45results about How to "Reduce condition number" patented technology

In illustrative implementations, a time-of-flight camera robustly measures scene depths, despite multipath interference. The camera emits amplitude modulated light. An FPGA sends at least two electrical signals, the first being to control modulation of radiant power of a light source and the second being a reference signal to control modulation of pixel gain in a light sensor. These signals are identical, except for time delays. These signals comprise binary codes that are m-sequences or other broadband codes. The correlation waveform is not sinusoidal. During measurements, only one fundamental modulation frequency is used. One or more computer processors solve a linear system by deconvolution, in order to recover an environmental function. Sparse deconvolution is used if the scene has only a few objects at a finite depth. Another algorithm, such as Wiener deconvolution, is used is the scene has global illumination or a scattering media.

The invention relates to the field of SCARA robot kinetic parameter identification, and discloses an improved SCARA robot kinetic parameter identification method. A Lagrange method is adopted for establishing an SCARA robot integral kinetic model comprising a friction term, and the SCARA robot integral kinetic model is subjected to linearization. The five-order fourier series is adopted as the stimulation track basic form, the polynomial with 5 degree is used for replacing the constant term in the traditional fourier series, and the joint angular velocity and angular acceleration are zero at the track start and stop time; and the minimum observing matrix conditional number serves as the target, and to enhance the global optimization capability, various coefficients of the stimulation trackare optimized through the niched pareto genetic algorithm based on the crowding-out mechanism. To avoid the transmission error directly caused to the joint angle two-time differential, the sampled joint angular velocity data are fitted into the fourier series form, and a joint acceleration signal is obtained through the differential. Considering the measuring noise influence, the WLS is adopted as the parameter estimation method.

Image and video processing using multi-scale amplitude-modulation frequency-modulation (“AM-FM”) demodulation where a multi-scale filterbank with bandpass filters that correspond to each scale are used to calculate estimates for instantaneous amplitude, instantaneous phase, and instantaneous frequency. The image and video are reconstructed using the instantaneous amplitude and instantaneous frequency estimates and variable-spacing local linear phase and multi-scale least square reconstruction techniques. AM-FM demodulation is applicable in imaging modalities such as electron microscopy, spectral and hyperspectral devices, ultrasound, magnetic resonance imaging (“MRI”), positron emission tomography (“PET”), histology, color and monochrome images, molecular imaging, radiographs (“X-rays”), computer tomography (“CT”), and others. Specific applications include fingerprint identification, detection and diagnosis of retinal disease, malignant cancer tumors, cardiac image segmentation, atherosclerosis characterization, brain function, histopathology specimen classification, characterization of anatomical structure such as carotid artery walls and plaques or cardiac motion and as the basis for computer-aided diagnosis to name a few.

The invention provides a element two-stage partition and twice polycondensation parallel method based on multiple document flows. The method includes the steps of firstly, dividing a finite element model into a plurality of initial sub-areas, and saving model information of each initial sub-area independently in a document; secondly, by each progress, simultaneously and independently forming a system equation of each sub-area, and performing polycondensation to remove internal freedom degree of each system equation; thirdly, by a local main process, forming packaging sub-area system equations, and performing polycondensation to remove internal freedom degree of each packaging sub-area system equation; fourthly, by the local main process, solving an interface equation, and solving the internal freedom degree of each packaging sub-area system equation by means of back substitution; fifthly, solving internal freedom degree of each initial sub-area according to results of the top packaging sub-areas by means of back substitution; terminating if iteration ends, or else starting again from the second step. On the basis of realizing distributed data storage, large-scale finite element parallel computing efficiency is increased by fully utilizing structural features of large-scale parallel computer systems by means of maximally realizing communication localization.

The present invention discloses a composite wave plate phase delayer optimal design method which comprises the following steps of (1) forming a composite wave plate phase delayer by n quarter zero order wave plates, (2) determining the change range of the central wavelength and optical axis azimuth of each quarter zero order wave plate, (3) obtaining the transmission Jones matrix of a kth quarter zero order wave plate, (4) obtaining the equivalent Jones matrix of the composite wave plate phase delayer, (5) obtaining the equivalent characteristic parameter of a composite wave plate, (6) obtaining the system matrix of an ellipsometer measurement system, (7) obtaining the system matrix condition number of the ellipsometer measurement system, (8) obtaining the maximum value of the ellipsometer measurement system in a wave band range Gamma, and (9) adjusting the central wavelength and optical axis azimuth of each quarter zero order wave plate, obtaining the maximum value of each adjustment, and then obtaining the minimum value in all maximum values. The method can be well adapted to the use need of the ellipsometer measurement system.

The invention discloses a pose estimation method for a visual inertial navigation odometer based on a sparse feature method. The method mainly comprises system initialization, inertial sensing unit state transfer, image feature point tracking, inertial sensing unit error item construction, system state quantity marginalization, visual measurement model error item construction and observability matrix constraints. the square root information filtering method is used to replace a high-latitude matrix inversion step of the traditional filtering method with a QR decomposition method with a fastercalculation speed, and line features are introduced at the same time to increase the smoothness of the estimated trajectory so as to construct a brand new visual inertial navigation odometer framework.

The present invention discloses a method of large-scale MIMO linear iterative detection under non-ideal channel. The method utilizes the incomplete Cholesky decomposition as the preprocessing part of the conjugate gradient iteration, combining the advantages of the two, so that the problem of reduced rate of convergence in the traditional conjugate gradient algorithm under non-ideal channel or in a expanded system.Furthermore, a dynamic threshold based on channel correlation coefficient and system scale is proposed as an important parameter of preprocessing, which makes the solution of the whole system well balanced between complexity and performance.

The invention discloses a calibration method used for a fully polarimetric microwave radiometer, which accomplishes calibration of the fully polarimetric microwave radiometer through steps as follows: setting initial values of a calibrating device of the fully polarimetric microwave radiometer, confirming a process matrix, confirming a matrix, confirming the condition number of the matrix, and confirming the minimum Kmin(S). The method gives a judgment basis for selecting vectors from a plenty of vectors which can be used in calibrating the fully polarimetric microwave radiometer so as to acquire higher calibrating accuracy. after the minimum condition number optimization is adopted, the obtained fully polarimetric microwave radiometer calibrating matrix has a smaller condition number, the matrix inversion result is enabled to be more stable, the error bound is smaller, and the calibrating accuracy of the fully polarimetric microwave radiometer can be improved. The method disclosed by the invention is characterized in that the error perturbation matrix is applied onto the calibrating inversion matrix of the fully polarimetric microwave radiometer, the calibration source temperature is taken as an optimized parameter, and the minimum condition number is taken as the basis for judging the ill-condition of the calibrating inversion matrix of the fully polarimetric microwave radiometer.

The invention provides a multi-support piezoelectric dynamometer calibration method. The method is characterized by, through a translational main force loading mechanism, ensuring that force lines ofthe main load force and axes of piezoelectric three-dimensional force sensors are overlapped; applying full scale of main force, repeating applying the force for many times, and collecting and repeating output voltage; carrying out pretreatment on the data, and obtaining a force-power relation expression at each loading point according to the principle of equilibrium of forces; and combining the force-power relation expressions, constructing a calibration equation set model, calculating condition number of a coefficient matrix, calculating force-power conversion coefficient of each piezoelectric three-dimensional force sensor and finishing calibration of a dynamometer. The method, based on an existing calibration device and by improving the thought of calibration, sets a plurality of calibrated variables as unknown parameters, carries out multi-point calibration and establishes a multi-variable one-time calibration equation set without decoupling hypothesis of each support force, thereby solving the problem of principle error due to force distribution hypothesis in the previous calibration; and meanwhile, the method is not sensitive to position precision of each calibration point,and the positions at the axes of the sensors are selected as the calibration points.

The invention relates to a three-dimensional reconstruction method in the field of computer vision, and particularly relates to an on-line tracking and dense reconstruction method, an on-line trackingand dense reconstruction system and on-line tracking and dense reconstruction equipment on mobile equipment. The invention aims to realize the on-line dense three-dimensional reconstruction on the mobile equipment. The three-dimensional reconstruction method comprises the following steps: acquiring a current depth image of a frame, calculating a three-dimensional point and a normal vector of eachpixel point under a camera coordinate system, and calculating a depth noise standard deviation and a gradient graph; selecting a point pair capable of providing enough constraint for an ICP algorithmbased on a geometric stability sampling method, and adopting a visual tracking and IMU information fusion method based on an ICP covariance matrix condition number; after calculating the posture of acamera, fusing a depth map into a TSDF model; carrying out light projection on fused volume data at the calculated posture of the camera so as to obtain a surface point cloud. The method is high in robustness and fast in calculation speed. Meanwhile, based on the method, a finally obtained three-dimensional model is reasonable and reliable.

The invention belongs to the computer image processing field and relates to a color image vignetting recovery method based on illumination surface modeling. The method comprises the following steps: converting an original color vignetting image into an HSI space; extracting foreground and background regions of the vignetting image; using a color image decomposition technology to realize the decomposition of a structure and a texture of the original color vignetting image and acquire a color structure image, and converting the color structure image into the HSI space; in the background region of the image, selecting central points according to an uniformity principle; calculating a radial distance between any two central points; modeling all the points to be modeled and acquiring a curved surface after the modeling; taking the curved surface acquired through modeling as a vignetting function so as to acquire the color image after the vignetting recovery. The method of the invention has a good adaptability and is easy to be realized in a real project.

The present invention relates to a method for configuring a management unit connected to at least one home automation equipment comprising at least one home automation device (D) and at least one central control unit (U), the method being implemented by a management unit (Sv) and comprising the following steps: configuring (EcfSv2) an alert (Al) corresponding to the triggering of an alert notification and / or an action when a triggering condition (Cnd) is produced relating to at least one state variable for a home automation device (D), a group of home automation devices (D), a type of home automation device (DT) or a group of types of home automation devices (DT); the step of configuring an alert (Al) being carried out on the basis of instructions of a first user (Usr1) having a first user profile type; declaring (ECfUsr26) the monitoring of an assembly of home automation devices (D) comprising at least one home automation device (D) for which at least one alert (Al) has been configured by a second user (Usr2) having a second user profile type, downloading the alert (Al) condition (Cnd) on at least one central control unit (U) to which at least one home automation device (D) belonging to the group of home automation devices (D) is connected, in order to configure the home automation device (D) for evaluating the condition (Cnd) and sending a notification of production of the condition to the management unit. The present invention also relates to a method for monitoring said equipment.

In illustrative implementations, a time-of-flight camera robustly measures scene depths, despite multipath interference. The camera emits amplitude modulated light. An FPGA sends at least two electrical signals, the first being to control modulation of radiant power of a light source and the second being a reference signal to control modulation of pixel gain in a light sensor. These signals are identical, except for time delays. These signals comprise binary codes that are m-sequences or other broadband codes. The correlation waveform is not sinusoidal. During measurements, only one fundamental modulation frequency is used. One or more computer processors solve a linear system by deconvolution, in order to recover an environmental function. Sparse deconvolution is used if the scene has only a few objects at a finite depth. Another algorithm, such as Wiener deconvolution, is used is the scene has global illumination or a scattering media.

The invention discloses an array antenna and a signal transmitting and receiving method and device. The array antenna comprises at least two active antennas, wherein every active antenna is wound with at least two different passive antennas, every passive antenna is coupled to a controllable load, and the controllable load is coupled to a control circuit. According to the array antenna, the at least two active antennas are provided, every active antenna is wound with the at least two different passive antennas and every passive antenna is coupled to the controllable load, so that when channels are in good conditions, the volumes of the channels can be adjusted by adjusting the controllable loads connected with the passive antennas, and when the channels are not in good conditions, by adjusting the controllable load coupled with the passive antennas simultaneously and adjusting the transmitting diagram of the active antennas according to adjustment vectors, the error rate can be reduced, and meanwhile, miniaturization can be achieved.

The invention discloses a method for characterizing grain size of magnetic nanometer grains. The method comprises the following steps of: on the basis of the Fourier analysis of a magnetization curve of an analysis model, working out the minimum number of discretization points needed for the characterization of the magnetization curve within a given tolerance error; providing a weighted error based optimal quantization method on the basis of the Lloyd-Max optimal quantization method; and applying the weighted error based optimal quantization method to the problem of solving a grain size distribution function according to a magnetization numerical equation of the magnetic nanometer grains, so that the condition number of a matrix equation is reduced, and the solving precision is improved.

The invention provides an MFTN joint channel estimation and equalization method under a time-varying frequency selective fading channel. The problems of complex channel estimation and strong coupling interference elimination of an MFTN system under the time-varying frequency selective fading channel can be solved. A small number of one-dimensional frequency domain cyclic postfixes are inserted, time domain segmentation detection is carried out on signals in a receiver, the two-dimensional self-interference length is truncated, a segmentation frequency domain receiving signal model of an MFTN system under a time-varying frequency selective fading channel is constructed, and the model has an equivalent channel matrix. The equivalent channel matrix has a block cyclic structure and each cyclic block is a diagonal matrix, so that the condition number of the equivalent channel matrix can be effectively reduced, the influence of an ill-conditioned problem introduced by time-frequency two-dimensional compression in the MFTN system on receiving demodulation performance is reduced, the method can be expanded to the MFTN system with time-frequency two-dimensional coupling interference, the two-dimensional strong self-interference of the MFTN system can be effectively eliminated, and the complex channel interference introduced by the time-varying fading channel can be processed.

The invention discloses a bilinear WLAV robust state estimation method considering matrix scale scaling. According to the bilinear WLAV robust estimation, an intermediate variable is introduced, a nonlinear measurement equation of a power system is converted into a two-step linearization process, the model is simplified, and the estimation time is shortened. However, when a long branch and a shortbranch occur in a power grid, leverage measurement and information matrix conditions are caused, and the numerical stability and bad data detection and identification of bilinear WLAV robust estimation are influenced. According to the method, the Jacobian matrix is scaled and stretched, so that lever point measurement caused by long and short branches can be effectively eliminated, the number ofconditions of the information matrix is reduced, the numerical stability is improved, and the engineering application value is achieved.

A system to repeat MIMO signals in a network by orthogonally polarizing the signals and maintaining the orthogonal polarization.

The embodiment of the invention provides a motion state estimation method and device, and the method comprises the steps: obtaining a plurality of measurement data through a first sensor, wherein eachmeasurement data at least comprises speed measurement information; and according to the measurement data corresponding to the target reference object in the plurality of measurement data, obtaining amotion state of the first sensor, the motion state at least comprising a speed vector of the first sensor. By adopting the embodiment of the invention, the more accurate motion state of the first sensor can be obtained.

The application discloses a least squares fitting method and a least squares fitting system, wherein the least squares fitting method performs least squares fitting on the data pairs in each division interval Perform data reconstruction on it; the data reconstruction includes translating the data pair to ensure that the translated data pair has the conditions for number domain mapping, and then mapping the translated data pair from the original number domain to a smaller preset In the number field, avoid the occurrence of ill-conditioned calculations due to the high degree of the polynomial of the corresponding function to be fitted after the translation of the data pair; finally adjust the order of magnitude of the data pair in the preset number field to reduce the normal equations condition number, thereby reducing the possibility of ill-conditioned calculations when performing least squares fitting on the reconstructed data pairs, thereby reducing the cost of the fitting process for function fitting using the least squares fitting method time.

The invention discloses a scattered field solving method applied to a simulated FSS structure. Compared with a traditional triangular-tetrahedron and RWG-SWG primary function VSIE method for solving the scattering field, the method solves the problem of frequency deviation, the number of unknown quantities and the memory overhead of calculation can be greatly reduced, and meanwhile the solving efficiency is higher. According to the method, the design of the FSS structure can be guided more efficiently and accurately, analogue simulation of a large-scale array is facilitated, and the method haspractical engineering application value.

Accelerated machine learning using an efficient preconditioner for Kernel Ridge Regression (KRR). A plurality of anchor points may be selected by: projecting an initial kernel onto a random matrix in a lower dimensional space to generate a randomized decomposition of the initial kernel, permuting the randomized decomposition to reorder its columns and / or rows to approximate the initial kernel, and selecting anchor points representing a subset of the columns and / or rows based on their permuted order. A reduced-rank approximation kernel may be generated comprising the subset of columns and / or rows represented by the selected anchor points. A KRR system may be preconditioned using a preconditioner generated based on the reduced-rank approximation kernel. The preconditioned KRR system may be solved to train the machine learning model. This KRR technique may be executed without generating the KRR kernel, reducing processor and memory consumption.

The invention provides a frequency domain equalization method based on generalized approximated message passing, which can solve the ill-conditioned regeneration in the message updating process, effectively alleviate the ill-conditioned problem of MFTN signals, and avoid the problem that a receiving demodulation algorithm is difficult to converge. The low-complexity balance of an MFTN system is solved when two-dimensional strong self-interference and color noise exist, the application limitation of a generalized approximated message passing algorithm communication system can be broken through, and the situation that a receiving demodulation algorithm is difficult to converge is avoided. According to the method, a block cyclic equivalent channel matrix with a cyclic block is constructed, only a small number of cyclic postfixes (CP) need to be inserted to construct a frequency domain system model, the condition number of the equivalent channel matrix in an MFTN system is effectively reduced, and therefore the influence of the MFTN signal on the performance of a receiver due to the ill-conditioned problem is relieved.

The invention discloses a calibration method used for a fully polarimetric microwave radiometer, which accomplishes calibration of the fully polarimetric microwave radiometer through steps as follows: setting initial values of a calibrating device of the fully polarimetric microwave radiometer, confirming a process matrix, confirming a matrix, confirming the condition number of the matrix, and confirming the minimum Kmin(S). The method gives a judgment basis for selecting vectors from a plenty of vectors which can be used in calibrating the fully polarimetric microwave radiometer so as to acquire higher calibrating accuracy. after the minimum condition number optimization is adopted, the obtained fully polarimetric microwave radiometer calibrating matrix has a smaller condition number, the matrix inversion result is enabled to be more stable, the error bound is smaller, and the calibrating accuracy of the fully polarimetric microwave radiometer can be improved. The method disclosed by the invention is characterized in that the error perturbation matrix is applied onto the calibrating inversion matrix of the fully polarimetric microwave radiometer, the calibration source temperature is taken as an optimized parameter, and the minimum condition number is taken as the basis for judging the ill-condition of the calibrating inversion matrix of the fully polarimetric microwave radiometer.

The invention discloses a quantum preprocessing method and device for a linear system, and the method comprises the steps: obtaining the element information of a first matrix A and a first vector b in the linear system, constructing a new matrix M for the preprocessing of the linear system according to the main diagonal element of the first matrix A, and carrying out the preprocessing of the linear system according to the inverse matrix of the new matrix M, and calculating a second matrix A'and a second vector b ', finally constructing a first quantum circuit representing quantum state evolution of specific class elements of the second matrix A' and a second quantum circuit representing quantum state evolution of specific class elements of the second vector b ', and respectively executing quantum state evolution operation for the first quantum circuit and the second quantum circuit. And obtaining the quantum states of the evolved first quantum circuit and second quantum circuit. The quantum preprocessing technology capable of meeting different linear systems is realized by utilizing the superposition characteristic of quantum, is used for simulating quantum calculation, and fills up the relevant technical blank in the field of quantum calculation.

A two-freedom-degree vibration restraining system applied to a high-altitude building robot operating platform comprises a force sensor, two driving rods with the same length, two driven rods with the same length, a horizontal adjusting motor, a magnetic powder clutch, a rotary type magneto-rheological damper, an installing plate and a vertical vibration reducing mechanism. The two-freedom-degree vibration restraining system is characterized in that two installing seat area arranged on the inner surface of the installing plate; each installing seat is connected with the front end of one driving rod through a driving rod controlling shaft, the tail end of each driving rod is connected with the front end of one driven rod, and the tail ends of the two driven rods are connected together; the installing plate, the driving rods and the driven rods form a planar five-connecting-rod mechanism; and the planar five-connecting-rod mechanism can move in a plane perpendicular to the inner surface of the installing plate and has two freedom degrees. The vertical vibration reducing mechanism is arranged on the outer surface of the installing plate.

The invention relates to a three-dimensional reconstruction method in the field of computer vision, and particularly relates to an on-line tracking and dense reconstruction method, an on-line trackingand dense reconstruction system and on-line tracking and dense reconstruction equipment on mobile equipment. The invention aims to realize the on-line dense three-dimensional reconstruction on the mobile equipment. The three-dimensional reconstruction method comprises the following steps: acquiring a current depth image of a frame, calculating a three-dimensional point and a normal vector of eachpixel point under a camera coordinate system, and calculating a depth noise standard deviation and a gradient graph; selecting a point pair capable of providing enough constraint for an ICP algorithmbased on a geometric stability sampling method, and adopting a visual tracking and IMU information fusion method based on an ICP covariance matrix condition number; after calculating the posture of acamera, fusing a depth map into a TSDF model; carrying out light projection on fused volume data at the calculated posture of the camera so as to obtain a surface point cloud. The method is high in robustness and fast in calculation speed. Meanwhile, based on the method, a finally obtained three-dimensional model is reasonable and reliable.

The invention discloses a unified per-unit calculation method for hybrid energy flow of an electricity-water interconnection system. The method comprises the following steps of 1) acquiring basic dataof the electricity-water interconnection system and initializing the basic data; 2) establishing a power flow model of the power system; 3) establishing a steady-state model of the water distributionsystem; 4) establishing a coupling link steady-state model; 5) based on the power system power flow model, the water distribution system steady-state model and the coupling link steady-state model, establishing a unified per-unit electric-water interconnection system energy flow model; and 6) resolving the unified per-unit energy flow model of the electricity-water interconnection system by usinga Newton method to obtain electricity-water mixed energy flow. The invention provides a unified per-unit method of an electricity-water combined system in order to simplify system description and improve calculation efficiency aiming at the conditions of large numerical value difference and complex conversion relation in calculation of named values in a power grid and a water distribution network.

The invention discloses a monitoring data noise reduction method based on urban pipeline leakage in the field of data noise reduction. The monitoring data noise reduction method comprises the steps of collecting leakage original data Xi (t) of infrasonic waves of a leakage pipeline; establishing an amplitude spectrogram and an effective frequency window of the frequency domain signal XF, i (s); dividing the frequency domain signal XF, i (s) into M sections of unit signals xm, removing the unit signals xm influenced by noise by using a differential energy model, and recombining to form a stable signal xc, i; carrying out Prony decomposition on the stable signal xc, i, and calculating an extreme value and a residue; after the extreme value and the residue are screened according to the effective frequency window, the real signal is reconstructed, and a complete signal x (t) after denoising is completed is obtained; according to the method, structural inherent information in leakage signals can be prevented from being lost, meanwhile, the condition number of a matrix is reduced, calculation is convenient and rapid, and stability is better.