29 results about "Data resampling" patented technology

The invention discloses an InSAR (Interferometric Synthetic Aperture Radar) and GNSS (Global Navigation Satellite System) robust adaptive fusion method for high-precision three-dimensional surface deformation monitoring, which belongs to the technical field of space geodetic survey and remote sensing, and comprises the following steps of: constructing a time sequence deformation field based on a small baseline set time sequence InSAR processing technology, and resampling GNSS observation data to a grid consistent with the InSAR by using Kriging spatial interpolation; a Helmert variance component estimation method is adopted to adaptively estimate variance components of InSAR and GNSS data, and reasonable weight fixing of a multi-source observation value is achieved; iterative reweighted least square estimation is introduced, outlier influences in various observation data are dynamically restrained through a Tukey double-weight function, robust optimization of a fusion model is achieved, and therefore a high-precision three-dimensional deformation field is reconstructed. The method comprises the steps of Helmert weight fixing, IRLS robust, variance component updating and a closed-loop feedback mechanism of weight matrix optimization.

The invention provides a frequency-wavenumber domain-based matrix array full-matrix capture three-dimensional ultrasonic imaging method and system, and belongs to the technical field of ultrasonic imaging. Five-dimensional time-space domain radio frequency data are collected through a two-dimensional full-sampling matrix array and are converted into a frequency-wave number domain through five-dimensional fast Fourier transform, angular frequency is mapped into axial wave number through Stolt interpolation, data resampling is achieved, then three-dimensional image frequency spectrum data are synthesized by defining spatial wave number coordinates of an image and conducting coherence stacking, and the image frequency spectrum data are obtained. And finally, reconstructing a high-quality three-dimensional ultrasonic image through three-dimensional inverse fast Fourier transform. Compared with a traditional delay summation beam forming method, the method has the advantages that the calculation efficiency is remarkably improved, the speed is increased by several times to dozens of times, the spatial resolution, the contrast noise ratio and the spot signal-to-noise ratio are remarkably improved, and the method is suitable for high-resolution real-time three-dimensional and four-dimensional ultrasonic imaging application.

The application relates to the field of meteorological data processing and satellite navigation application, and provides a multi-source meteorological data fusion method and system based on a Beidou space-time reference. The method comprises the following steps: acquiring Beidou data and multi-source meteorological data to be fused, taking the coordinated universal time timestamp output by a Beidou ground-based enhancement station as a global time reference, and setting a sliding window; taking the world geodetic coordinate system corresponding to the Beidou positioning data as a space reference, resampling the multi-source meteorological data to a preset standard grid, and combining terrain elevation data to construct a unified space coordinate system; calculating the confidence degrees of fused static terrain features and dynamic change features based on Beidou satellite orbit residual and clock difference parameters, adjusting the feature weights of the fused static terrain features and dynamic change features, and generating an enhanced fusion feature matrix.

This invention discloses a method and system for resampling time-domain data of electromechanical-electromagnetic simulations while preserving time-frequency domain features. It includes: constructing electromechanical and electromagnetic transient time-series trajectory samples of a new energy power grid system; processing the trajectory samples separately using an improved Proni method to obtain frequency domain response features; reconstructing the electromechanical and electromagnetic transient time-series trajectories separately using linear interpolation and adaptive resampling techniques based on the frequency domain response features; and fusing the time-domain and frequency-domain features in the reconstructed time-series trajectory data to form a comprehensive feature vector, thereby obtaining electromechanical-electromagnetic simulation time-domain data that preserves time-frequency domain features. This invention solves the asynchronous sampling problem of electromechanical-electromagnetic time-domain simulation data and can be used to support the feature analysis of inconsistent data at different time scales in electromechanical and electromagnetic systems. It has broad application prospects and significant practical implications in fields such as scheduling and control of new energy power grids.

A kind of coal mine personnel action recognition method applied in low-illumination environment belongs to video recognition field.The video data of personnel operation in low-illumination environment of coal mine is resampled, and then converted into video data under normal light by style conversion module.After global feature extraction module extracts global space-time features of video frame sequence, target detection module detects personnel in video, and extracts feature representation of personnel action pipeline from video frame sequence.Finally, pipeline features are respectively input into regression module and action classification module to calculate the position of each pipeline and the probability of belonging to each action.The advantages of the method are as follows: style conversion module does not need paired data for training, which reduces the difficulty of data acquisition;in target detection module, video target detection is converted into a set of set predictions, which does not need any prior knowledge and subsequent processing, and realizes end-to-end detection of personnel in video.

A device includes a receiver to receive a packet over a channel at a first frequency and generate a sampled stream of data at a first sample rate corresponding to the first frequency. A data resampler circuit includes a re-timer engine to determine, using a ratio between the first sample rate and a crystal oscillator (XO)-divided sample rate, re-timer values. The data resampler circuit includes a time shifting circuit to re-sample data values of the sampled stream of data associated with locations of the plurality of re-timer values. A correlation circuit uses the re-sampled data values, and the re-timer values to match an expected data pattern to a corresponding data pattern detected in a frame of the packet.

The invention provides a domain generalization laser radar point cloud semantic segmentation method and system, and belongs to the field of laser radar semantic segmentation, and the method comprises the steps: obtaining point cloud data of a source domain, and carrying out the preprocessing; resampling the preprocessed source domain point cloud data to generate a resampling point cloud; generating a mixed point cloud through a cross-scene mixed enhancement strategy; inputting the initial point cloud data, the re-sampling point cloud data and the mixed point cloud data into a semantic segmentation network to extract point cloud features; unified alignment of point cloud features in a semantic space is realized by using a text encoder of a contrast language-image pre-training model; and carrying out model training by adopting a contrast loss function, selecting a target domain point cloud, inputting the target domain point cloud into the trained model, and outputting a point cloud semantic segmentation result. According to the method, through the synergistic effect of multi-density resampling, cross-scene mixing and unified semantic space alignment, the domain offset problem is effectively relieved, and the semantic segmentation precision and generalization ability of the model on an unknown target domain are remarkably improved.

The application discloses an exoplanet light variation signal classification method based on data enhancement and integrated learning, which comprises the following steps: splicing light variation signal data of different observation seasons of the same target, and pre-processing the data; resampling the data to a fixed length; dividing the resampled data set into a training set, a test set and a validation set; performing data enhancement on the training set by over-sampling and under-sampling respectively to obtain sub-training set 1 and sub-training set 2; building a one-dimensional CNN network model for the two training sets respectively, and optimizing the network by adjusting the number of hidden layers, the number of neurons, the activation function and the model optimizer; obtaining the best classification model as base classifier 1 and base classifier 2 by adopting N-fold cross-validation respectively; combining the base classifiers by adopting a weighted voting method to obtain a light variation signal classifier, and obtaining a classification result according to the signal classifier. The application solves the problem of unbalanced exoplanet light variation signal categories, and enables the model to obtain more features to improve the generalization ability.

This invention relates to a small target detection method based on multi-scale information fusion, comprising the following steps: acquiring an existing dataset and preprocessing the dataset, with all preprocessed images constituting a training set; constructing a small target detection model, which uses the YOLOv5 backbone feature extraction network as the base model, replacing several modules within it; introducing data resampling operations on the input dataset and training it, saving the model parameters after optimal convergence to obtain the optimal small target detection model; calling the optimal small target detection model, inputting the test image for detection, and outputting the recognition result. Experimental results demonstrate that this method improves the ability to extract multi-scale features and effectively improves the detection performance of small-scale targets, with test accuracy improved compared to existing methods.

The invention discloses an out-of-distribution data learning method and system for agent navigation. The method aims at improving the capability of utilizing out-of-distribution data in a navigation task of a machine learning model. By optimizing the navigation strategy, a large amount of distributed external navigation data which is easy to obtain but does not reach the target position can be effectively utilized. According to the main technical scheme, the method comprises the steps that a navigation data set containing expert demonstration and out-of-distribution demonstration is collected; training a reverse environment model and a reverse navigation strategy; reverse data enhancement and data resampling are executed through self-paced learning; and finally, combining off-line behavior cloning and reinforcement learning, adopting a behavior cloning method for marked data containing expert behaviors, implementing approximate dynamic planning on unmarked data, and establishing an optimization target and an optimization navigation strategy in an out-of-distribution state. The method has the characteristics of easiness in data acquisition, easiness in algorithm implementation and the like, and based on few expert demonstration, the decision-making ability of the intelligent agent in a complex navigation environment is steadily improved.

The invention relates to the field of artificial intelligence and point cloud segmentation, and provides a segmentation method and system for disaster scene point cloud data, and the method can effectively promote the network to improve the discrimination capability of new objects through carrying out the data resampling of new objects in a secondary subset in the network continuous learning of students. The teacher network prediction result is combined with the new-class original real label to generate a pseudo label, and training loss is obtained based on the difference between the generated pseudo label and the student network prediction result; point cloud features of a teacher network and a student network are subjected to synchronous layered pooling, and distillation loss is obtained based on the difference between the pooling features of the teacher network and the student network, so that the network does not need to construct total loss based on an old class, and the problem that a computer memory is greatly increased due to old class continuous storage is solved; and meanwhile, the trained teacher network on the main subset is utilized to guide the students to learn on the network, so that the post-disaster objects in the new class can be quickly identified.

The application relates to a touch screen resampling method and device, electronic equipment and a storage medium. The method comprises the following steps: performing scene recognition on the electronic equipment to obtain a scene recognition result; determining a corresponding target resampling time difference according to the scene recognition result; the target resampling time difference is a time difference between a sampling time point and a time point of a screen refresh event; after listening to the screen refresh event each time, determining the sampling time point according to the target resampling time difference and the time point of the screen refresh event; resampling target touch screen data reported before the screen refresh event according to the sampling time point to obtain supplementary touch screen data. The method can improve the response effect of touch.

The invention discloses a method for reducing resource overhead in a data resampling process. The method comprises the following steps: calculating a ratio of an input sampling rate to an output sampling rate of a rate matching module and an absolute difference value between 1 and the ratio; when the input sampling rate is smaller than the output sampling rate, data are read from the FIFO buffer according to a read enable signal and a read pointer, and after each time of reading, the read pointer is moved forwards by one bit, and absolute difference values are accumulated; when the accumulation result of the absolute difference is greater than or equal to 1, keeping the read pointer unchanged and reading again for interpolation; when the input sampling rate is greater than or equal to the output sampling rate, writing data into the FIFO buffer according to the write enable signal and the write pointer, and after each time of writing, forwards moving one bit of the write pointer and accumulating absolute difference values; and when the accumulation result of the absolute difference is greater than or equal to 1, pulling down the write enable signal and discarding the current write-in data to perform value extraction. According to the invention, under the condition of not depending on a high-frequency clock and a large-depth FIFO, the resource overhead in the data resampling process is reduced.

The invention provides a suspension road spectrum reduction method for accelerating a rack endurance test, and belongs to the field of rack endurance tests. The method comprises the following steps: step 1, recognizing a suspended road spectrum characteristic pavement and splitting load data; step 2, deleting burrs of the suspension time domain load data; 3, determining the resampling frequency of the suspension time domain load data; step 4, executing the resampling of the suspension time domain load data, and correcting the resampled data; step 5, deleting the data of the time period less than the threshold load to obtain reduced time domain load data of each characteristic pavement suspension; step 6, obtaining complete reduced suspension time domain load data by splicing the suspension time domain load data of each characteristic pavement; step 7, calculating a false damage amount after the suspension time domain load data is reduced, and judging whether the false damage amount is greater than or equal to a threshold value or not: if so, turning to step 8; if not, turning to the step 3; and step 8, carrying out a rack endurance test. According to the method, the bench test period is shortened, and the accuracy is improved.

The invention discloses an automobile multi-source data time alignment and fusion method and system. The method comprises the steps of obtaining first data and second data with non-uniform time references, and performing standardization processing on time; in the second data, window data are extracted from each candidate position through sliding window search, time coverage is judged, the window data are resampled to a first data time point, and a similarity distance is calculated by adopting a dynamic time warping algorithm; selecting the position with the minimum distance to establish a time mapping relation; and performing data fusion to generate fused data covering the complete or partial time range of the second data. The system comprises a data acquisition and preprocessing module, a sliding window matching module based on dynamic time warping, a time alignment relationship establishment module, a data fusion module and a data output module. According to the method, the relation between the time type and the sampling frequency of the data is not limited, universality is achieved through self-adaptive resampling, and the problems that in the prior art, the requirement for data characteristics is harsh, and the application range is narrow are solved.

The application discloses a waveform data resampling method, device, functional module and system. The waveform data resampling method comprises the following steps: using target waveform sampling data to initialize a jump point detection data, and identifying a jump sampling point in the jump point detection data; using a current interpolation algorithm to perform interpolation processing between the jump sampling point and a previous adjacent sampling point of the jump sampling point; after updating the jump point detection data by using the interpolation data, returning to perform the operation of identifying the jump sampling point until the use of the interpolation algorithm is completed for a preset number of times; performing resampling processing according to interpolation positions recorded by each interpolation algorithm in the interpolation algorithm set before interpolation processing to obtain resampling data of a to-be-processed channel; performing data waveform recovery according to the resampling data to obtain to-be-stored data, and storing the to-be-stored data. The technical scheme of the embodiment of the application can greatly improve the sampling precision of the waveform data and meet the occasions with high data precision requirements.

The invention relates to a head and neck tumor multi-modal image database construction method and a computer storage medium, and relates to the technical field of multi-modal image database construction.The method comprises the steps that S1, CT and MRI data of head and neck tumors and related clinical, pathological and gene cross-scale multi-modal data are collected through an image archiving and communication system (PACS) and a hospital information system (HIS); data quality control is carried out; s2, performing data anonymization and data desensitization processing on the acquired CT and MRI images; s3, carrying out standardization processing on all CT and MRI images, wherein the standardization processing comprises data resampling and data annotation; s4, multi-modal CT and MRI images and corresponding marks are classified, arranged and updated in real time; and S5, carrying out cleaning, conversion and quality control on the unstructured data. The system has the advantages that the blank of head and neck tumor image artificial intelligence databases at home and abroad is filled; a solid foundation is laid for research, development and training of an artificial intelligence model, and the method has great value for promoting precise diagnosis and treatment, mechanism research and clinical transformation of head and neck tumors.

The invention discloses a multi-modal data prediction method and system based on a Markov model, and relates to the technical field of data processing, and the method comprises the steps: collecting a multi-modal data set, carrying out the normalization processing, carrying out the data resampling through employing an interpolation method, carrying out the sequence alignment of asynchronous multi-modal data through employing a dynamic time warping (DTW) algorithm, and carrying out the prediction of the asynchronous multi-modal data. The method comprises the steps of calculating the Euclidean distance between every two modal points, constructing a dynamic programming matrix, backtracking to generate an optimal path as correction of a time period between multi-modal data, constructing a Bayesian Gaussian process probability model, calculating conditional probability distribution, determining a missing point prediction value, and generating a feature embedding vector of same-modal data through principal component analysis. According to the method provided by the invention, the coupling association of the features and the states is further strengthened through modeling of the conditional probability, and the limitation that a traditional Markov chain cannot process complex context dependence in a multi-mode system is made up.

The application provides a data synchronization method and device for multi-device cooperative collection, and belongs to the technical field of automobile testing. The synchronization method comprises the following steps: a step voltage signal is sent to all data collection devices participating in collection through a general hardware trigger module; each data collection device receives and records the step voltage signal through an analog voltage channel; the original data collected by each data collection device is subjected to time axis resetting with the moment when the step voltage signal reaches a preset voltage threshold as a time zero point; the data obtained after resetting is resampled at the same sampling frequency, and the data length is aligned, so that the multi-device data is synchronized and aligned. The step voltage signal is sent through the general hardware trigger module, and the starting moment of collection of each device is unified; meanwhile, the time deviation of data collection is solved by data resampling of software, the data collected by collection devices of different brands and models is synchronized and aligned, and high-precision and high-reliability cooperative collection is realized.

The application discloses a depth domain seismic wavelet extraction method based on a generalized seismic wavelet model. The method is characterized in that, the depth domain logging data information can be fully utilized, the depth domain seismic record is directly synthesized in the depth domain, the depth domain seismic wavelet can be efficiently and accurately extracted from the depth domain generalized seismic wavelet model only through 3 to 5 iterations, the method has high stability, and the effective information loss problem caused by domain conversion or data resampling is avoided.

This application discloses a method for reducing resource overhead during data resampling, comprising: calculating the ratio of the input and output sampling rates of the rate matching module and the absolute difference between 1 and the ratio; when the input sampling rate is less than the output sampling rate, reading data from a FIFO buffer according to a read enable signal and a read pointer, moving the read pointer forward by one position and accumulating the absolute difference after each read; when the accumulated absolute difference is greater than or equal to 1, keeping the read pointer unchanged and reading again for interpolation; when the input sampling rate is greater than or equal to the output sampling rate, writing data to a FIFO buffer according to a write enable signal and a write pointer, moving the write pointer forward by one position and accumulating the absolute difference after each write; when the accumulated absolute difference is greater than or equal to 1, pulling the write enable signal low and discarding the currently written data for sampling. This invention reduces resource overhead during data resampling without relying on a high-frequency clock and a large-depth FIFO.

The present application relates to the field of grid data application, and particularly to a grid data resampling method and system in a high concurrency scenario, the system comprising: an extensible sampling module, an extensible cache computing module and a storage area, the storage module storing grid data, pyramid data, coordinate system data and index table data; the extensible cache computing module performs pyramid processing on newly uploaded grid data to obtain pyramid data; and a plurality of extensible cache computing modules work in parallel; the extensible sampling module comprises a plurality of independently running modules, which are used to respond to the demand for resampling, resample the pyramid data and feed back to the user. In the high concurrency scenario, distributed cache storage technology and distributed resampling technology are used to reduce the complexity of system design and optimize the efficiency of client data request.

The invention discloses a multi-dimensional convolutional network model and a pedestrian trajectory tracking method based on the same. Accelerometer, gyroscope and attitude data of an IMU sensor of a mobile phone are collected at a preset frequency through a dynamic sliding window, a cubic spline interpolation model is constructed for data resampling, and the problem of equipment sampling frequency fluctuation is solved. Sensor data is converted from a carrier coordinate system to a global coordinate system by adopting quaternion coordinate conversion, and is reconstructed into three-dimensional tensor input through a differentiation rule. According to the collaborative architecture, time sequence features are extracted through the one-dimensional residual network, spatio-temporal context modeling is carried out through the two-dimensional residual network, receptive fields are enhanced and features are compensated through expansion convolution, and cross-modal joint representation is constructed in combination with physical parameters. A double-branch regression layer is designed, a main branch predicts a displacement vector, an auxiliary branch outputs uncertainty, and stable training is achieved through an improved negative log-likelihood loss function. According to the method, the trajectory tracking precision is effectively and remarkably improved, and quantitative uncertainty evaluation indexes are provided at the same time.

The embodiment of the present application relates to a kind of time series anomaly detection method and device, the method comprises: obtaining the time series data set to be processed, each time series data in the time series data set at least includes time point and corresponding numerical value;The data in the time series data set is preprocessed, and the preprocessing includes data resampling, missing value processing, data labeling, data formatting and anomaly injection;Based on the time series data set after preprocessing, an anomaly detection model is constructed;Anomaly detection is carried out using the anomaly detection model.The technical scheme provided in the embodiment of the present application, by the reasonable data preprocessing link to time series data set, improve the data quality;By providing different sequence anomaly detection model, and more flexibly link it when monitoring in real time on line, while ensuring accuracy also makes its execution efficiency improves.

The invention discloses a machine learning liver cancer microvessel invasion probability prediction system based on image omics and clinical features. The system comprises a data collection module which collects clinical CT data of a patient and a marked focus area; the data integrity checking module is used for extracting DICOM header file information and carrying out data integrity checking; the image format conversion module is used for converting the CT data into an NRRD format, reading a DICOM (Digital Imaging and Communications in Medicine) sequence, keeping space information and intensity information of an original image and generating a standardized medical image format file; the image preprocessing module is used for carrying out data resampling and intensity standardization; the radiomics feature extraction module is used for extracting quantitative features from the image and the segmentation mask; the data set division module is used for dividing into a training set and a test set; the feature selection module is used for performing feature selection on the training set; the model building and training module is used for carrying out model building and training; the model performance evaluation module is used for evaluating the model; and the probability prediction module is used for inputting the test set into the evaluated model and outputting a prediction probability.

The application discloses a method and device for reducing image noise, a computer readable storage medium and a terminal, and relates to the technical field of image processing.The method comprises the following steps: performing first data processing and second data processing on a to-be-processed image respectively to obtain a first image and a second image, wherein the first data processing is increased by at least one round of data resampling processing compared with the second data processing; determining a fusion weight value of each pixel according to the first image and the second image; and performing pixel-level fusion processing on the first image and the second image by using the fusion weight value to obtain a fused image.The application can remove chroma noise while retaining the color of a real object in an image and reducing color loss.

This invention discloses a method, system, equipment, and storage medium for constructing a distribution network operation architecture, relating to the field of distribution network operation architecture. The method includes: extracting monitoring data from different equipment operations to establish a distribution network dataset with the same time-stamp; resampling all equipment data to the same timestamp sequence based on the kernel density formula; determining and performing alignment operations between the current equipment operation monitoring data and the same timestamp sequence based on the equipment's sampling frequency and timestamp; obtaining the quality assessment score and weight coefficient of the equipment operation monitoring data, and updating and optimizing it; constructing a distribution network operation architecture evaluation model and outputting the current operation status labels of the distribution network equipment; and dynamically adjusting the equipment operation parameters in the distribution network operation architecture based on fault tree analysis and expert experience rules. This achieves dynamic sampling of equipment operation data and applies the distribution network operation architecture evaluation model to solve the problem of synergistic operation of distribution network economy and reliability under multiple constraints.

Various systems and methods are provided for processing computed tomography (CT) data in a CT imaging system. The CT imaging system comprising an X-ray source configured to emit X-rays, an X-ray detector configured to generate sampled digital detector data and a digital processor configured to process the sampled digital detector data. The method comprises filtering, in the digital processor, the sampled digital detector data to generate filtered detector data and resampling, in the digital processor, the filtered detector data to generate resampled detector data. The resampling comprises a reduction in data size of the filtered detector data. The filtering is performed on at least part of the sampled digital detector data according to a filtering setting and the resampling is performed on at least part of the filtered digital detector data according to a resampling setting, wherein the filtering setting and resampling setting are decoupled.