1016 results about "Reconstruction error" patented technology

The invention relates to an on-line video abstraction generation method based on depth learning. An original video is subjected to the following operation: 1) cutting the video uniformly into a group of small frame blocks, extracting statistical characteristics of each frame image and forming corresponding vectorization expressions; 2) pre-training video frame multilayer depth network and obtaining the nonlinearity expression of each frame; 3) selecting the front m frame blocks being as an initial concise video, and carrying out reconstruction on the concise video through a group sparse coding algorithm to obtain an initial dictionary and reconstruction coefficients; 4) updating depth network parameters according to the next frame block, carrying out reconstruction and reconstruction error calculation on the frame block, and adding the frame block to the concise video and updating the dictionary if the error is larger than a set threshold; and 5) processing new frame blocks till the end in sequence on line according to the step 4), and the updated concise video being generated video abstraction. With the method, latent high-level semantic information of the video can be excavated deeply, the video abstraction can be generated quickly, time of users is saved, and visual experience is improved.

The invention discloses an abnormity detection method for satellite telemetry data through an adversarial network autoencoder, and the method comprises the steps: breaking the limitation of a traditional empirical model, and employing a pure data driving model; on the basis of a variational autoencoder, introducing a confrontation network idea, using a bidirectional LSTM (Long Short Term Memory) (Long-short term memory network) as a discriminator, and judging whether satellite telemetry data is abnormal or not by using errors of reconstructed data and original data; aiming at the redundancy problem of a satellite sensor, the conventional situation is broken through, and a Markov distance is used for measuring a reconstruction error. In combination with periodicity of satellite orbit operation, a dynamic threshold determination method based on a periodic time window is provided. The method has the advantages that pure data driving is adopted, expert experience is not needed, and the method can be suitable for various occasions; By combining the respective advantages of the variational auto-encoder and the generative adversarial network, the proposed network has the characteristics of high training speed and relatively easy convergence; eliminating redundant data influence between satellite telemetry data by adopting a Mahalanobis distance. According to the periodicity of the satellite, the dynamic threshold method based on the periodic time window is provided, and the misjudgment rate is reduced.

The invention provides a light distribution optimization method for two-dimensional reconstruction of a combustion flow field by combining a tunable semiconductor laser absorption spectrum technology and a computed tomography technology. The method comprises the following steps: a light matrix calculation step, a combustion flow field reconstruction step and a light distribution optimization step and particularly comprises the following steps: step 1, discretizing a region to be tested; pre-setting a light space distribution manner and calculating the length of light penetrating through the discretized region to obtain a light distribution matrix; step 2, selecting a suitable absorption spectrum line and calculating a light projection result of an initial gas distribution model; reconstructing temperatures and concentrations of the tested flow field by using an algebra iterative algorithm and calculating a reconstruction error and a standard deviation; and step 3, wholly searching by using a self-adaption simulated annealing method to obtain an optimal light distribution range and partially searching by using a sequential quadratic programming method to obtain optimal light distribution.

The invention discloses a method for reconstructing partitioned images by compressive sensing on the basis of structural dictionaries, aiming to overcome the defects that the effect of reconstructing the partitioned images by compressive sensing is not ideal in the prior art because the image blocks with different structures can not be expressed most sparsely on the basis of the signal dictionaries. The method comprises the following steps: (1) constructing a training sample library; (2) classifying image blocks; (3) training the structural dictionaries; (4) constructing an observation matrix; (5) observing the image blocks; (6) reconstructing the structural dictionaries; (7) acquiring reconstruction errors; (8) reconstructing images; and (9) outputting the reconstructed images. The images are reconstructed on the basis of the image blocks under all the structural dictionaries by adopting the error weighting and summing method, and the reconstruction quality is improved.

The disclosure relates to a method and a device for single frame super resolution reconstruction based on sparse domain reconstruction, The disclosure mainly solves the technical problem in the prior art that the reconstructed image with high quality cannot be obtained by selecting the appropriate interpolation function according to the prior knowledge of the image. The disclosure adopts the first paradigm of the example mapping learning to train the mapping M of the low resolution feature on the sparse domain Bl to the high resolution feature on the sparse domain Bh and the mapping of the high resolution feature on the sparse domain Bh to the high resolution feature YS, equalizing the mapping error and the reconstruction error to the mapping operator M, the reconstructed high-resolution dictionary Φh and the reconstructed high-resolution sparse coefficient Bh, the better solution to the problem, can be used for graphics processing.

The invention discloses a SAR (Synthetic Aperture Radar) image segmentation technique based on dictionary learning and sparse representation, and mainly solves the problems that the existing feature extraction needs a lot of time and some defects exist in the distance measurement. The method comprises the following steps: (1) inputting an image to be segmented, and determining a segmentation class number k; (2) extracting a p*p window for each pixel point of the image to be segmented so as to obtain a test sample set, and randomly selecting a small amount of samples from the test sample set to obtain a training sample set; (3) extracting wavelet features of the training sample set; (4) dividing the training sample set by using a spectral clustering algorithm; (5) training a dictionary by using a K-SVD (Kernel Singular Value Decomposition) algorithm for each class of training samples; (6) solving sparse representation vectors of the test sample on the dictionary; (7) calculating a reconstructed error function of the test sample; and (8) calculating a test sample label according to the reconstructed error function to obtain the image segmentation result. The invention has the advantages of high segmentation speed and favorable effect; and the technique can be further used for automatic target identification of SAR images.

The invention discloses a semi-supervised intrusion detection method based on a depth generation model. The method comprises the steps of: 1, preprocessing data: converting symbol attributes in a dataset into numerical attributes, and then normalizing all the numerical attributes; 2, converting high-dimensional feature representations of labeled and unlabeled data into low-dimensional representations of a new feature space by using the variational self-encoding technology in the generation model, adding a constraint to low-dimensional feature vectors to obey Gaussian positive distribution soas to obtain a hidden variable z, and training a classifier by using the hidden variable z in combination with a labeled sample; 3, reconstructing labeled sample data: jointly generating a new labeledsample by using the hidden variable z in combination with label class information; 4, reconstructing an unlabeled sample: predicting the probability of each class of an unlabeled sample by using thehidden variable z, and then generating a new unlabeled sample in combination with the hidden variable z; and 5, calculating a reconstruction error of the model with the newly generated labeled and unlabeled samples, and training and optimizing model parameters in combination with a classification error till convergence.

Failed disk reconstruction time in a RAID storage system is decreased when dealing with non-catastrophic disk failures by using conventional parity reconstruction to reconstruct only that part of the disk that actually failed. The non-failed remainder of the failed disk is reconstructed by simply copying the good parts of the failed disk to the reconstructed copy. Since the good parts of the failed disk are simply copied, it is possible to reconstruct a failed disk even in the presence of disk failures in the secondary volumes. The copying and reconstruction starts at the stripe level, but may be carried out at the data block level if a reconstruction error occurs due to secondary media errors.

The invention discloses a data subspace clustering method based on multiple view angles, which comprises the steps of extracting multi-view-angle characteristics in a multi-view-angle database; for the multi-view-angle database, selecting a specific linear reconstruction expression method and determining a regularization constraint method corresponding to the linear reconstruction expression method; determining reconstruction error weight of each view angle characteristic in multi-view-angle characteristics; according to the selected reconstruction expression method and the obtained reconstruction error weights of different view angle characteristics, learning to obtain a linear expression matrix for reconstructing all samples in the multi-view-angle database, wherein the linear expression matrices are used for expressing a relationship among the samples in the database and element values are used for expressing reconstruction coefficients for corresponding samples in the line to reconstruct corresponding samples in the row; correspondingly processing the linear expression matrix to obtain an affinity matrix for measuring the similarity of the samples in the multi-view-angle database; and using a spectral clustering algorithm to partition the affinity matrix to obtain multi-view-angle data subspaces.

A method for training a neuron network using a processor in communication with a memory includes determining features of a signal using the neuron network, determining an uncertainty measure of the features for classifying the signal, reconstructing the signal from the features using a decoder neuron network to produce a reconstructed signal, comparing the reconstructed signal with the signal to produce a reconstruction error, combining the uncertainty measure with the reconstruction error to produce a rank of the signal for a necessity of a manual labeling, labeling the signal according to the rank to produce the labeled signal; and training the neuron network and the decoder neuron network using the labeled signal.

The invention belongs to the technical field of image super-resolution reconstruction, and discloses a multi-scale residual attention network image super-resolution reconstruction method based on attention. The method comprises the steps of selecting a common image data set as a to-be-experimented image set, dividing the to-be-experimented image set into an image training set and an image test set, and performing image preprocessing, designing a multi-scale residual structure unit module, introducing a channel attention mechanism, and building a multi-scale residual attention neural network model based on channel attention, inputting the preprocessed image training set into a multi-scale residual attention neural network model based on channel attention for model training, and inputting the preprocessed image test set into the trained model for testing to obtain a finally reconstructed high-resolution image. According to the method, a basic unit is enabled to focus on extraction of high-frequency information, important feature map information in a channel is better highlighted, important information in an image is better extracted, and reconstruction errors are reduced.

The invention provides an ultrasonic image low-rank analysis based thyroid lesion image identification method. The method comprises three steps of extracting and describing image block-shaped features based on superpixel hierarchy partition: extracting image features in a multi-scale and hierarchical way by taking a superpixel as a unit, removing redundant information of the image by virtue of the superpixel, lowering the complexity of a subsequent image processing task, and giving consideration to acquisition of global and local information; identifying thyroid based on feature space low-rank reconstruction error analysis: according to the low-rank property of image structure information, calculating the similarity between test data and a dictionary in a manner of optimizing a lowest rank, calculating a reconstruction error, and identifying a thyroid region in combination with a graph-cut segmentation algorithm; and detecting a thyroid lesion based on local low-rank decomposition: dividing a data matrix into a matrix with the low-rank property and an error matrix with the sparsity by adopting a low-rank decomposition method, calculating a sparse error, performing significance detection, and determining a lesion region.

The invention provides a video heat prediction method based on deep belief networks. The method comprises following steps: selecting input variables according to video features; normalizing and quantifying influence factors, thus preprocessing training data; determining a single-layer restricted Boltzmann machine reconstitution dimension according to the selected input variable and feature reconstruction errors; forming the deep belief networks by multi-layer stacked restricted Boltzmann machines and a BP (Back Propagation) neural network; adjusting the deep belief networks through a global learning algorithm, thus obtaining an optimum video prediction model; placing to-be-tested video test data in the optimum video prediction model for heat prediction analysis and watching quantity prediction analysis. The invention also provides a video heat prediction system based on the deep belief networks. According to the method and the system of the invention, an online video prediction model based on the deep belief networks is provided; the deep neural network is applied in the online video prediction field; and the prediction accuracy and reliability can be improved.

The invention relates to a non local joint sparse representation based hyperspectral image super-resolution reconstruction method. The method comprises: firstly, performing dictionary training on a low-spatial-resolution hyperspectral image with an online dictionary training method to obtain a corresponding spectral dictionary; secondly, performing joint sparse representation on similar pixel vectors by virtue of a full-color image of a same scene, and reconstructing a high-resolution image; and finally, processing the reconstructed high-resolution image by utilizing an iterative reverse projection technology to obtain a high-resolution hyperspectral image with smaller reconstruction error and higher visual quality. According to the method, the similar pixel vectors are subjected to non local joint sparse representation by utilizing the non local self-similarity property of the image, so that the visual quality of the reconstructed image is improved and structural features of edges, textures and the like of the image are reconstructed more effectively in an empty region while image spectral information is kept complete; and multiple wavebands of the hyperspectral image are subjected to sparse representation and reconstruction at the same time, so that the hyperspectral image with relatively high definition and identification degree can be reconstructed.

The invention discloses a graph embedding low-rank sparse representation recovery sparse representation face recognition method and belongs to the technical field of computer vision and mode recognition. The method comprises the steps that firstly, a graph embedding low-rank sparse representation recovery method is provided, a clean training sample data matrix with the high discrimination power can be recovered from a training sample data matrix, and meanwhile a training sample data error matrix is obtained; secondly, the sparse representation coefficient of face data to be recognized is obtained by using the clean training sample data matrix as a dictionary, using the training sample data error matrix as an error dictionary, and adopting the norm optimization technology; thirdly, according to the sparse representation coefficient of the face data to be recognized, class association reconstruction is carried out on the face data to be recognized; finally, a face image to be recognized is recognized based on the class association reconstruction error of the face data to be recognized. According to the graph embedding low-rank sparse representation recovery sparse representation face recognition method, the problem that the face recognition effect is poor on the conditions that a training sample image and an image to be recognized are polluted by noise or partially blocked can be solved.

The invention discloses a face identification method based on multiscale weber local descriptor and kernel group sparse representation. The face identification method comprises the following steps: firstly normalizing the size of face images and smoothing the images by utilizing a gaussian filter; extracting differential excitation ingredients of the multiscale weber local descriptor of the images and extracting direction information by utilizing an Sobel operator; extracting the multiscale weber local descriptor of the face images according to the multiscale differential excitation and the direction information and mapping the multiscale weber local descriptor to a kernel space by utilizing a histogram intersection kernel; then with a kernel matrix obtained by a training sample as a sparse dictionary, calculating group sparse representation coefficients of a kernel vector obtained by a test sample; and finally reconstructing a multiscale weber local descriptor vector of the test sample according to the group sparse representation coefficients and distinguishing the test sample by utilizing the minimum reconstruction error. According to the face identification method, the multiscale weber local descriptor and the kernel group sparse representation algorithm are fused for face identification, and the identification accuracy rate is greatly improved.

The invention discloses a sparse representation and dictionary learning-based bearing fault classification method. The method is characterized by comprising the following steps of: preprocessing acquired history bearing vibration signals to obtain vibration signals under different working states, respectively constructing corresponding sub-dictionaries and combining the sub-dictionaries into a redundant dictionary; and online acquiring a bearing vibration signal by using a sensor, solving a sparse coefficient of the signal under the redundant dictionary by using a generalized orthogonal matching pursuit algorithm, and classifying the vibration signal through a reconstruction error so as to identify the bearing working state. According to the method, relatively good classification effect can be obtained, the dictionary training process can be accelerated and the ability, of being adapted to target signals, of the dictionary is improved, so that the method is capable of realizing the sparse decomposition of complex vibration signals more efficiently and is used for fault identification.

The invention discloses an image coding method and system based on dictionary learning. The method comprises the steps that image blocks are randomly selected from an image sample to carry out sparse representation based on batch processing, and a dictionary matrix is solved through an alternative optimization method; according to the obtained dictionary matrix, the sparse representation processed in batch is utilized to carry out sparse representation on the image blocks to be coded, and quantification entropy coding is carried out on nonzero coefficients in sparse coefficient vectors; decoding and sparse reconstruction are carried out on the coded image at a decoding end according to the obtained dictionary matrix. The system comprises a dictionary learning module of sparse representation based on batch processing, an image coding module and an image decoding module. According to the method and system, the problem that the training sample in dictionary learning is large in scale is resolved, meanwhile reconstruction errors of the image sample are reduced, and rate-distortion performance of image compression is significantly promoted.

The invention provides a sparse-coding license plate character recognition method based on shape and contour features. The method comprises a sparse dictionary learning process and a dictionary-utilizing character recognition process. The method mainly comprises the following steps: firstly, a training image set is formed by pre-processing standard license plate images; secondly, feature extraction is performed on the training image set, so that a training feature set is formed; thirdly, a sample region feature and a chain code histogram feature of the raining feature set are introduced into an objective function, sparse dictionary learning is performed on license plate character samples off line to obtain dictionaries corresponding to all characters, and a dictionary set is formed by all the dictionaries; fourthly, feature extraction is performed on test sample data; fifthly, test sample features are subjected to sparse representation in each dictionary, and license plate character recognition is performed through reconstruction errors. Since region features and boundary features of character images are considered at the same time, the sparse-coding license plate character recognition method is a fast and robust license plate character recognition method.

The invention relates to a CNN-based HRRP target recognition method for a rejectable radar, which comprises the following steps of: (1) acquiring HRRP time-frequency domain characteristic data of radar high-resolution range profile; (2) selecting training sample set and testing sample set; (3) constructing convolution neural network; (4) setting the adjustable cost function of convolution neural network; (5) training the convolution neural network; (6) obtaining the output results of the convolution neural network; 7) judging whether that reconstruction error is great than the threshold value,if so, rejecting the judgment of the target, otherwise, obtaining the recognition result. A multi-layer convolution neural network is introduced, high-dimensional features are extracted from radar HRRP data in time-frequency domain, which can effectively solve the problem that the accuracy of target recognition is not high due to the limited information of target features in the prior art, and has adjustable rejection ability to targets outside the database, and has better target recognition performance than the general methods.

The invention discloses a method for performing anomaly detection by a weighted convolutional autoencoder-long short-term memory network (WCAE-LSTM network). The method is devoted to perform anomaly detection and positioning by learning a generation model of a mobile pedestrian, thereby guaranteeing the public safety. The invention provides a novel double-channel framework, which learns generationmodes of an original data channel and a corresponding optical flow channel and reconstructs data by utilizing the WCAE-LSTM network, and performs the anomaly detection on the basis of a reconstruction error. In addition, for the problem of complex background, it is proposed that a sparse foreground and a low-rank background are separated by adopting modular robust principal component analysis decomposition; and a weighted Euclidean loss function is designed according to obtained background information, so that background noises are inhibited. The designed WCAE-LSTM network can not only perform the anomaly detection globally but also roughly locate an abnormal region locally; and through the joint consideration of global-local anomaly analysis and optical flow anomaly analysis results, finally robust and accurate detection of abnormal events is realized.

The invention provides a method for outputting images by the utilization of a phase modulator. The method for outputting the images by the utilization of the phase modulator comprises the steps that (1) an initial inputting optical field is determined; (2) Fourier transform is conducted on the current inputting optical field; (3) amplitude of an objective image is used for replacing amplitude of a result of the Fourier transform in the step (2), wherein the objective image is an image which is obtained after an original objective image is softened; (4) Fourier transform is conducted on the result of the step (3); (5) a component of amplitude of the inputting optical field obtained in the step (1) is used for replacing amplitude of the result of the step (4) in a component mode, so that a new inputting optical field which is iterated next time is obtained; (6) the step (2), the step (3), the step (4), the step (5) are repeated till the condition of stopping iteration is met, and therefore a modulated phase needed by the phase modulator for generating the objective image is obtained; (7) the modulated phase obtained in the step 6 is used for outputting a reconstructed image. The method can effectively restrict reconstruction errors and spots, and is suitable for various complex objective images.

The invention discloses a spectrum reconstruction method and system based on a multichannel imaging system, which are used for performing spectrum reconstruction according to the camera response values of a training sample to-be-selected set and a to-be-reconstructed sample set acquired by the multichannel imaging system. The method comprises the following steps: selecting training samples with most similar camera response values from the training sample to-be-selected set for each to-be-reconstructed sample in the to-be-reconstructed sample set respectively to obtain an initially-selected sample set; removing identical samples, and constructing a matrix by the spectral reflectivity information of all samples to perform PCA (Principal Component Analysis) analysis; computing the accumulating contribution rate of a characteristic value; screening the samples to obtain a training sample sub-set, performing spectrum reconstruction according to each training sample sub-set for each to-be-reconstructed sample, and selecting one reconstruction result according to a reconstruction error minimum principle for serving as the final spectral reflectivity value of each to-be-reconstructed sample. By adopting the spectrum reconstruction method and system, the chromaticity accuracy and spectrum accuracy of each reconstructed sample can be remarkably increased, and high-fidelity color reproduction can be performed. Meanwhile, the method and the system are easy and convenient to implement, and are suitable for popularization and application.

The invention relates to a sensing network clustering type space time compression method based on network coding and compression sensing. Targeted at problems of performance defects of reconstruction errors and computing complexities which are not low enough in existing research schemes during exploration of correlation of time and space of sensing data of a wireless sensor network, the invention brings forward a clustering type space time compression method with reference to network coding and a compression sensing theory; time space correlation of sensing data is deeply excavated; through design of appropriate network coding coefficients and observation matrix elements, network coding and the compression sensing theory are fused and unified in a real number domain; data reconstruction is ensured to be feasible and a high success rate is ensured; through construction of sensor node (cluster head node) independent codes and combination with a node combination decoding idea, reconstruction of compressed data of the method is enabled to have lower reconstruction errors. Meanwhile, exploration is carried out on the correlation between the time and the space step by step to guarantee low complexity of the reconstruction process.

The invention discloses a method for tracking an infrared small target based on sparse representation and transfer limited particle filtering. The method includes the following steps: first, introducing a blocking binary indication vector representing a normal target region and a noise background outlier region, and building a sparse representation model of an infrared small target according to atarget dictionary template; then, extracting a high-frequency region of an infrared image through saliency detection as prior information in the state transition process of a particle filter model, and predicting the state of the target and sampling particles under the limit; and finally, building a particle filter observation model based on target sample particle sparse reconstruction error difference, and estimating and tracking the state of the target according to an online template update strategy. The state estimation ability of random particles is enhanced, and the adaptability and tracking accuracy of particle sparse representation to small and weak moving targets are improved.

The invention discloses a fault detection and identification method for civil aircraft system based on LSTM-AE depth learning framework and relates to the technical field of condition monitoring and fault diagnosis of complex civil aircraft system, and can be used to realize the monitoring and identification of flight faults. The invention comprises the following steps: selecting time series dataof multi-state parameters of an aircraft in flight under a certain stable condition, and according to the characteristics of the monitored object, the time series data of state parameters under suitable conditions are selected for the training of the system reconstruction model, then the fault-free state of civil aircraft system is modeled and reconstructed by making full use of the long-time series-dependent memory ability of LSTM model. The fault monitoring and identification are realized by further analyzing the reconstruction error of its state parameters. The invention solves the problemof insufficient fault monitoring means of civil aircraft system, utilizes the advantage of deep learning in big data analysis to mine massive operation and maintenance data of civil aircraft, and provides important support for fault monitoring of civil aircraft system and route fault isolation.

The invention discloses a human face recognition method and device for quickly processing human face shading, and the method comprises the steps: obtaining a test human face sample and a standardized learning dictionary; solving a local restriction code of an unshaded part of the test human face sample through employing a target function with the local restriction, and modeling a shaded region of the test human face sample through a Markov random field till the detection is completed; obtaining the sparse representation of the unshaded part of the test human face sample through the target function with l2-normal constraint; generating a reconfigured human face, corresponding to each class, class by class, and solving a reconfiguration error between the reconfigured human face of each class and the test human face sample; searching a corresponding class with the minimum reconfiguration error, determining the class as the class of the test human face sample, and outputting the class. The method can process and recognize a human face which is shaded or polluted by noise, is higher in recognition speed under the condition of guaranteeing the recognition success rate, achieves the better recognition of the human face, and is more suitable for an actual scene.

The invention discloses an end-to-end network flow reconstruction method based on compression sensing in a dynamic network. In a large-scale IP backbone network, the OD flow is selected by a random walk method; the flow value of partial OD flow acquired by a router subset is described by constructing a sparse flow matrix; a compressing sensing reconstruction module is established by adopting a main information analysis method; a relationship between the OD flow generated by the router sublet and all the end-to-end OD flow in the IP backbone network is descried by using the module, and further all the end-to-end OD flow of the whole IP backbone network is determined. The method can more accurately acquire end-to-end network flow detail characteristic, cannot consume mass hardware resources, can track the dynamic changes of the OD flow in a real time manner, and the reconstruction error is less.

The invention provides an interactive image segmentation method for multiple foreground targets. The method comprises the following steps of: performing linear reconstruction on pixel colors in an image local window, and repeatedly modifying color reconstruction coefficients by using linear projection; repeatedly performing the linear reconstruction on pixel class label vectors in the image localwindow by using the modified color reconstruction coefficients, and estimating to acquire local reconstruction errors; accumulating the local reconstruction errors to acquire a global reconstruction error; building an interactive image segmentation model of the multiple foreground targets; and performing cluster analysis on the same class of pixels which are labeled by a user to acquire a clustering center; acquiring a group of polynomial functions by adopting regression estimation by taking the clustering center as a training sample; mapping the unlabelled pixels by using the polynomial functions to acquire an initial solution; solving the segmentation model; and determining the class attribution of the unlabelled pixels, and outputting a segmentation result. The interactive image segmentation method has a wide application prospect, and the problem that the multiple foreground targets are difficult to segment simultaneously in the prior art is solved.