48 results about "Maximum entropy method" patented technology

The invention provides a multi-source information fusion based rainfall estimation method. The multi-source information fusion based rainfall estimation method includes: combining ground station observation rainfall data in a study area with a multi-source satellite rainfall data body in the study area to form a multi-source data set in the study area; establishing a dynamic Bayesian theory basedBayesian rainfall prediction model; using the maximum entropy method to obtain a nonlinear optimal solution of the Bayesian rainfall prediction model, and then determining the optimal weight and uncertainty information of each satellite data source; and generating an estimation result using the multi-source information fusion rainfall in the study area. The advantages of the invention are that theresult of multi-data source fusion analysis can reduce the uncertainty of the regional rainfall estimation due to the inaccuracy of a single type of rainfall information; and more reliable data inputand richer and more refined modeling data are provided for strengthening regional high-precision disaster warning, avoiding the flood risk or small watershed rainstorm flood estimation.

The invention discloses an infrared thermal imaging system-based fault recognition method. The method comprises the following steps of: 1, shooting electric equipment by adoption of an infrared thermal imaging system so as to obtain an infrared image of the electric equipment, and preprocessing the infrared image of the electric equipment; 2, carrying out threshold value segmentation on the preprocessed infrared image by adoption of a maximum entropy method; 3, carrying out edge detection on the infrared image after the threshold value segmentation by taking a Prewitt operator as an edge detection operator; 4, carrying out expansion processing on the infrared image after the edge detection by using an expansion algorithm; and 5, fusing the infrared image of the electric equipment after theexpansion processing with a visible image of original electric equipment so as to differentiate heating targets in the electric equipment from surrounding background. According to the method, an infrared temperature image processing technology is utilized to search temperature heating abnormal points and automatically judge operation condition and fault information of the electric equipment, so that the operation condition of the electric equipment is online and rapidly detected.

The invention relates to a poor information theory fusion-based product life characteristic information extraction method, which comprises the following steps of: acquiring original information of a small sample; transforming the original information into large-sample generating information by using a right self-service method and performing effective maximum likelihood processing, and acquiring maximum likelihood estimated values of large-sample content of two parameters, namely a Weibull distribution shape parameter and a scale parameter; extracting density functions of the two parameters by using a maximum entropy method; giving a confidence level, and calculating estimation intervals and expected values of the two parameters through the density functions of the shape parameter and the scale parameter respectively; and giving a failure probability, and acquiring the product life characteristic information through Weibull distribution life of the two parameters and reliability calculation thereof. The method has no requirement on completeness of the original information of the small sample, does not need priori information of the shape parameter and the scale parameter, can effectively recover total original characteristics of the product life, disclose nature of the product life information, more accurately acquire the product life characteristic information and reduce the experimental quantity of the product.

The invention relates to the technical field of image processing and provides an image enhancement algorithm based on a gauss hybrid model. According to the method, at first, luminance components of a color image are counted into a histogram, and mixture gauss modeling is carried out on the histogram; secondly, an improved EM algorithm is used for carrying out gauss hybrid model estimation on the histogram, a parameter of expectation maximization of a likelihood function is found out, and the optimum cluster quantity is determined through self-adaptation; thirdly, partition is carried out on the histogram according to an intersection point of adjacent clusters, and a plurality of sub-histograms are obtained; finally, the mapped clusters are found out according to the fact that area proportions of the sub-histograms with mapping relations are equal, the mapping function is adjusted in a micro mode according to application of the characteristic that the maximum entropy method tends to the human vision, and the final enhanced image is obtained. By the adoption of the image enhancement technology, the algorithm effectively improves the contrast ratio of the image, and increases the processing speed. The enhanced image obtained through the method achieves good effects in subjective visual perception aspect and objective evaluation aspect.

The invention discloses an improved Canny image edge detection method with noise immunity. On the basis of a Canny algorithm, the improved algorithm combines adaptive median filter to replace Gaussianfilter to denoise the image, so as to filter salt and pepper noise interference better. Then combined with the maximum inter-class variance method and the maximum entropy method, the double thresholdselection method is improved to get the high and low thresholds, so as to achieve the edge detection of the image, so that the target image retains the edge information as much as possible, while filtering out the unnecessary interference edges. The invention realizes the edge detection of an image under the condition of salt and pepper noise pollution and a double threshold selection method withstrong adaptability, and has high application reference value.

In a measurement method for measuring the epitaxial film thickness of a multilayer epitaxial wafer, a reflectivity spectrum of a multilayer epitaxial wafer having at least two epitaxial layers of different electric characteristics is measured by using infrared radiation in a far infrared region of at least 500 cm-1 or less, and frequency-analysis is performed on the reflection spectrum thus obtained by a maximum entropy method, and the film thickness of each epitaxial layer is calculated on the basis of the analysis spectrum thus obtained.

The invention provides an improved adaptive Gaussian mixture foreground detection method. The method comprises: firstly, performing learning by utilizing a Gaussian mixture model to form an initialized Gaussian mixture background model; secondly, for a new input video sequence, performing sampling at an interval of N frames, obtaining an image frame by utilizing weighted time-domain mean filtering, and performing background model updating by taking the image frame as an input of Gaussian mixture modeling; automatically determining whether background mutation exists in a current frame by Poisson distribution, if the background mutation does not exist, keeping normal sampling interval and learning rate, otherwise, reducing an interval frame number and increasing the learning rate, updating the background model, and extracting a current background frame; and finally, performing difference by utilizing the current frame and the current background frame, obtaining an adaptive threshold with a maximum entropy method, performing weighted mean on the obtained threshold, and performing foreground detection. According to the method, motion interferences of tree leaf shake, water ripples and the like in a video scene are effectively overcome, the calculation amount of frames is reduced through periodic sampling, and the timeliness is improved.

The invention relates to a planeness assessment method for reducing the number of measuring points, and has the following specific steps: the plane error values at a plurality of point positions on a plane are measured according to the existing planeness measuring method; the probability distribution of the measured plane errors is estimated by the maximum entropy method according to the limited measuring point values; according to the estimated probability distribution, the estimated value of the plane errors between the measuring points is generated at other positions of the measured plane, that is, the measuring points are interpolated; the planeness error assessment methods such as least envelope zone method, least square method or diagonal method are adopted to assesses the planeness error based on the measured value of the planeness error and the estimated value of the plane error at interpolation position; the planeness errors are respectively calculated according to the probability, statistical theory and a plurality of estimated values at interpolation position, and the mean value of the calculated results is taken as the final planeness error assessment value. On the basis of maximum entropy method, the invention uses limited measuring data sample to determine the probability distribution of the measured plane error.

The invention discloses a reliability verification test method based on mixed Bayesian prior distribution. According to the method, the prior distribution of unknown parameters is determined by adopting a conjugate prior distribution method; two different sets of parameters are obtained through a prior moment method and a maximal entropy method, namely different prior distributions are obtained; the weight of the two prior distributions is determined according to the second type of maximum likelihood method, parameters obtained through the prior moment method and the maximal entropy method are fused according to the weights, further, the prior distribution obtained finally is more accurate than prior distribution obtained through the separate adoption of anyone of the prior moment method and the maximal entropy method and better fits to the real distribution. In the method, the minimum number of zero-failure use cases required in the reliability verification test is calculated according to the Bayesian prior distribution and test information; therefore, compared with a method not based on prior knowledge, the method can be used for effectively reducing the quantity of use cases used in the test.

A time series data analyzer includes a segment condition input section, an analysis condition input section, and an optimum condition deriving section for analyzing all segments based on the segment conditions and analysis conditions inputted in the respective input sections, under all analysis conditions by a maximum entropy method and a nonlinear least squares method. The time series data analyzer derives the optimum segment length and the optimum lag value in correspondence to selected results, and an analysis execution section executes analysis by the maximum entropy method by setting the optimum analysis conditions derived as described above. The trending of the spectrum of electroencephalogram data is used as an indicator of the state of the subject based on the findings that the spectrum of electroencephalogram data is an exponential spectrum and the gradient changes depending on the state of the subject.

The invention provides a nuclear magnetic resonance echo data inversion method based on two-parameter regularization and a nuclear magnetic resonance echo data inversion device thereof. The nuclear magnetic resonance echo data inversion method comprises the steps that nuclear magnetic resonance echo data are acquired; an inversion method target function is constructed according to a target function corresponding to a Tikhonov method and a target function corresponding to a maximum entropy method; the optimal regularization parameters of the inversion method target function are acquired according to the optimal regularization parameters of the target function corresponding to the Tikhonov method and the optimal regularization parameters of the target function corresponding to the maximum entropy method; and the nuclear magnetic resonance echo data and the optimal regularization parameters of the inversion method target function are substituted in the inversion method target function for solution so that the inversion result of the nuclear magnetic resonance echo data is obtained. According to the nuclear magnetic resonance echo data inversion method based on two-parameter regularization and the nuclear magnetic resonance echo data inversion device thereof, the high-precision inversion result of the nuclear magnetic resonance echo data can be obtained.

The invention discloses a night monitoring and identification method based on neural network enhancement. The night monitoring and identification method includes the steps: firstly, obtaining neighborhood information of pixels in a night monitoring image; carrying out adaptive enhancement on the image based on the quadratic Taylor series; obtaining adaptive enhanced images, extracting regional features and edge features from the self-adaptive enhanced image and respectively inputting the regional features and the edge features into corresponding neural networks; and performing significance calculation on a feature recognition result output by the neural network, fusing the calculated significance region images to obtain a comprehensive significance image, finally segmenting the comprehensive significance image by using a maximum entropy method to obtain a binary image, and extracting a target image from the adaptive enhanced image based on the binary image. According to the night monitoring and identification method, the contrast of the image can be improved, and self-adaptive image enhancement is performed for the problem of uneven illumination of each part of the image, and the generated saliency map can effectively cover the boundary of the target area and well inhibit the saliency of the background area.

The invention discloses an analysis system of emotional tendency of a text. The analysis system comprises a sample training module, an entity extracting module, a characteristic extracting module and an emotional tendency recognizing module. The sample training module is used for receiving texts to by analyzed and training the sample to acquire a discrimination template. The entity extracting module is used for extracting the entities of texts to be discriminated and filtering texts without entities. The characteristic extracting module is used for extracting tendency relative characteristics of the texts. The emotional tendency recognizing module is used for discriminating the tendency of the texts according to maximum entropy method. According to the arrangement, the forums and blogs belongs to the field of the enterprise user are collected, the texts of the webs are extracted, the emotional tendency of the texts and aimed entities are acquired through the analysis system, a chart showing image changes of the enterprise and the competitor is automatically generated, and accordingly the emotional tendency of the text is accurately judged through text classification.

The present invention relates to the field of traffic, especially to a maximum entropy method used for traffic subnetwork trip matrix estimation. The method comprises the following steps of: S1: selecting and establishing an abstracted sub traffic network, wherein the network is formed by a node set N and a road section set A, and the N comprises a starting point set R and a terminal point set S;S2: establishing and solving the maximum entropy model of a traffic subnetwork trip matrix; S3, in the abstracted sub traffic network, employing the maximum entropy model, performing initialization toobtain a feasible solution of the maximum entropy model, designing an algorithm to find and solve a current solution decreasing direction of decreasing of a target function value of the maximum entropy model; S4: performing linear search, performing solution, determining an optimal [Alpha], and determining the optimal step of the decreasing; S5: updating the feasible solution; and S6, allowing the algorithm to end the examination. The maximum entropy method used for the traffic subnetwork trip matrix estimation takes easily obtained flow of each road section of the whole network as unique input of the model to establish the maximum entropy problem so as to improve the algorithm efficiency, allow the method to be utilized in a large network and improve the prediction precision. The maximumentropy method used for the traffic subnetwork trip matrix estimation can be used for assessment of influences of different network changes on the subnetwork flow.

The present invention relates to a pork intramuscular fat content nondestructive testing method based on computer vision. The method comprises the steps of adopting a camera calibration method to calibrate a CCD digital camera, selecting the pork musculi oculi just now purchased from a slaughter house as an experiment sample, shooting and sampling the cross-section of the pork musculi oculi, carrying out the pre-processing on an obtained sample image via an improved sample block repair method, and recovering the original information of a light reflection area of the image; combining a maximum entropy method and an iteration method to carry out the image segmentation on the pre-processed musculi oculi image, and extracting the pork marbling; extracting 291 characteristic values, such as a fat number index, a fat distribution index, a fat texture index, etc., from an obtained marbling image, establishing a pork intramuscular fat content prediction model according to the characteristic values and a chemical method detection result, and finally predicting the pork intramuscular fat content via the model. By utilizing the method of the present invention, the pork intramuscular fat content can be predicted very well, and the pork nutrition detection is more objective, more accurate and more efficient.

There is provided a computer calculation processing method for determining risk-neutral probability distribution with the use of a model-independent maximum entropy method. There are provided an input device (101) into which arbitrarily set information is inputted, a market (102) which is the source of various market information, a calculator (103) which receives various inputs, performs the calculation processing of the present invention and issues an output instruction, an interface (104) which collects information from the market and inputs it to the calculator (103), a storage device (105) which stores the program of the present invention and various data generated during the calculation processing, and a display device (10). Output of a result of the calculation processing can be stored in the storage device (105) in addition to the display device (106), printed on a printing device or transmitted via a communication line.

The invention provides a technology for maximum power point tracking and diagnosis of a photovoltaic power generation system. A maximum power point tracking and diagnosis system for the photovoltaic power generation system comprises a light ray collecting module, an image processing module, a maximum power point tracking (MPPT) module and a signal conversion module. The technology comprises the steps of first collecting a light ray intensity image, wherein a collecting system is mainly structurally formed by a light intensity sensor and a temperature sensor; performing light ray image analysis, using a partitioning algorithm to divide the light ray image into several parts which are obviously different in gray values, selecting the partitioning algorithm with an Otsu method and a maximum entropy method as the core, and performing precise image display on a display screen by segmenting out a highest light intensity part. The technology has a very good power tracking effect and can improve working efficiency.

The invention relates to a well distribution method based on offshore oilfield reservoir subdivision. The method includes the steps of 1, in connection with existing oil formation division in a fluvial deposition reservoir, subjecting a logging curve of an actual drilled oilfield to maximum entropy method calculation, and preliminarily determining depositional cycles according to changes in knee points of an INPEFA curve; 2, in each high-level deposition cycle, determining low-level deposition cycles, and calculating channel sand thickness to a sand thickness to formation thickness ratio in each low-level deposition cycle; 3, in a target oilfield reservoir, dividing the low-level deposition cycle having 3-10 m of channel sand thickness and greater than 40% of sand thickness to formation thickness ratio, into a first development layer series, dividing the low-level deposition cycle having 3-8 m of channel sand thickness and 30%-40% of sand thickness to formation thickness ratio, into a second development layer series, and dividing the low-level deposition cycle having 1-4 m of channel sand thickness and less than 30% of sand thickness to formation thickness ratio, into a third development layer series; 4, using horizontal wells at positions satisfactory to the first development layer series of step 3, using the combination of horizontal wells and directional wells at positions satisfactory to the second development layer series, and using directional wells at positions satisfactory to the third development layer series.

The invention provides a method for constructing nonlinear motion model of CGCS2000 frame sites. The method comprises the steps of adopting the time series gross error detection and time series missing value interpolation methods based on singular spectrum analysis to preprocess day analysis time series to obtain the clean and complete time series, conducting singular spectrum decomposition on the preprocessed time series after, obtaining the principal component weight of the time series of the sites, combining the maximum entropy method and the nonparametric test and oscillation pair identifying method to separate signals and noise of the time series, extracting the useful signals, conducting dynamic feature analysis on the sites, adopting the time series prediction method based on the singular spectrum analysis method, conducting site position predication on the extracted useful signals, carrying out modeling based on the predication result, and obtaining the nonlinear motion model of the sites. According to the method, the precision of the constructed nonlinear motion model is obtained from multiple aspects, and the high-precision maintenance and updating can be dynamically carried out on the CGCS2000 frame for a long time.

The invention discloses a maximum entropy principle-based shallow sea water flow field acoustic tomography method. According to the method, transmitting-receiving nodes are arranged in a sea area to be tested, so that measured acoustic signals can be obtained; the transmission delay difference of the same path in an opposite direction is obtained; prior flow field information is utilized to construct the relationship of the measured delay difference and flow field distribution, and an inversion mathematical model is constructed; the inversion mathematical model is solved by using the maximum entropy principle; and a convergence solution is obtained by using an iterative method. As indicated by simulation instances, the maximum entropy principle-based shallow sea water flow field acoustic tomography method can be utilized to construct two-dimensional flow field information of the target sea area by using the transmission delay difference and achieve higher least-square accuracy.

The invention relates to a surface defect detection method of robust based on machine vision. The method comprises the following steps: 1) performing gray-scale treatment on the input colorful images to obtain a grey-scale image; 2) when the damage gray value in the grey-scale image is higher than that of the background, inverting the color of the gray-scale image; 3) based on the provided binarizition threshold optimization function, calculating the optimal threshold value; 4) based on the optimal threshold value, performing binaryzation on the image; and 5) performing contour detection on the image after binarizition. According to the provided method for calculating the optimal threshold value by the binarizition threshold optimization function, compared with the traditional binaryzation operation such as an iterative threshold method, an otsu method, a bimodal average method and an one-dimensional maximum entropy method, the method has strong robustness, damage position can be accurately extracted from the image containing the surface defect, and the proportion of the non damaged image with false detection as damaged is greatly reduced. The parameter can be conveniently adjusted, and the realization process is simple.

The present invention provides a method of selectivity estimation in which preprocessing steps improve the feasibility and efficiency of the estimation. The preprocessing steps are partitioning (to make iterative scaling estimation terminate in a reasonable time for even large sets of predicates), forced partitioning (to enable partitioning in case there are no “natural” partitions, by finding the subsets of predicates to create partitions that least impact the overall solution); inconsistency resolution (in order to ensure that there always is a correct and feasible solution), and implied zero elimination (to ensure convergence of the iterative scaling computation under all circumstances). All of these preprocessing steps make a maximum entropy method of selectivity estimation produce a correct cardinality model, for any kind of query with conjuncts of predicates. In addition, the preprocessing steps can also be used in conjunction with prior art methods for building a cardinality model.

The invention relates to an improved natural language characteristic precise extracting method based on deep learning. The method includes the steps of establishing a natural language condition maximum entropy model by means of a maximum entropy method when basically analyzing natural language, selecting natural language property characteristics by means of an IFS algorithm on the basis of the model, selecting characteristics meeting reality by matching the natural language property characteristics, and precisely extracting the natural language characteristics by means of a deep learning method. When natural language characteristics are extracted by means of the improved extracting method, compared with a traditional extracting method, the improved extracting method has the advantages that the extracting accuracy is improved, the error rate is decreased, and certain practicability is achieved.

By combining a univariate dimensionality reduction method (UDRM) and a maximum entropy method (MEM), the invention provides an efficient random uncertainty analysis method for processing multi-peak distribution random variables. According to the method, an MEM is expanded from a fourth-order moment constraint to an n-order moment constraint, then the order of a response statistical moment (or themoment constraint of the MEM) during response distribution convergence is determined through UDRM + MRM circulation, and on the basis, response probability distribution and response point probabilities are obtained through a UDRM + MEM method. According to the method, multi-peak distribution random variables and multi-peak distribution responses can be processed at the same time, so that the problem that the Jacobian matrix G is close to singularity or illness is well solved, and the solvability of an equation set is improved; Meanwhile, on the premise that the result accuracy is guaranteed, the requirement for the sample scale can be greatly reduced, and high calculation precision can be obtained only through a small number of samples.

The invention discloses a power grid probabilistic load flow analysis method based on generalized semi-invariants and a maximum entropy method. The power grid probabilistic load flow analysis method comprises the following steps: S1, establishing a power system load flow model considering frequency; S2, establishing a load random output model based on a normal distribution function; S3, establishing a new energy random output model by using a Gaussian mixture model; S4, calculating the probabilistic load flow of the power system based on a generalized semi-invariant method; and S5, fitting theprobability density curve of the output variable by adopting a maximum entropy method. According to the method, the problem of frequency and voltage fluctuation of the power system under the influence of input variable randomness can be effectively solved; the method has the advantages of accuracy, practicability and high efficiency; therefore, the out-of-limit risk of accessing large-scale new energy to the power system can be comprehensively evaluated, weak links of the power system are discovered, and further consumption of the new energy is promoted.

The invention discloses a moving vehicle tracking method for resisting strong shadow interference based on a non-subsampled shear wave domain zero tree structure, which is high in accuracy and robustness and has self-adaptive capability, and comprises the following steps of: after a video frame is converted into an HSV color space from an RGB color space, carrying out non-subsampled shear wave transformation; assuming that the transformation coefficient obeys Gaussian distribution, calculating a weighted mask of each scale by adopting a mean value and a standard deviation of the transformationcoefficient; according to the zero tree distribution characteristic of the multi-scale transformation coefficient, correcting a fine-scale weighted mask by utilizing a coarse-scale weighted mask, andcarrying out linear combination on the weighted masks of each scale and each color channel to obtain a public mask; calculating a self-adaptive segmentation threshold value by using a maximum entropymethod based on least square fitting, and binarizing the public mask; and determining a moving vehicle area in a voting mode, and tracking the target vehicle by adopting a mean shift algorithm.

The invention relates to a tunnel monitoring measurement data analysis method based on a maximum entropy method reliability theory. The method mainly comprises the following steps of acquiring point cloud data of a primary support of a tunnel in a tunnel construction process; computing a distance between corresponding points between point cloud faces adjacent in time in order to determine a displacement of the support at the same position of the tunnel in a specific duration, and obtaining a displacement matrix generated by the deformation of the primary support of the tunnel; and computing based on the displacement matrix by utilizing a maximum entropy method in order to obtain a probability density function of the deformation of the primary support of the tunnel, determining a tunnel monitoring measurement index, drawing a convergence time curve according to the measurement index in order to determine whether the deformation of the tunnel is converged, and judging whether the deformation of the primary support of the tunnel meets the requirements for construction at each stage. According to the method, the analysis of the deformation of the primary support of the tunnel construction can be realized, an analysis result error is small, and the actual deformation degree of the support can be accurately reflected.

The invention relates to a wind electricity maximum installed capacity prediction method which uses a maximum entropy principle. The wind electricity maximum installed capacity prediction method which uses the maximum entropy principle mainly achieves the purposes of achieving rapid and exact prediction for the wind electricity maximum installed capacity, and laying a foundation for reasonable planning of a power grid. The technical scheme is that the wind electricity maximum installed capacity prediction method which uses the maximum entropy principle includes steps: 1) building a chance constraint planning model of the wind electricity maximum installed capacity; 2) using a pattern search method combined with a maximum entropy method to solve the chance constraint planning model, and then obtaining prediction value of the wind electricity maximum installed capacity of a system. The electricity maximum installed capacity prediction method which uses the maximum entropy principle is mainly used in the wind electricity technology field.

The invention relates to a Weibull type product reliability acceptance scheme design method based on expert information. The method comprises the following steps: acquiring expert information of a Weibull type product to be accepted, determining a nonlinear programming model through a maximum entropy method, and determining distribution parameters; determining a parameter distribution formula of the Weibull-type product according to the distribution parameters, and obtaining the probability that the service life of the product is greater than an inspection upper limit, the probability that the service life of the product is less than an inspection lower limit and the probability that a fault product occurs for a given time in a timing truncation acceptance test through a sampling-based algorithm; and determining a producer risk and a user risk in combination with a parameter distribution formula, and determining an acceptance scheme package of the product according to the producer risk and the user risk. Expert information is effectively utilized, two kinds of risk values can be remarkably reduced when the number of invested test samples is small, and the test time is shortened.

The invention discloses a method for detecting corrosion of submarine pipelines. The method comprises the following steps of: acquiring alternating-current impedance, acquiring detection frequency and determining pipeline conditions. The detailed steps are as follows: loading alternating-current disturbance signals to one end of a pipeline, receiving alternating-current response signal generated according to the corrosion state of the pipeline at the other end of the pipeline, and acquiring alternating-current impedance according to the alternating-current disturbance signals and the alternating-current response signals; changing the frequency of the alternating-current disturbance signals loaded on the pipeline, acquiring the alternating-current impedance of the pipeline under different frequencies and selecting maximum value and minimum value and corresponding maximum detection frequency and minimum detection frequency; and respectively loading current signals with maximum detection frequency and minimum detection frequency to one end of the pipeline, receiving the current signals at the other end of the pipeline, extracting electrochemical noise from the current signals and carrying out fourier transformation and maximum-entropy-method transformation on the electrochemical noise to obtain characteristic parameters of a power density spectrum.