169 results about "Separation algorithm" patented technology

Embodiment provide a surveillance camera adapted to be connected to an internet protocol network, the surveillance camera including at least one facial processor, at least one facial recognition algorithm embodied in suitable media, at least one facial recognition algorithm executable with digital format image data by at least one facial processor detecting faces, execution of at least one facial recognition algorithm producing unique facial image data, execution of at least one facial separation algorithm producing facial separation data, at least one facial processor in communication with at least one facial signature database to obtain reference data, execution of at least one facial signature algorithm comparing facial separation data and reference data to identify correlations, at least one compression algorithm producing compressed image data, and a network stack configured to transmit to the network unique facial image data for each detected face and compressed image data.

The invention discloses a high-speed railway user separation method based on signaling data of an A interface and an Abis interface. The method comprises the following steps of: acquiring all signaling data of the A interface and the Abis interface of a high-speed railway cell; establishing a high-speed railway resident user database, and determining high-speed railway resident users in the high-speed railway cell; acquiring call behavior characteristics of high-speed railway users which are not the resident users according to the acquired signaling data; and separating the high-speed railway users from high-speed expressway users and normal users according to the call behavior characteristics. By the method, user types such as a high-speed railway user, a resident user along a railway and a low-speed user along the railway can be effectively separated by establishing a high-speed railway user separation algorithm model, relevant functions of a high-speed railway road test simulation system can be realized according to the separated high-speed railway users, and uncertain problems of a small quantity of testing samples, high contingency and single testing terminal during the conventional high-speed railway simulation test can be effectively solved.

The invention provides a method for sound field separation by double plane vibration speed measurement and two dimensional space Fourier transform, which is characterized in that a measurement plane S1 and an auxiliary measurement plane S2 parallel therewith and Delta h away therefrom are disposed in the measurement sound field; normal direction particle vibration speed is measured on two planes; wave number domain transfer relationship between normal direction particle vibration speed on two measurement planes is established by the two dimensional space Fourier transform; sound pressure, normal direction particle vibration speed radiated by sound sources on both sides of the two measurement planes is separated according to the transferring relation. The method adopts two dimensional space Fourier transform as the sound field separation algorithm, making computation speed fast, and implementation simple; normal direction particle vibration speed on two measurement plane is used as the input for separation, so that the separated normal direction particle vibration speed has higher accuracy comparing with using sound pressure on two measurement planes as the input. The method can be widely applied for near-field acoustic holographic measurement under internal sound field or noise environment, material reflection coefficient measurement, scattering sound field separation.

The invention discloses an algorithm for separating vibration signal blind sources under a strong noise environment. The method comprises the following steps: 1, de-noising a group of given mixed signals containing noise through a time delay autocorrelation method to acquire de-noised mixed signals; 2, performing mean value removal and steady whitening pretreatment on the mixed signals acquired in the first step to further reduce the influence of the noise signals on the separation result; and 3, calculating second-order and fourth-order cumulated amount of the initial separate signals, using the sum of diagonal elements of second-order and fourth-order cumulated amount matrixes as a cost function, and maximizing the cost function to similarly diagonalize the joints of the cumulated amount matrixes so as to realize the separation of the signals of the independent sources and acquire orthogonal separation matrixes. The method combines the conventional de-noising method and the blind separation algorithm to realize the separation of the mixed signals under the strong noise environment, and has the advantages of good separation effect, high convergence rate and de-noising effect free from the limitation of the set threshold value compared with the conventional algorithm.

The invention provides a method for single sweep quantitative magnetic resonance T2 imaging based on overlapping echoes and relates to a magnetic resonance imaging method. Echo signals of two different evolution time are generated by adding two excitation pulses with the same deflection angle in single sweep. Though the evolution time of the two echo signals is different, the T2 weighing of the two echo signals is caused to be different, the two echo signals come from the same imaging slice, the two echo signals can be separated through prior knowledge between the two echo signals, namely the structural similarity in combination with edge sparsity, and the two echo signals are separated by utilizing a separation algorithm corresponding to sparse conversion matching; and finally the two signals obtained through separation undergo T2 calcualtion to obtain a quantitative T2 image. The quantitative T2 imaging in single sweep is acquired, the time of quantitative T2 imaging is reduced from the second grade and even minute grade to the ms grade, and the quality of the acquired T2 image can be equivalent to the quality of an image obtained in an EPI sequence in conventional single sweep.

The Target Separation Algorithms (TSAs) are used to improve the results of Automated Target Recognition (ATR). The task of the TSAs is to separate two or more closely spaced targets in Regions of Interest (ROIs), to separate targets from objects like trees, buildings, etc., in a ROI, or to separate targets from clutter and shadows. The outputs of the TSA separations are inputs to ATR, which identify the type of target based on a template database. TSA may include eight algorithms. These algorithms may use average signal magnitude, support vector machines, rotating lines, and topological grids for target separation in ROI. TSA algorithms can be applied together or separately in different combinations depending on case complexity, required accuracy, and time of computation.

The invention discloses a water and fat separation imaging method in magnetic resonance imaging. A fast spin echo sequence based on a two-point Dixon method is adopted, wherein each refocusing radio-frequency pulse corresponds to two read-out gradients having the same polarity and a rephasing gradient having the opposite polarity, the two read-out gradients corresponding to each refocusing radio-frequency pulse are respectively divided into a front portion and a rear portion through the echo center of each of the two read gradients, the area of the rear portion of the front read gradient is less than that of the front portion of the front read gradient, and the area of the front portion of the rear read gradient is less than that of the rear portion of the rear read gradient. The method comprises operating the fast spin echo sequence; conducting fast Fourier transform (FFT) on each collected echo signal, rebuilding and obtaining water and fat in-phase images and water and fat antiphase images; conducting Fourier transform on the data executing portion of each echo signal; carrying out the water and fat separation algorithm on the water and fat in-phase images and the water and fat antiphase images to obtain pure water images and pure fat images. The invention further provides a corresponding device which can resolve the problem of image artifacts.

The invention provides a single-scanning quantitative magnetic resonance diffusion imaging method based on dual echo, and relates to a magnetic resonance imaging method. According to the method, two echoes with the same evolution time are generated through two small-angle excitation pulses with the same turning angle, so that the same transverse relaxation time is achieved; a displacement gradient is added after each excitation pulse to achieve central displacement of the two echo signals in a signal space, and a diffusion gradient is added after the first excitation pulse, so that diffusion reduction only exists in the first echo signal; accordingly, signals under different diffusion factors are obtained. The two echo signals are from one imaging slice, so that the two echo signals can be separated through priori knowledge of the two echo signals by matching sparse conversion with a corresponding separation algorithm. Finally, a quantitative ADC image is obtained by performing apparent diffusion coefficient calculation on two signals obtained through separation. Single-scanning quantitative ADC imaging is obtained through the method, and the quality of the obtained ADC image is good.

The invention relates to a water-fat separation magnetic resonance imaging method. The method is particularly applicable to a high field and super high field magnetic resonance system. The method comprises the following steps of 1, setting an integrated stable-state free precession imaging sequence capable of adjusting multiple echo times; 2, imaging the same imaging area and acquiring a magneticresonance signal corresponding to each echo time; 3, conducting Fourier image rebuilding on the acquired magnetic resonance signal to obtain a magnetic resonance complex number image corresponding toeach echo time; 4, integrating magnetic resonance complex number image information corresponding to the multiple echo times, wherein the magnetic resonance plural number image information comprises anamplitude and a phase; adopting a water-fat separation algorithm to solve images of separated water and fat. The water-fat separation magnetic resonance imaging method can be widely applied to various magnetic resonance high field imaging systems, particularly a super high field imaging system.

The invention relates to a multi-source distribution type remote sensing distinguishing method and system for geothermal resources. The multi-source distribution type remote sensing distinguishing method for the geothermal resources comprises a data acquisition step which is used for obtaining multi-source remote sensing images; a data processing step which comprises performing color coding and mapping processing on multispectral remote sensing image data through a temperature and emissivity separation algorithm to obtain temperature color coding image information; establishing a geologic structure interpretation sign of the geothermal resources for radar remote sensing image data and performing geologic structure interpretation on the radar remote sensing image data; a data integration step, namely performing superposition, projection transformation and rectification on three routes of information obtained from the data processing step, integrating to obtain positions of the geothermal resources and achieving multi-source distribution type remote sensing distinguishing of the geothermal resources. The multi-source distribution type remote sensing distinguishing method for the geothermal resources achieves accurate distinction of the geothermal resources due to processing, interpretation and integration of the multi-source remote sensing image data.

The invention discloses a testing and analyzing system for rotating errors of a main shaft. The testing and analyzing system comprises a hardware part and a software part, wherein the hardware part comprises a bracket or magnetic stand, laser displacement sensors, signal connecting wires, a signal adapter plate and a signal acquisition card, wherein the bracket or magnetic stand is fixedly arranged on the main shaft; the laser displacement sensors and the bracket or magnetic stand are connected to acquire axial displacement runout of the main shaft; the signal output end of each laser displacement sensor is connected with the input end of the signal acquisition card through the signal connecting wire and the signal adapter plate; the signal adapter plate is used for outputting more than two laser displacement sensors according to a unified format; a software system is arranged on a computer; the computer communicates with the output end of the signal acquisition card by using testing software, acquires a displacement signal of each laser displacement sensor, then performs analysis processing on a filtering algorithm and a separation algorithm and separates and displays roundness error and rotating error. According to the testing and analyzing system disclosed by the invention, the roundness error and the rotating error of the main shaft can be acquired quickly and accurately; the testing is simple and practical.

In an imaging method and device for water / fat separation in MRI using a two-point Dixon FSE sequence, each refocusing RF pulse corresponds to two readout gradients of the same polarity, each being center-divided into a smaller rear part and a larger front part, and one rephasing gradient of opposite polarity. In running the FSE sequence, each echo signal acquired is subjected to an FFT, to reconstruct an image with water and fat in phase and an image with water and fat in opposed phases. Data of each echo signal are subjected to a partial Fourier transform; and the in-phase water / fat image and the opposite-phase water / fat image are subjected to a water / fat separation algorithm, to obtain a pure water image and a pure fat image.

The invention discloses a multi-frequency alternative-ejecting real-time ultrasonic elastography method which comprises the steps of: 1, alternatively ejecting N kinds of frequency ultrasonic waves by an ultrasonic probe in a process of extracting tissues by using hands; 2, carrying out displacement estimation on adjacent same frequency echoes by using a two-dimensional weighting phase separation algorithm; 3, equally weighting N frames of continuous different frequency displacement images to generate a compound displacement image; 4, carrying out gradient operation on the compound displacement image to generate a strain image; and 5, carrying out downsampling on the strain image and carrying out gray scale mapping, and scanning and converting into an elastic image. According to the method, displacement images with different noise modes are generated by using different ejecting frequencies, the displacement estimation error is reduced through compounding of the displacement images, and the elastography pseudomorphism noise caused by speckle noise is inhibited. The signal to noise ratio of the elastic image generated by the compound image is higher than that of an elastic image generated by compounding the former frequency sub-displacement image, and therefore, the property and the quality of elastography are improved.

The invention relates to a fault diagnosis method of a lithium ion battery sensor of an electric vehicle based on an observer, and belongs to the technical field of battery management. The method comprises the following steps: determining parameters of a lithium ion battery, and establishing an electrothermal coupling dynamic model of the lithium ion battery of the electric vehicle; performing anopen circuit voltage test and an HPPC experiment on the measured battery to obtain characteristic parameters of the battery under different ambient temperatures; establishing a relationship between battery OCV and SoC, and identifying the parameters in the electrothermal coupling dynamic model by using a recursive least square method with a forgetting factor to obtain a quantitative relationship among the battery parameters, the ambient temperatures and the battery SoC; and importing current, a voltage and a temperature sensor measured value into a fault diagnosis and separation algorithm of the lithium ion battery based on the observer, estimating a quantity of state through an extended Kalman filter algorithm to generate a residual, and performing residual evaluation by using a CUSUM test method, and finally achieving fault diagnosis and separation of the lithium ion battery sensor according to combined response conditions of different residuals.

A double-echo based single-sweep quantitative magnetic resonance T2* imaging method relates to magnetic resonance imaging. Two low-angle excitation pulses of the same deflection angle are used, an evolution period is added after the first excited pulse, two echoes of different evolution time are generated, the two echoes include different horizontal relaxation time, and a shift gradient is added after each excitation pulse so that two echo signals are offset in the center of a signal space. The two echo signals are from the same imaging slice, so that the two echo signals can be separated by utilizing prior knowledge including structural similarity and sparsity of combined edges of the signals, and a proper separation algorithm corresponding to sparse transformation cooperation can be used to separate the two echo signals. T2* calculation is carried out on the two signals obtained by separating to obtain a quantitative T2* image.

The invention discloses a gradient-echo multi-echo water and fat separation method. The method includes the following steps of 1, using a three-dimensional gradient-echo N echo sequence for conductingimaging scanning on a magnetic resonance imaging region, collecting N pieces of echo data, and using a GRAPPA technology for accelerating data collection during collection, wherein N is greater thanor equal to 4; 2, using the GRAPPA technology for fitting and restoring to-be-collected phase encoding data of the N pieces of echo data within respective K spaces, and then conducting preliminary data processing to obtain complete image domain data of N echoes; 3, substituting the image domain data of the N echoes into a water and fat separation algorithm of a multi-peak water-fat model, introducing T2* variables, conducting iterative calculation, obtaining a T2* distribution map of an imaged object, and meanwhile obtaining water and fat images through calculation. According to the method, the T2* distribution can be accurately estimated, the signal attenuation between the echoes is effectively corrected, and phase changes caused by inhomogeneity of a magnetic field are reduced, so that results of the quantitative analysis images are more accurate and stable. A magnetic resonance imaging system using the method is also provided.

The invention relates to a profiled fiber automatic recognition method based on scale invariant feature extraction and vector machine classification. The method is characterized by comprising the following steps: (1) inputting the cross-section micro-sample images of a profiled fiber to be recognized and pre-processing the images; (2) carrying out binary processing of the pre-processed images to obtain a binary image of the fiber; (3) extracting the fiber object in the binary image by a separation algorithm to obtain the independent image of each fiber in the binary image; (4) extracting the eigenvector of the existing profiled fiber by an SIFT (scale invariant feature transform) algorithm to serve as the input, and training by an SVM (support vector machine) classification algorithm; and (5) using the independent image obtained in the step (3) as the input to determine the type of profiled fiber, to which the fiber on each independent image belongs, by the classification algorithm trained in the step (4). By utilizing the method, automatic classification and recognition of profiled fibers in complex background, tilting fibers, deformed fibers, broken fibers, fibers in illumination changing environment and the like can be achieved, and the problem of poor effectiveness of image feature extraction is solved.

The invention provides a plant health state detection method based on a thermal imaging technology. The plant health state detection method based on the thermal imaging technology provided by the embodiment of the invention comprises the following steps: firstly, reading a temperature data file comprising plants and reference leaves and carrying out a vegetation soil separation algorithm and a reference temperature algorithm to obtain a vegetation soil separated data file and a canopy temperature Tc; obtaining a reference temperature Tp according to temperature data of thermal infrared images of the reference leaves; and finally, calculating according to the canopy temperature Tc, the reference temperature Tp and an air temperature Ta to obtain a vegetation transpiration diffusion coefficient, so as to obtain a health condition grade of the plants. Compared with the prior art, a three-temperature model is used for calculating the vegetation transpiration diffusion coefficient in the embodiment and the plant growth vigor is further evaluated; the operation is simple and the efficiency is high.

The invention discloses a graph theory-based Chinese medicinal tongue nature and tongue coat separation algorithm, which comprises the following steps of: converting an acquired tongue image into a hue, saturation and value (HSV) model, and dividing a tongue coat initial segmentation area by using an H value; dividing the tongue image into a set of areas; forming an initial segmentation result by using the initial segmentation area and an independent vertex in the image; sequentially stacking differences among area combination operator calculation areas; when valid elements are in the stack, separating the areas if the difference of the stack top areas is greater than one of the internal differences of two areas, determining that the stack top elements are invalid, repeating the step, otherwise, merging the two areas, and deleting the stack top elements and the differences between the related areas of the two areas; and calculating related contents of the newly merged areas, stacking the differences among the areas, and resetting the invalid elements in effect. According to the characteristic of the H value of the HSV space, the tongue coat initial segmentation area is determined, and segmentation efficiency and precision of the algorithm are improved; a threshold function adaptive algorithm is constructed, and the nature and coat crossing detail distinguishing capacity of the algorithm is improved; and the iterative algorithm of the segmentation process is improved by using area combination operators, and smooth nature and coat transition areas are segmented according to the human eye perceptiveness.

The invention discloses a radar radiation source identification method based on singular value decomposition and one-dimensional CNN network, and mainly solves the problems of complex identification time and low identification precision of a radiation source identification technology in the prior art. The method comprises the following steps: adopting a separation algorithm to carry out signal separation and splitting a radar time sequence into a matrix G; performing singular value decomposition on the matrix G, and extracting a diagonal element forming vector lambda from the decomposed sigmamatrix; making a training set, a verification set and a test set from singular value vectors extracted from a plurality of groups of radar data; designing the one-dimensional CNN network structure aiming at the vector lambda; training the one-dimensional CNN network by using the training set; testing and training the performance of the network by using the verification set, and judging whether thenetwork is available or not; the test set is sent to a trained network, and the network output is a radiation source category. The method reduces the identification time of the radiation source underthe condition of ensuring the available identification rate, and can be used for the identification of the radar radiation source under the complex electromagnetic environment.

The invention relates to the jointed rock deformation modulus testing technology and provides a jointed rock deformation modulus testing method based on building a free interface incident wave separation algorithm in viscoelastic media. The method includes: testing jointed rock deformation modulus; building the reflection coefficient calculation formula of stress waves, in the viscoelastic media, on a free interface; acquiring incident P waves and SV waves according to testing point waveform separation; performing amplitude compensation of spherical dispersing; calculating the elastic modulus and viscose coefficient of a jointed rock; calculating the jointed rock deformation modulus. The method has the advantages that rock deformation modulus under different peak frequency dynamic load effects is obtained through testing; when peak frequency is small, the deformation modulus is static; when peak frequency is large, the deformation modulus is dynamic; corresponding deformation modulus can be selected according to the dynamic load feature borne by a rock project to perform related dynamic analysis, and the method is simple to operate and low in testing cost.

The invention relates to a parameter self-turning method for a torque / propeller pitch controller of a megawatt asynchronous double-feed wind driven generator. The method comprises the following steps: after a unit is started, a yaw system firstly works to an appointed position, subsequently enters into a self-check program and enters into a free rotation process after passing through the self-check program; when the rotation speed of a generator reaches a switching speed, the blade angle of a variable pitch system is completely opened, and the speed of the generator continuously increases; when the speed of the generator reaches a grid-connection speed, a torque controller controls the rotation speed of the generator to increase according to a certain slope; and when the rotation speed of the generator reaches a rated rotation speed (rated power), a variable current controller keeps the torque of the generator constant, and a variable pitch controller controls the rotation speed of the generator to be stabilized in the rated rotation speed. The propeller pitch and torque controller comprises a PID (Proportional-Integral-Derivative) parameter self-turning controller of a BP (Back Propagation) neural network with an integral separation algorithm, the controller can automatically adjust the parameters according to rules, and the parameters are optimized through on-line training according to convergence rules, so that the power generation amount of the wind driven generator is increased, and the load of the wind driven generator is effectively reduced.

The invention provides a radar repeated-frequency separation method based on pulse time delay matching. The radar repeated-frequency separation method utilizes a separation algorithm, pulse sequence repeated-frequency testing is accurate, and the problems such as misinformation, missed reporting and excessive reporting of repeated-frequency separation in the actual usage process can be decreased.

The invention belongs to the field of computer graphic images and relates to a seven color separation algorithm based on a cellular element partition neugebauer equation. By further optimizing the neugebauer equation color separation algorithm, a seven color separation model is built on the basis of the cellular element partition neugebauer equation, a cellular element search algorithm suitable for the color separation model is designed, the calculated amount is small, search accuracy is high, and the problem that the neugebauer equation color separation precision is poor is solved simultaneously.

The embodiment of the invention discloses a method and a device for intercepting voice of a target person in a video. The method comprises the following steps of using a lip-shaped voice activity detection model, giving a first mark to a video frame, subjected to voice activity, of a target person in the audio and video file, a second mark is given to the video frame, not subjected to the voice activity, of the target person; obtaining a first marker sequence, continuously setting a preset number of first start-stop time points of the video frames containing the first mark in the first mark sequence; determining a second start-stop time point of a corresponding voice frame in the audio and video file, Therefore, the corresponding voice segment in the audio and video file is directly intercepted according to the second start-stop time point. According to the method and the device, the voice segment file of the target person is obtained through the human-voice separation algorithm, human-voice separation is realized, and the technical problems that the requirement of the current human-voice separation algorithm on the definition of audio is high, the audio needs to be subjected to noise reduction processing first and then subjected to human-voice separation, the noise influence is large in a noisy environment, the voice interception difficulty is high, and the voice interceptionefficiency is low are solved.

The invention discloses a water and fat separation method based on under-sampling k-space data. The method includes steps of, step one, respectively acquiring k-space trajectory parameters and k-space raw data according to magnetic resonance (MR) raw data; step two, selecting a parameter system matrix according to the k-space trajectory parameters; step three, acquiring echo data of each channel according to the k-space raw data; step four, executing regularization iteration image reconstruction by using the parameter system matrix and the echo data of each channel; and step five, subjecting reconstructed images to variable projection (VARPRO) water and fat separation calculating to obtain a final image. The water and fat separation method based on the under-sampling k-space data is applicable to more scanning tracks, and by means of a VARPRO water and fat separation algorithm, a phase diagram, a B0 field image, a water image and a fat image which are irrelevant to main magnetic field inhomogeneity are separated; the water and fat separation method based on the under-sampling k-space data has advantages of VARPRO iteration field image calculating and is insensitive to seed points; and the water and fat separation method based on the under-sampling k-space data is smaller in calculated amount and fewer in calculating requirements, image merging is easier to finish, and a system function which is capable of achieving parallel imaging can be introduced.

The invention discloses a pigment separation method of centralized image channel difference based on optical density space, which comprises the following steps of acquiring an RGB color skin image under polarized light and extracting skin color; dividing the image into a plurality of sub-image blocks and discarding the sub-image blocks containing non-skin pixels; obtaining the pure color density of each sub-image block; obtaining a matrix optimum solid color concentration matrix; calculating the concentrations of melanin and hemoglobin in the image, and introducing the gray distribution and color distribution of the pigment concentration. According to the present invention, a light source interference term in a conventional pigment separation model is eliminated without principal componentanalysis, and a simple and efficient constraint condition for judging the rationality of the pure color concentration matrix is designed, thereby avoiding, in the existing pigment separation algorithm, the inaccuracy and robustness of the pigment concentration extraction result due to the interference of light source intensity and the data loss in the principal component analysis process, as wellas the low efficiency of the algorithm due to the complexity of the determination process of the pure color concentration matrix.

The embodiment of the invention discloses a wireless sensor network media access control method based on multipacket reception and a communication system. The method comprises the following steps: a first node in an idle state sensing a carrier of a channel, and if the channel in the sensing range of the first node is idle, sending RTR broadcast information; when a second node has a data packet to be sent to the first node, returning RTS response information to the first node after standing back one time; the first node selecting M target nodes from the second node to send ATS broadcast information according to the current state of the channel; M target nodes sending the data packets to the first node; the first node separating and identifying the data packets separately sent by the M target nodes through a multipacket separation algorithm and returning ACK broadcast information to the corresponding target nodes sending the data packets after the data packets are correctly identified. According to the embodiment of the invention, the control information expenditure in a network can be reduced, and the network throughout is increased.

The invention provides an engineering drawing vector conversion and primitive semantic extraction method. The method comprises the following steps: S1, utilizing a scripting language to extract information data in a PDF (Portable Document Format) drawing document or generate corresponding BMP (Bitmap) image data, and further pre-processing the document in the image format; S2, utilizing the seed segment algorithm to vectorize straight-line primitives in the image data; S3, vectorizing circles / circular arcs in the image data; S4, identifying the primitive relationship and utilizing the text / graphics separation algorithm based on connected region to remove labeling frames and graphics information on the drawing images; and S5, outputting the result. Through pretreatment of the text / graphics separation algorithm and the improvement on the seed segment vector conversion method, the engineering drawing vector conversion and primitive semantic extraction method improves the vector conversion accuracy and processing speed, and better solves the difficult problems in extracting the pattern semantic information through the rule matching algorithm, thereby providing a better data foundation for intelligent machine drawing reading and three-dimensional reconstruction.