50 results about "Constrained least squares" patented technology

A method for accurately synthesizing a full-aperture data map from a series of overlapped sub-aperture data maps. In addition to conventional alignment uncertainties, a generalized compensation framework corrects a variety of errors, including compensators that are independent in each sub-aperture. Another class of compensators (interlocked) include coefficients that are the same across all the sub-apertures. A constrained least-squares optimization routine maximizes data consistency in sub-aperture overlap regions. The stitching algorithm includes constraints representative of the accuracies of the hardware to ensure that the results are within meaningful bounds. The constraints also enable the computation of estimates of uncertainties in the final results. The method therefore automatically calibrates the system, provides a full-aperture surface map, and an estimate of residual uncertainties. Therefore, larger surfaces can be tested with greater departures from a best-fit sphere to greater accuracy than was possible in the prior art.

An inversion-based algorithm for computing the time frequency analysis of reflection seismograms using constrained least-squares spectral analysis is formulated and applied to modeled seismic waveforms and real seismic data. The Fourier series coefficients are computed as a function of time directly by inverting a basis of truncated sinusoidal kernels for a moving time window. Spectra may be provided that have reduced window smearing for a given window length relative to the discrete Fourier transform irrespective of window shape, and a time-frequency analysis with a combination of time and frequency resolution that is superior to the short time Fourier transform and the continuous wavelet transform. The reduction in spectral smoothing enables enhanced determination of spectral characteristics of interfering reflections within a short window. The degree of resolution improvement relative to the short time Fourier transform increases as window length decreases.

A fast combinatorial algorithm can significantly reduce the computational burden when solving general equality and inequality constrained least squares problems with large numbers of observation vectors. The combinatorial algorithm provides a mathematically rigorous solution and operates at great speed by reorganizing the calculations to take advantage of the combinatorial nature of the problems to be solved. The combinatorial algorithm exploits the structure that exists in large-scale problems in order to minimize the number of arithmetic operations required to obtain a solution.

The invention provides a luminosity diagram-based geometry and surface material reconstructing method. The method comprises the following steps of: (1) segmenting a material, constructing an initialized material weight figure, establishing Diffuse Albedo; (2) estimating a surface normal vector diagram; (3) iterating and solving a normal vector by using an Alternating-Constrained-Least-Square method and constructing a bidirectional reflectance distribution function (BRDF) material substrate; and (4) enhancing a depth filed. On the basis of a normal vector diagram, the least square surface depth information is obtained by a Poisson equation. By the method, different material areas and material attributes and the surface depth information of an object in a picture can be recovered automatically from a group of picture sequences with known light conditions reliably; and the object is subjected to three-dimensional reconstruction rapidly.

The present invention is embodied in a system and method for extracting structure from multiple images of a scene by representing the scene as a group of image layers, including reflection and transparency layers. In general, the present invention performs layer extraction from multiple images containing reflections and transparencies. The present invention includes an optimal approach for recovering layer images and their associated motions from an arbitrary number of composite images. The present invention includes image formation equations, the constrained least squares technique used to recover the component images, a novel method to estimate upper and lower bounds on the solution using min- and max-composites, and a motion refinement method.

The invention discloses a non-linear constrained least square-based single station positioning method. The method comprises the steps of first, building an NLOS signal transmission model of a macro cell single base station according to a signal transmission characteristic of a macro cell; then, converting a positioning problem into a non-linear constraint least square problem by using a position relationship among a multi-path signal parameter, a base station, a target and a scatterer; resolving the least square problem by using an LM (Levenberg-Marquardt) algorithm; and when a residual is large, introducing an aquasi-Newton algorithm, and improving the LM algorithm so as to obtain a target estimation position. The method solves a problem that a traditional positioning algorithm depends on the number of base stations, performs positioning by using a single base station, does not need clock synchronization, and lowers the algorithm complexity; in the NLOS environment, the multi-path signal is used as a positioning path but does not suppress multiple paths, so that a problem of large positioning error in the NLOS environment is solved; the variable constraints are increased by using a ring model, so that the least square method is more effective, and the single station positioning of the macro cell reaches the positioning accuracy of 20-30 meters.

The invention discloses a low delay filter design method for sampling rate conversion in an electronic transformer. The method comprises the steps of first adopting cascade connection of an interpolator and an extractor to achieve conversion of random fraction ratio sampling frequency, enabling an anti-image filter in the interpolator and an anti-aliasing filter in the extractor to be combined into a low pass filter; then adopting a mean square error minimization standard to solve a coefficient vector of the filter, enabling a filter transmission band amplitude and a stop band amplitude to serve as constraint conditions, and enabling mean square error minimization to serve as an optimized target; and finally enabling a solving process of the filter coefficient vector based on constraint least square method design to be converted into solving of a positive definite quadratic programming problem so that the filter coefficient vector can be obtained by direct solving. According to the method, the problems that large output delay is caused in the direct linear convolution filtering process and the higher the filter order, the larger the group delay are solved, and rapidity of protection action is greatly improved.

The invention provides a point cloud accurate registration method based on a robust constraint least square algorithm. The method comprises the following steps: firstly, collecting point cloud data with a certain overlapping degree observed by a measured object from different visual angles according to a ground three-dimensional laser scanner, and carrying out coarse registration on a point cloudto be registered and a target point cloud through a PCA algorithm, so that the two point clouds have a relatively good initial position; secondly, establishing a kd-tree in the target point cloud, searching for points corresponding to the point cloud to be registered, and then reserving the points with the large feature degree according to the normal vector features of the point cloud; and finally, unifying the point cloud to be registered and the target point cloud to the same coordinate system through rotation and translation by using a robust constraint least square algorithm according to the obtained corresponding point pairs. The method effectively solves the problems that the original point cloud data size is large, the precision is not high and trigonometric function solving is complex, avoids falling into a local optimal solution in the registration process, shortens the registration time, and improves the precision of point cloud registration.

The invention discloses a trend-aware robot trajectory planning method in a dynamic environment. The method comprises the following steps of: cutting, discretizing and timing a path planned by a global path planner to change the path into a discrete initial trajectory, and awaring and estimating the speed of a dynamic obstacle to predict the motion trajectory of the dynamic obstacle; then constructing trend trajectories of the robot and the dynamic obstacle according to the initial trajectory of the robot and the motion trajectory of the dynamic obstacle, and constructing constraint conditionsthrough the intersection and overlapping degree of the trend trajectories of the robot and the dynamic obstacle; and finally, mapping the constructed constraint conditions to a hypergraph, convertingthe hypergraph into an unconstrained least square optimization problem with constraint approximation, solving the problem, and optimizing the robot trajectory. Obstacle trend constraints are introduced, the collision probability of the robot and the dynamic obstacle can be effectively reduced, the track generation stability is improved, and a safe and reliable robot motion track is provided for automatic driving, social services and other scenes.

The invention relates to a layered interface rebuilding method based on electricity/ultrasonic bimodal fusion. The layered interface rebuilding method includes the steps that a horizontal center plane of a pipe is defined as a reference plane, a set of forward-direction control points are selected on a target interface, and the distance from the forward-direction control points to the reference plane is recorded as forward-direction distance; the center forward-direction distance dc obtained through ultrasonic serves as the constraint condition of an electricity imaging objective function, and the constrained least-squares problems are built; electricity imaging objective functions with apriori constraints are solved with the shape rebuilding algorithm, and the optimum solution of shape parameters is obtained; the obtained shape parameters are converted into the corresponding target interface with shape functions, and the geometric silhouette of the target interface is drawn. The layered interface rebuilding method has the advantages of being high in rebuilding accuracy.

The invention discloses a watershed scale vegetation net primary productivity remote sensing fine inversion method. The method comprises the following steps: acquiring a vegetation type distribution diagram, remote sensing image data and site meteorological data of a measured area; interpreting the vegetation type distribution diagram; calculating the remote sensing image data to obtain normalized vegetation index time sequence data; carrying out spatial interpolation on the station meteorological data to obtain raster data with the same resolution as the normalized vegetation index time sequence data; improving the resolution of the data by using a fully constrained least square mixed pixel decomposition and sub-pixel spatial gravitation model; configuring static parameters; and using the improved NPP estimation model to invert the vegetation net primary productivity NPP. On the basis of existing low-resolution remote sensing data, the method for improving the product resolution by means of sub-pixel mapping is provided, the vegetation net primary productivity is accurately estimated, the estimation precision of the net primary productivity is improved, and remote sensing fine inversion of the vegetation net primary productivity in the watershed scale is achieved.

The invention discloses a waveform synchronization method and system for virtual instruments. A random noise sequence is added to an input signal; the input signal containing the random noise sequence is sampled; curve fitting is carried out on a first frame of signal and a second frame of signal with the random noise sequence added through a constrained least squares learning method; whether the first frame of signal and the second frame of signal after curve fitting match each other is determined; if the first frame of signal and the second frame of signal after curve fitting match each other, the matching translation parameter is obtained through a genetic algorithm, the waveform after matching is obtained, the signal after matching is reconstructed, and the tail data redundant relative to the first frame of signal in the signal after matching is moved to the header data missing relative to the first frame of signal in the signal after matching. By using the waveform synchronization method and system for virtual instruments provided by the invention, when an ADC chip in a virtual instrument collects data of a periodic signal, the lateral translation jitter of the signal is small.

The invention relates to a high accuracy surface modeling method based on big data. The method comprises the following steps: creating geographic coordinate information and to-be-tested variable sampling values of all sampling points; turning a to-be-tested regional space into a grid point form through discretization, and establishing a sampling equation; conducting high order difference discretization on a partial differential equation set of a surface, obtaining a corresponding algebraic system, combining the algebraic system and the sampling equation to form an equality constrain least squares problem, converting the problem into a solve cross cutting object function minimum value problem, and then converting the minimum value problem into a solve symmetry indeterminate equation set; randomly selecting iteration initial values; blocking a coefficient matrix, and storing the block matrixes after decomposition; solving an HASM equation set through a block line projection iterative method, and determining whether a solving result is convergent; determining whether a solution meets demands of a Gauss and C.odazzi equation set; and outputting a high accuracy simulation surface model. Large-scale problems can be solved, the needed storage space is small, and defects of HASM for solving large-scale problems are prevented.

The invention discloses a high-spectral image demixing method based on an end member cluster. Adaptive accurate demixing of a high-spectral image is achieved, through extraction of the end member cluster and abundance inversion based on the end member cluster. The method includes the steps: (1) performing sparse representation based on a global image; (2) screening alternative end members based onvoting; (3) establishing the end member cluster by extracting spectral shape characteristics; (4) constructing a sub-block over-complete dictionary; (5) traversing the entire image, and selecting anoptimal end member cluster based on block sparsity according to each pixel; (6) performing abundance estimation by the aid of a fully-constrained least square method, and outputting an end member spectrum and an abundance result. According to spectrum variation in a high-spectral image, the method can effectively decrease errors caused by the spectrum variation, demixing accuracy is improved, andadaptivity of an algorithm is high based on extraction of the end member cluster of the image.

The invention provides a method for establishing a constrained least square maximum entropy quantile function model. The method comprises the steps of 1), establishing an unconstrained least square maximum entropy quantile function model; and 2), establishing a constrained maximum entropy quantile function model and method based on tail constraint of a Weibull distribution model the unconstrained least square maximum entropy quantile function model. The constrained least square maximum entropy quantile function model and method provided by the invention can be successfully applied to an aeronautic structure reliability analysis quantile function estimation problem under small sample condition. Compared with a traditional probability statistics method, the method has the advantages that high computing efficiency is achieved and complex random variables can be processed. Compared with a classic maximum entropy quantile function model, the method is high and stable in computing precision.

The invention provides a fully-constrained least square linear spectrum hybrid analysis method of a hyperspectral image. The method comprises the following steps of: 1) initially analyzing a mixing ratio; 2) screening an end member; 3) secondarily analyzing the mixing ratio; 4) inputting a mixed pixel to be analyzed, namely p; and 5) inputting d types of d end member arrays. The invention provides a new first come last served-linear spectral mixture analysis (FCLS-LSMA) analysis method in a single end member mode. The method has the advantages of high speed and optimal analysis effect theory.

The invention relates to a method for predicting the composition of mixed crude oil through near-infrared spectroscopy of crude oils. The method comprises: (1) respectively determining the near-infrared spectrums of single-crude-oils and mixed crude oil, carrying out second-order differential treatment, and calculating the polar coordinate projecting scores of the absorbance of each single-crude-oil and the mixed crude oil at two spectral regions such as W1 of 4000-5000 cm<-1> and W2 of 5000-6500 cm<-1>; and (2) combining the polar coordinate projecting scores of all the single-crude-oils to form a matrix X, combing the polar coordinate projecting scores of the mixed crude oil according to the same polar coordinate projecting score aligning order as the matrix X to form a vector y, substituting the y into the following formula y=bX, and calculating the vector b by using a non-negative constrained least-square method, wherein the vector B is recorded as b<inverted v>, and is the contents of each single-crude-oil contained in the mixed crude oil. According to the present invention, the method has advantages of simple calculation process, rapid predicting and strong practicality.

The invention relates to the technical field of piezoelectric ultrasonic transducer parameter measurement equipment, in particular to a piezoelectric ultrasonic transducer parameter measurement device and method. The output end of a low-ripple direct-current stabilized power supply is electrically connected with the input end of an LCD display screen and touch screen module circuit, the input end of a CPU, the input end of a DDS direct digital frequency synthesizer, the input end of a constant-current power amplification circuit, the input end of an analog switch circuit, the input end of a numerical control low-pass filter, the input end of a numerical control gain amplification circuit and the input end of a high-pass ADC circuit. The CPU is electrically connected with the LCD display screen and the touch screen module circuit through a bus. The device is reasonable in design, convenient to operate and small in overall machine body and has the advantages of being high in signal precision and calculation speed, the solved parameters of the electric ultrasonic transducer are more accurate and visual, small-step frequency sweep is adopted, admittance circle data are obtained through constrained least square fitting, the calculated parameter error of the piezoelectric ultrasonic transducer is smaller and more accurate, the frequency sweeping speed is higher, and the required measurement time is shorter.

The invention belongs to the technical field of digital image processing, and discloses a feature fine matching method and application based on a minimum spanning tree, and the method comprises the steps: converting an image into an HSL space, and carrying out the convolution of an L channel through a multi-scale Gaussian function; correcting the illumination component of the image to enhance a brightness non-uniform region in the image; equalizing brightness components among the three-line array CCD images after dodging; adopting surf feature matching for the preprocessed three-line array CCD dodging image to serve as rough matching feature points; calculating an affine transformation matrix by roughly matching the affine transformation relation of the feature point pair; generating an image minimum spanning tree according to the color in the image and the distance information, and taking the image minimum spanning tree as a weight to constrain a least square matching algorithm; and calculating a matching window correlation coefficient, continuously iterating the search window, and outputting by taking the center point coordinate as a matching point pair. The sparse matching provided by the invention has good registration precision, the average matching precision of the algorithm is 97.21%, and the average matching time is 6.3443 s.

The invention discloses a co-prime array partial polarization signal parameter estimation method based on zero interpolation, which is used for estimating a partial polarization signal arrival direction and polarization parameters under an underdetermined condition. The method comprises the following steps of arranging receiving ends according to a co-prime array; constructing a partial polarization signal receiving model and sampling; constructing array element non-repetitive virtual array output and a model noise covariance matrix thereof; initializing an interpolation output signal; initializing an iterative variable; solving a linear equation constraint least square; and estimating the direction of arrival and polarization parameters of the signal. Compared with the prior art, the method is advantaged in that a co-prime array is adopted to replace a nested array during underdetermined estimation of signal parameters, the aperture between array elements is increased, mutual coupling of the array elements is reduced, the signal direction of arrival and polarization parameters can be estimated at the same time, and during processing of co-prime array data, zero interpolation is utilized to interpolate virtual array holes of the co-prime array data; a discontinuous array is changed into a continuous uniform virtual linear array, and the number and precision of estimable signals are improved by using all array element information and degrees of freedom.

The invention provides an insulator surface charge inversion method based on a digital image processing technology, which comprises the steps of: dividing the surface of an insulator into a pluralityof grids; setting a unit analog charge at the geometric center of the surface, and calculating surface potential distribution formed by the analog charge; performing two-dimensional discrete Fourier transformation on the unit analog charge distribution to obtain a unit analog charge distribution matrix in a frequency domain, performing two-dimensional discrete Fourier transformation on the surfacepotential distribution to obtain a surface potential distribution matrix in a frequency domain, calculating a conversion matrix through the unit analog charge distribution in the frequency domain andthe surface potential distribution in the frequency domain, constructing a constrained least square filter, and obtaining an optimal solution of a filter coefficient by using an iterative algorithm;and obtaining an optimized constrained least square filter, calculating an estimated solution of the surface charge density in a frequency domain, and obtaining surface charge density distribution ina space domain through two-dimensional inverse Fourier transformation. The method is small in calculation amount and high in precision, and is helpful for promoting the research process of insulator surface charge distribution characteristics.

The invention discloses a Mode5 preamble pulse jitter value calculation method and system based on a constrained least square algorithm. The method specifically comprises the following steps: firstly, sampling an IFF signal in real time to obtain original sampling data, and synchronizing sliding delay; frequency domain signal detection and time domain parameter measurement are carried out on original sampling data to obtain pulse string information, and then a tree structure is used to judge whether a signal is a Mode5 signal. If the judgment is successful, carrying out accurate time delay on the original sampling data, and respectively carrying out FFT (Fast Fourier Transform) on the data after the synchronous sliding time delay and the accurate time delay to obtain a complex number result; and carrying out conjugate multiplication on the complex results to obtain a cross-correlation spectrum, finally calculating the arc tangent of the cross-correlation spectrum by using a CORDIC algorithm, carrying out phase ambiguity resolution processing on the arc tangent result in a system detection range, and carrying out linear fitting on the processing result by using an RLS algorithm to obtain a preamble pulse jitter value of the Mode5 signal. The calculated leading pulse jitter value is high in precision and good in robustness, the signal individual recognition technology can be supported, and the reconnaissance requirement of recognition friend or foe equipment at the present stage is met.