96 results about "Spectral vector" patented technology

A method of spectral-morphometric analysis of biological samples, the biological samples including substantially constant components and suspected variable components, the method is effected by following the steps of (a) using a spectral data collection device for collecting spectral data of picture elements of the biological samples; (b) defining a spectral vector associated with picture elements representing a constant component of at least one of the biological samples; (c) using the spectral vector for defining a correcting function being selected such that when operated on spectral vectors associated with picture elements representing other constant components, spectral vectors of the other constant components are modified to substantially resemble the spectral vector; (d) operating the correcting function on spectral vectors associated with at least the variable components for obtaining corrected spectral vectors thereof; and (e) classifying the corrected spectral vectors into classification groups.

The present invention relates to a real-time wideband compressor for multi-dimensional data. The compressor comprises a plurality of compression engines for simultaneously compressing a plurality of data subsets of a set of input data vectors and providing compressed data thereof using one of SAMVQ or HSOCVQ data compression. Each compression engine comprises an along spectral vectors codevector trainer as well as an across spectral bands codevector trainer. The compression engines are programmable to perform either along spectral vectors codevector training or across spectral bands codevector training in combination with one of the SAMVQ or HSOCVQ techniques without changing hardware. The compressor further comprises a network switch for partitioning the set of input data vectors into the plurality of data subsets, for providing each of the plurality of data subsets to one of the plurality of compression engines, and for transmitting the compressed data. The real-time wideband compressor is highly advantageous in, for example, space applications by programmable enabling performance of different techniques of codevector training as well as different techniques of VQ. Furthermore, after the compression process is started the compression process is performed autonomously without external communication.

The invention provides a Hyperspectral remote sensing image classification method based on attention mechanism and convolution neural network.featuring that that original hyperspectral remote sensingimage is reduced in dimension by principal component analysis, and the reduce hyperspectral data is sampled into blocks. After that, 3D convolution and pooling operations are carried out to obtain theintermediate feature map. Then, each spectral vector of the intermediate feature is multiplied with the spectral attention module and each spatial feature is multiplied with the spatial attention module to obtain an attention enhancement sample. After that, the convolution operation and attention enhancement operation are performed again. Then the intermediate feature map obtained by 3D convolution operation is inputted into the classifier for classification. The invention has the advantages that the classification cost is reduced, the classification performance is improved, the adaptive feature thinning is realized through the extraction and enhancement of the sample features, and the classification accuracy of the hyperspectral remote sensing image is further improved.

The invention belongs to the field of remote sensing image processing, and particularly relates to a fusion method of multispectral remote sensing images with different spatial resolutions. The method comprises the steps of dividing the multispectral image with high spatial resolution into n types of surface features, calculating the spectral vector mean value of all types of the surface features in the multispectral image with high spatial resolution, then conducting mixed pixel decomposition on the multispectral image with high spatial resolution, showing the multispectral image with low spatial resolution by the percent of all types of surface features in each pixel of the multispectral image with high spatial resolution, solving the spectral vector mean value of all types of the surface features in the multispectral image with low spatial resolution, and finally calculating the pixel value of each pixel of a fused image and generating the fused image. The method can fuse the multispectral image with high spatial resolution and the multispectral image with low spatial resolution, and enable the image fused to be sharper.

The invention discloses an optical device spectrum response measurement method based on double sideband modulation and stimulated Brillouin scattering effect, comprising steps of dividing an optical carrier wave which is outputted by a light source into two paths, performing frequency-beating in a photoelectric detector by a scanning frequency double sideband signal and a carrier having the frequency shifted after the signal passes through the optical device to be detected, obtaining two radio frequency signals which have two different frequencies and carry spectral response information of the optical device to be detected at the scanning frequency double sideband signal frequency position, using a radio frequency amplitude phase extraction module to respectively extract amplitude phase information of two radio frequency signals to obtain an amplitude-frequency response and a phase frequency response of the optical device to be detected at the optical detection signal frequency, changing the wavelength of the optical detection signal and repeating the above process to obtain the spectral vector response information of the optical device to be detected. The invention also discloses an optical device spectrum response measurement device based on the double sideband modulation. Compared with the prior art, the optical device spectrum response measurement method and measurement device greatly improve the measurement range and the measurement efficiency.

The invention discloses a spectral response measurement method of an optical device and belongs to the technical field of optical measurement. According to the method, optical detection signals of the single wavelength are divided into two paths, frequency shifting with the fixed frequency shifting amount is carried out on one path of signals, and the other path of signals pass through the optical device to be detected; then frequency beat is carried out on the two paths of light, and radio-frequency signals carrying the spectral response information of the optical device to be detected at the optical detection signal frequency positions are obtained; a radio-frequency amplitude phase extraction device with the same working efficiency and the frequency shifting amount is used for extracting the amplitude phase information of the radio-frequency signals, and the amplitude frequency response and phase frequency response of the optical device to be detected at the optical detection signal frequency positions are obtained; the wavelength of the optical detection signals is changed, the process is carried out repeatedly, and the spectral vector response information of the optical device to be detected is obtained. The invention further discloses a spectral response measurement system of the optical device. Compared with the prior art, the spectral response measurement method and system of the optical device have the advantages of being capable of achieving the high-precision measurement of the amplitude frequency response and phase frequency response of the optical device and greatly lowering cost at the same time.

The invention discloses an SAM weighted KEST hyperspectral anomaly detection algorithm (SKEST). The method includes the steps: firstly, deducing the SKEST algorithm; and secondly, calculating the SKEST value of each image element in a hyperspectral image by the aid of a double-rectangular window, performing threshold segmentation and detecting abnormal points. In the SKEST algorithm, based on the KEST (kernel Eigen space separation transformation) algorithm, a weight factor is introduced into each sample in a DCOR (difference correlation) matrix of a high-dimensional Eigen space detection point neighborhood by means of SAM (spectral angle mapper) measurement, and the weight factor of each sample depends on an included angle between the spectral vector of the sample and a data center of the detection window. Therefore, abnormal data in the detection window are suppressed, the contribution of main compositional data is highlighted, and the DCOR matrix can more effectively describe target and background data distribution difference. Besides, the SAM is robust to spectral energy, and by the aid of a radial basis function, the SKEST algorithm considers both spectral energy difference and spectral curve shape difference of signals, and accordingly conforms to hyperspectral data characteristics more effectively.

The invention provides a high-spectral ship detection method based on combination of spatial and spectral information, and the method can be used to solve the problems in accuracy and false alarm in ship object detection in a present ship detection method. A monitoring-free spectral subinterval screening method based on inter-class variation is related to, and the monitoring-free spectral waveband screening method is accurate, rapid and highly robust. Relative stability of a water domain is added to threshold segmentation, interference of a non water area of relatively low brightness value in a detection result is weakened greatly, a sea-land segmentation result of high accuracy can be obtained, and the method is simple, rapid and adaptive. Space features of an RX abnormal detection result are enhanced, namely, a 2D intrinsic mode is used to decompose a reconstruction image, and the method is self-adaptive and relies on structure of data itself. According to the method, the spectral feature of a substance is used to analyze main materials of the substance, and all spectrums can be obtained by linear combination of spectral vectors in a spectrum dictionary.

The present invention relates to a method and system for compressing a continuous data flow in real-time based on lossy compression. In real-time data compression, a series of multi-dimensional data subsets acquired in a given period of time are treated as a regional data cube for the purpose of dividing a continuous series of data subsets into a plurality of data cubes. In a first embodiment implementation of parallel processing using a plurality of compression engines is facilitated by separating a data cube into a plurality of clusters comprising similar spectral vectors. By separating the data cube into clusters of similar spectral vectors no artificial spatial boundaries are introduced substantially improving image quality. Furthermore, the spectral vectors within a cluster are more easily compressed due to their similarity. In a second embodiment a predetermined number of 2D focal plane frames in a boundary area of a previous regional data cube close to a current regional data cube are included in a training set used for codevector training for the current region. Therefore, no artificial boundary occurs between the two adjacent regions when codevectors trained in this way are used for codebook generation and encoding of the spectral vectors of the current regional data cube substantially reducing image artifacts between adjacent regions. A remedy for the single bit error problem is provided in a third embodiment. Full redundancy of compressed data for a regional data cube is obtained by combining the previous regional data cube and the current regional data cube for codebook training. In order to obtain redundancy for the index map, the codebook is used to encode the current regional data cube as well as the previous regional data cube producing a baseline index map for the current regional data cube and a redundant index map for the previous regional data cube. Therefore, full redundancy for a regional data cube is provided allowing restoration of a regional data cube if its codebook and/or index map are corrupted or lost due to single bit errors.

The invention discloses a projection structure sparse coding-based hyperspectral image classification method, which mainly solves the problem that in the prior art, neighborhood information of a hyperspectral image can not be effectively used for classification. The method comprises the realizing steps of: (1) reading in hyperspectral image data; (2) determining a training sample set and a test sample set in spectral vectors with labels of the hyperspectral image; (3) solving a projection matrix according to the training sample set; (4) according to the projection matrix of training samples of the hyperspectral image, solving a sparse coefficient of the test sample set; (5) according to the projection matrix and the sparse coefficient of the test sample set, judging a test sample by using an error judgment function, thus obtaining a surface feature category label of the test sample. The method has the advantages of high classification precision, small processing expenditure cost of high dimensional data and usage for surface feature differentiation of the hyperspectral image.

An apparatus for determining a resonant frequency of a wind turbine tower is provided. The apparatus includes a processing unit configured to receive an acceleration measurement value, the acceleration measurement value representative of the acceleration of the wind turbine tower in the direction parallel to a rotor rotational axis of the wind turbine and/or in the direction perpendicular to both the rotor rotational axis and the tower axis of the wind turbine. The apparatus includes a memory configured to store a series of acceleration measurement values, and the processing unit includes a Fourier transform module configured to calculate a spectral vector based on calculating a convolution-based fast Fourier transform of the series of acceleration measurement values, and includes a resonant frequency calculation module configured to calculate the tower resonant frequency based on the calculated spectral vector.

The invention belongs to the technical field of radar, and discloses a jagged Doppler frequency shift selection method for DDMA waveform. The method comprises the following steps: determining the number of discrete points in a non-fuzzy Doppler scope and the Doppler frequency shift resolution; calculating the number of clutter Doppler points and the number of times of search, and constructing a Doppler frequency shift matrix; for each element of the Doppler frequency shift matrix, taking the coordinate thereof as the Doppler frequency shift number of a corresponding transmission channel, constructing a corresponding Doppler spectral vector after determining that the transmission channel is in the non-fuzzy Doppler scope, getting a synthetic Doppler spectral vector, and setting the element as the length of a non-overlapping clutter range in the synthetic Doppler spectral vector; and searching the maximum element in the Doppler frequency shift matrix, and calculating the optimal Doppler frequency shift corresponding to each transmission channel according to the coordinate of the maximum element. Through the method, the requirement on high-pulse repetition frequency of DDMA waveform is reduced, and the blind speed problem of a target during application of DDMA waveform is relieved effectively.

The invention provides a hyperspectral image target detection method based on a variational self-coding network, which mainly solves the technical problem of low detection precision in the prior art,and comprises the following steps of: obtaining a to-be-detected hyperspectral image and a real spectral vector of a to-be-detected target; constructing a variational self-coding network, and trainingthe variational self-coding network; obtaining a feature map of the to-be-detected hyperspectral image; calculating a spectral vector corresponding to the position of the maximum pixel value in eachfeature map in the to-be-detected hyperspectral image; calculating a spectral angle between each spectral vector and a real spectral vector; obtaining a fusion image; obtaining an initial detection image of the to-be-detected hyperspectral image; and obtaining a final detection target of the to-be-detected hyperspectral image. According to the method, the frequency band interference in the hyperspectral image can be reduced, redundant information is reduced, a target and a complex background in the hyperspectral image are better distinguished, the detection precision of a target point is improved, and meanwhile, the complexity of data processing is reduced.

The invention discloses a hyper-spectral image ground object recognition method based on a sparse kernel representation (SKR), which mainly solves the defects that the recognition time is long and the recognition accuracy is not high when the dimensions of a sample are decreased to be low in the existing method. The method comprises the recognition steps of: firstly, using the spectral vectors ofknown tags in a hyper-spectral image as a dictionary for sparse coding, wherein the spectral vectors of the known tags are arranged by classifications and the spectral vector samples of all unknown tags form a test sample set; secondary, using a neighbor method to construct a central sample matrix, respectively mapping test samples and the dictionary to a feature space by constructing a sparse kernel function to obtain the mapped dictionary and the mapped test samples, and conducting line normalization to the mapped dictionary; and finally using the normalized dictionary to conduct sparse coding to the mapped test samples and judging the classifications of the test samples through an error discriminant. The hyper-spectral image ground object recognition method disclosed by the invention has the advantages that the recognition accuracy can be ensured to be high, the ground object recognition of the hyper-spectral image can be completed rapidly at the same time and the subsequent processing of the recognized ground objects is facilitated.

The invention discloses a GPU-based parallel realizing method of an abundance estimation algorithm. The GPU-based parallel realizing method comprises the steps of: 1, storing hyperspectral mixed pixeldata H to a GPU end; 2, acquiring end element vectors m, and allocating the acquired end element vectors m to p parallel threads on a CPU side one by one; 3, acquiring vectors m<~> which correspond to the end element vectors m and are orthogonal in a vector space [m<1>, m<2>, ..., m<i-1>]; 4, acquiring an orthogonal basis corresponding to an end element set M, and calculating a corresponding inner product value; 5, calculating projection of a spectral vector of each pixel in a d<~T> direction; 6, and calculating an abundance result alpha p corresponding to each end element mp, andstoring the abundance result alpha p to the GPU end. According to the GPU-based parallel realizing method, the abundance value is calculated by adopting a GPU parallel processing kernel function, theparallel mechanism runs much faster than on the traditional CPU serial mechanism, and the operating efficiency of real-time abundance estimation is greatly improved.