effects as the electrocardiogram signal denoising method.

The invention discloses an image processing method and device, and belongs to the field of mobile communication. The method comprises the following steps: acquiring an overall zoom coefficient of an original image; determining a target scaling coefficient group enabling a target energy function to be minimum according to the total scaling coefficient, wherein the target energy function is used forrepresenting the difference between the adjusted image and the original image, the target scaling coefficient group comprises a target scaling coefficient of each triangular mesh of the original image, the triangular meshes are meshes obtained by triangulating the original image, the target scaling coefficient is obtained by adjusting the overall scaling coefficient according to the significancevalue of the triangular mesh; and according to the target zoom coefficient group, carrying out zoom processing on pixel points in each triangular mesh to obtain a target image. According to the embodiment of the invention, the problem that image distortion is easily caused in a picture zooming process in the prior art is solved.

The invention discloses a filtering method and system for Beidou high-frequency deformation monitoring data, comprising the steps of: S100 using sample data to initialize the state equation and measurement equation of the amplified Kalman filtering method; S200 using double gross error discriminant statistics to treat filtering The gross error type is judged for the observed value, and the gross error is corrected according to the gross error type; S300 performs noise filtering on the gross error corrected observation value to be filtered to obtain the two-dimensional state vector filter value; S400 calculates the observation value based on the two-dimensional state vector filter value Feedback the filtered value of the observed value to S200 as the filtered value of the observed value of the previous epoch; S500 repeats S200-S400 until the filtered observed values ​​of all epochs are completed. The present invention uses double gross error discriminant statistics to realize effective identification of gross errors, performs personalized correction according to gross error types, and greatly retains effective information in monitoring data, thus ensuring the reliability of filtering results.