45 results about "Discriminant function analysis" patented technology

The invention relates to a discriminant analysis-based high road real-time traffic accident risk forecasting method. The method comprises the following steps of: building a high road accident risk discrimination model for a detection area; substituting real-time traffic flow characteristic parameters into the high road accident risk discrimination model; and judging whether the risk of traffic accident exists or not. According to the method, traffic accidents can be forecasted in real time by using the real-time traffic flow characteristic parameters acquired by high road traffic detection equipment, the method has relatively high forecasting precision, and technical defects and shortages in the prior art for analyzing traffic safety by using aggregated statistics are overcome. The methodhas practical engineering application value in the aspects of discrimination of the risk of the high road traffic accidents and forecast of the traffic accidents.

The welding quality classification apparatus relating to the present invention is an apparatus, wherein a data point indicating feature information of a welded joint to be classified whose welding quality is unknown is mapped to a point in a mapping space which has a dimensional number higher than the number of the features constituting the feature information, and the welding quality of a welded joint to be classified is classified based on which of regions of two welding qualities, which are formed by separating the mapping space with a decision boundary, contains the mapped point, and wherein a discriminant function is determined by adopting a weight which minimizes the sum of the classification error corresponding to classification accuracy of a training dataset and a regularization term having a positive correlation with the dimensional number of the discriminant function as weight for each feature constituting the discriminant function indicating the decision boundary.

A structure reliability dynamic response surface method based on discriminant analysis, comprising the steps as follows: determining a random variable; sampling via using Markov chain Monte Carlo method, determining an initial training sample point, calculating the function value of the initial training sample point, and determining status value thereof; establishing a training sample set and training a classification response surface to obtain a trained classification response surface; randomly sampling N sample points and estimating the status value, and then calculating failure probability; judging whether the failure probability meets condition of convergence, and stopping if the failure probability meets condition of convergence, otherwise, finding the failed sampling point, finding and calculating the function value of the most probable failure point and determining the status value; adding to the training sample set by using the most probable failure point and status value thereof as a new sample, and repeating the following steps until the condition of convergence is met. The method of the invention is simple in theory and efficient in calculation, which provides an efficient path for analyzing the reliability and precision of the complex structure that the single calculation thereof is time-consuming.

The invention discloses a knock response detecting device for tiny flaws on poultry eggs and a method thereof. The detecting device mainly comprises a detecting platform, a data acquisition device, an A / D converter and a computer, which are connected in turn. The method comprises the following steps: fixing a poultry egg on a platform base, and adopting a detection method of single-point excitation and three-point response; after the poultry egg is excited, triggering a response signal by a sensor signal, and collecting the response signal generated by a sensor through the data acquisition device; after conditioning and converting the signals through the A / D converter, carrying out fast Fourier transform to obtain a response frequency curve between 0 and 7,500Hz; after the response frequency is subjected to normalization treatment, taking highest amplitude to extract characteristic value and frequency characteristic value extracted by high-low amplitude in turn, and using principal component analysis and linear discriminant function analysis to carry out pattern recognition respectively. The method is not limited by environmental noise, eliminates interference of subjective factors during manual operation, and can accurately distinguish out the tiny flaws on the poultry eggs.

The invention discloses a method for discriminating reservoir fluid by establishing a gas logging chart on the basis of discriminant analysis, relating to the technical field of oil and gas exploitation and development. The method includes the steps of: a, collecting the gas logging peak value data and the oil test results in an oil test completion well; b, according to the categories of the oil test results, grouping the gas logging peak value data of the logging display sections which have the same category of oil test results; c, establishing discrimination factors by taking the gas logging peak value data of the logging display sections as a base; d, selecting the established discrimination factors, and formulating discrimination functions which can effectively discriminate the property of the fluid on the basis of discriminant analysis; and e, according to the coincidence rates of the formulated discrimination functions, preferentially selecting two discrimination functions with the highest coincidence rates to be used as x axis and y axis to establish a gas logging chart to discriminate the property of the reservoir fluid. The method disclosed by the invention solves the problem that conventional gas logging charts are difficult for effectively discriminating the reservoir fluid, so the dominant regions of different fluids can be well marked out.

A decision algorithm of discriminating a defect type and / or a defect level is automatically produced in a short time. The accuracy of defect classification in terms of a defect type or a defect level is improved. More specifically, feature values of surface defects acquired in the past are represented as points in a feature space, constraint is set that when the feature space is mapped into a space by an assumed mapping function, a discriminant plane can be defined in the mapped space such that mapped feature values are linearly separable perfectly with respect to the discriminant plane, an objective function is defined as the distance between the discriminant plane and a point located closest to the discriminant plane, an optimal discriminant plane is determined by solving a quadratic programming problem such that the objective function is maximized under the constraint, the determined discriminant plane is employed as a discriminant function, a feature value of a surface defect whose type is unknown is substituted into the discriminant function, and it is determined, from the resultant value of the discriminant function, whether the surface defect whose type is unknown is of the particular surface type.

What is disclosed is a non-contact system and method for determining cardiac function parameters from a vascular pattern identified from RGB and IR video signals captured simultaneously of a region of exposed skin of a subject of interest. In one embodiment, a video of a region of exposed skin is captured using a video camera that captures color values for pixels over visible channels and an IR camera that measures pixel intensity values in wavelength ranges of interest. Pixel intensity values are processed to generate a vascular binary mask that indicates pixel locations corresponding to the vascular pathways. The IR images are registered with corresponding data from the camera's visible channels such that pixels that correspond to the vascular pattern can be isolated in each frame of the video of visible color data. Once processed, pixels associated with the isolated vascular patterns are analyzed to determine desired cardiac function parameters.

The invention relates to a deep recurrent neural network-based cardiac function automatic analysis method and belongs to the technical field of medical image analysis. The method includes the following steps that: S1, a cardiac nuclear magnetic resonance film is acquired, and the cardiac nuclear magnetic resonance film is pre-processed; S2, a recurrent neural network model of multi-task learning is constructed, and underlying general image features are extracted; S3, the extracted underlying general image features are inputted into the two-layer long- and short-memory recurrent neural network,space-time dependence relations are constructed; S4, a target loss function is constructed; S5, and parameters in the recurrent neural network are trained and optimized through a stochastic gradientdescent method according to the loss function constructed in step the S4; and S6, after the training of the recurrent neural network model is completed, the pre-processed cardiac nuclear magnetic resonance film is inputted into the trained recurrent neural network, and thirteen parameters in cardiac function analysis are measured. With the method of the invention adopted, the manual delineation ofventricular structures is not required, and end-to-end cardiac function analysis can be automatically performed.

If a threshold discrimination is performed with variable Z=0 using discriminant analysis, that is useless unless know-how is accumulated through visual judgment and actual operation. A discriminant function is computed using a plurality of parameters which make pass / fail judgment factors and the results of that pass / fail judgment. With respect to the discriminant function, a histogram is generated for pass category and for fail category. Then, a threshold is determined based on the standard deviation in the individual categories so that an intended rate of flowout and rate of overcontrol will be obtained. The acceptability of pass / fail judgment objects is judged based on the threshold. Thus, the rate of flowout and the rate of overcontrol can be controlled as intended. Further, high-performance pass / fail judgment can be implemented without accumulating know-how.

The invention discloses a ship radiation signal recognition method based on multi-kernel learning and discriminant analysis. According to the method, pretreatment, auditory sense model feature extraction, dimensionality reduction and classifier classification and judgment are sequentially conducted on a ship radiation signal sample, wherein in the stage of dimensionality reduction, a method based on multi-kernel learning and discriminant analysis is adopted, alternate optimization is utilized, and optimization operation is conducted on kernel mapping coefficients and linear multi-kernel combination coefficients respectively under the goal of kernel discriminant analysis optimization represented in a graph embedding mode. Compared with the prior art, on the aspect of ship radiation signal recognition, the recognition performance of a system can be improved effectively.

The invention discloses a device for detecting cracks on a poultry egg shell by using a volatile matter and a method thereof. The detection device mainly comprises a computer, an activated carbon filter, a first built-in pump, a gas tank, an air valve, a material container, a switch, a gas sensor array, a second built-in pump, a data acquisition unit and an A / D converter. The data acquisition unit acquires response signals of the gas sensor array; the signals are transmitted to the computer after being conditioned and converted through the A / D converter so as to extract a representative characteristic value in a curve of each sensor; by taking the extracted characteristic value as an input value, a visualized distinguishing is performed by using a principal component analysis and a linear discriminant function; and a pattern recognition is performed by using a classical BP algorithm and a BP neural network based on a genetic algorithm. The method excludes the interference caused by subjective factors of manual operation and overcomes the shortcoming of destructive detections, thereby achieving the non-destructive detection on fine cracks on the poultry eggs nondestructively and accurately.

The invention discloses a hyperspectral target detection method based on combination of sparse expression and discriminant analysis. The method comprises the steps as follows: (1) a to-be-detected picture is pre-detected; (2) a background sample set and an initial dictionary are constructed according to a pre-detection result; (3) the background sample set is purified by the aid of reconstruction errors and discrimination information, target samples are rejected, and the purified dictionary is constructed; (4) a sparse coefficient is obtained with a Lasso method on the basis of the purified dictionary with the discrimination information; (5) the experimental result is counted and the target detection precision of a hyperspectral image is calculated. Compared with an existing method, the hyperspectral target detection method has the advantages that only background samples are used during construction of the dictionary, and the problem of imbalance of target samples and background samples for dictionary learning due to few target samples is solved; during construction of the dictionary and formulation of judgment rules, distance-based discrimination information is used. With the adoption of the method, the discrimination information is fully fused into the dictionary learning method, and the detection efficiency is improved.

The invention discloses a method for identifying a shale gas reservoir while drilling by utilizing a discriminant analysis method. The method comprises the following steps that (a) drilled shale gas horizontal well data in the same block are integrated and classified; (b) data analysis is performed on the classified region data, effective shale gas reservoir discriminant parameters are preferably selected, and discriminant factors are established; (c) a discriminant function which can effectively identify the shale gas reservoir is established by utilizing the disriminant factors; (d) according to the established discriminant function, interpretation while drilling is performed on the shale gas reservoir. According to well logging interpretation and oil production testing data of a drilled shale gas horizontal well, the discriminant factors established through optimized while-drilling parameters are utilized, and the discriminant function which can effectively identify the shale gas reservoir is established through discriminant analysis. In this way, while-drilling identification on the shale gas reservoir is achieved, and the problem that a while-drilling interpretation method of a shale gas reservoir has defects is solved.

The invention provides a regular expression compressing method for DFA (Discriminant Function Analysis), comprising the following steps of: compressing each state of each row or column of regular expressions in a way similar to run length coding; and if states with unique state skip condition exist in each row or column of coded regular expressions, forming a hash table by using the states. The invention has the advantages of greatly reducing the storage space required by DFA storage and storing more regular expressions within a limited space.

The invention discloses an image identification method for two-dimensional probability linear-discriminant analysis based on L1 norm. The method is used for handling the situation when dimensionalityof two-dimensional data is reduced and there are exceptional values in images, and denoted as L1-2DPLDA. The method specifically includes the steps of creating an L1-norm model on original image data;utilizing an EM algorithm to figure out a solution to the model to obtain a projection matrix; utilizing the projection matrix to classify unknown images. Compared with traditional vector data, the method is different in that dimensionality can be reduced in the line and column directions of the two-dimension data, the spatial features of the images can be maintained, and meanwhile, under the situation when there are exceptional values in the original data, the method has robustness, and an image identification rate with higher accuracy is obtained.

The invention discloses an image interestingness dichotomy prediction method combining discriminant analysis and multi-kernel learning, and the method comprises the steps: inputting image data, and forming a data set; inputting the data set in the step 1, and determining three clues including an unusual clue, an aesthetic clue and a general preference clue in the data set; carrying out any featurefusion by adopting discrimination correlation analysis or multiple discrimination correlation analysis; and a simple multi-kernel learning algorithm is adopted for classification. According to the image interestingness dichotomy prediction method, compact expression of different interestingness characteristics in each clue and interestingness multi-source heterogeneous characteristics of expression between the clues are considered, a compact and discriminative interestingness characteristic set is formed, and simultaneous characterization and modeling of multi-source interestingness information are realized.

The invention relates to a product image rapid generation method based on a series confrontation network. According to the method, texture structure reasoning and color reasoning are realized by usingtwo models respectively; And the whole network structure is formed through series connection, and product image generation is completed through a discriminant function. According to the network structure, the complexity of the problem is decomposed, the training difficulty is reduced, and the generation effect is improved.

The invention discloses a three dimensional fragment category detection method based on histogram feature kernel optimized discriminant analysis, relating to a three dimensional fragment category detection method. The invention solves the problem of inaccurate detection existing in the present three dimensional fragment category detection methods. The category detection method comprises the following steps of: scanning a fragment to be detected to obtain three dimensional surface data of the fragment; carrying out feature extraction on the three dimensional surface data of the fragment, obtained in the step 1 to obtain a three dimensional surface feature vector of the fragment; carrying out kernel optimized discriminant analysis on the three dimensional surface feature vector of the fragment, obtained in the step 2 to obtain a feature vector of kernel optimized discriminant analysis; and finally utilizing a nearest neighbour classification to carry out category detection on the feature vector of kernel optimized discriminant analysis, obtained in the step 3 to obtain the category of the fragment. The invention overcomes the insufficiencies of the prior art, can accurately detect the category of the three dimension fragment and can be applied to the technical field of category detection, classification and the like of the three dimensional fragment.

The invention is applicable to the field of improvement on a computer vision technology, and provides a face recognition method based on joint sparse discriminant analysis. The method comprises steps: A, a database for storing face training images with known identities is built; B, detection on a face with a to-be-detected identity and data acquisition are carried out, and a face database with authentication is built; C, feature extraction is carried out on to-be-authenticated face data and training data, and a discriminant analysis model based on joint sparse representation is built; and D, by using the training data after feature extraction in the step C and the to-be-authenticated data and in combination of corresponding label information, a nearest neighbor classifier is used for discriminant analysis, and a face recognition result is calculated. By adopting the joint sparse discriminant analysis to recognize the to-be-authenticated face image, the authentication accuracy is improved, and the recognition robustness in a condition with multiple changes of the to-be-authenticated face in aspects such as light, angle, expression, disguise and gesture can be effectively improved.

The invention discloses a method for discriminating reservoir fluid by establishing a gas logging chart on the basis of discriminant analysis, relating to the technical field of oil and gas exploitation and development. The method includes the steps of: a, collecting the gas logging peak value data and the oil test results in an oil test completion well; b, according to the categories of the oil test results, grouping the gas logging peak value data of the logging display sections which have the same category of oil test results; c, establishing discrimination factors by taking the gas logging peak value data of the logging display sections as a base; d, selecting the established discrimination factors, and formulating discrimination functions which can effectively discriminate the property of the fluid on the basis of discriminant analysis; and e, according to the coincidence rates of the formulated discrimination functions, preferentially selecting two discrimination functions with the highest coincidence rates to be used as x axis and y axis to establish a gas logging chart to discriminate the property of the reservoir fluid. The method disclosed by the invention solves the problem that conventional gas logging charts are difficult for effectively discriminating the reservoir fluid, so the dominant regions of different fluids can be well marked out.

The present invention provides a polymerization reactor fault diagnosis method based on combination of DKPCA (dynamic kernel principal component analysis) and FDA (Fisher's discriminant analysis), and relates to a polymerization reactor fault diagnosis method. The method aims at features in a polyvinyl chloride (PVC) polymerization production process that the number of fault types is relatively large, the types are complex, and so on, and provides a polymerization reactor fault diagnosis method based on combination of DKPCA and FDA. Fault diagnosis is performed on a PVC polymerization process by using the kernel principal component analysis and the dynamic kernel principal component analysis algorithm separately, and meanwhile, further separation of the fault data is performed by using an FDA method. A simulation research result shows that the kernel principal component analysis has the relatively good fault diagnosis accuracy on the PVC polymerization process, and the FDA can further realize fault separation, so that faults can be monitored in an actual PVC polymerization production process.

The invention discloses a digital ball identification method based on sparse representation and discriminant analysis, which comprises the following steps of: putting each digital ball in a digital ball set under a monocolor background, continuously acquiring a single image or a plurality of images by using a single camera, automatically positioning the digital ball in each image, extracting visual features, and establishing sparse representation for all features to form a training sample feature set; putting one or more digital balls to be identified in the same scene, acquiring a single image or a plurality of images, automatically positioning all digital balls in each image, extracting sub-images, extracting visual features of the sub-images, and establishing the sparse representation of the sub-images by utilizing the training sample feature set; and identifying by adopting a discriminant analysis method, wherein for the condition of a plurality of images, joint posterior discrimination is adopted to improve identification precision. By fully utilizing the sparse representation and the discriminant analysis, the method is applied to identification of a single ball or a plurality of balls under the monocolor background and has good identification effect.

The invention discloses an exponent regularization and null space linearity discriminant analysis-based fault diagnosis method. Regularization discriminant analysis is combined with null space discriminant analysis, advantages of NSLDA and RLDA in terms of mode identification are integrated, a regularized intra-class sample matrix Sw1 is used for replacing a class sample matrix Sw in null space discriminant analysis, small sample problems can be further solved; in discrimination criteria for discriminant analysis, an exponential function is introduced, the regularized intra-class sample matrix Sw1 and an inter-class sample matrix Sb are respectively subjected to exponential operation, and therefore more characteristic information can be obtained; faults can be effectively and accurately identified, fault diagnosis precision can be effectively improved, and a new train of thought is put forward for small fault diagnosis based on data driving.

The invention relates to a RGB-D target recognition method based on quaternion generalized discriminant analysis, including quaternion-based RGB-D image representation method, define quaternion generalized discriminant analysis, two-way quaternion generalized discriminant analysis recognition method base on average row and average column; wherein the quaternion-based RGB-D image representation method solves that problems of data redundancy and extra computational overhead when the existing quaternion color image representation method adopts four-dimensional quaternion to represent three-dimensional color images, The kernel function is introduced into the field of quaternion subspace analysis, and the quaternion generalized discriminant analysis is defined, which solves the problem that theexisting quaternion subspace analysis algorithms are not effective in processing quaternion nonlinear signals. Finally, the bidirectional quaternion generalized discriminant analysis based on averagerows and average columns is used to eliminate the influence that aiming at the RGB-D recognition process, the computational complexity of quaternion kernel matrix eigendecomposition is too large, which improves the recognition effect of the target recognition method.

The invention discloses a hyperspectral image dimension reduction method, which comprises the following steps: firstly, dividing an original hyperspectral image into non-overlapped superpixels by using an over-segmentation method; next, as the pixel point in one super-pixel usually belongs to the same kind of objects, describing the spatial information by using an intra-class graph in the invention; and finally, introducing the intra-class graph based on the super-pixel level into an LGDE model as a regular item. In addition, in order to effectively capture the nonlinear characteristics of thehyperspectral image, the invention expands the linear LGDE into a kernel version. An original pixel point classification method (RAW), a principal component analysis (PCA), a linear discriminant analysis (LDA) method, a spectral space linear discriminant analysis (SSLDA) method, a local reservation projection (LPP) method, a collaborative graph-based discriminant analysis (CGDE) method, a sparsegraph-based discriminant analysis (SGDE) method, and a local graph-based discriminant analysis (LGDE) method are compared. Under the same experiment conditions, the classification result of the methodis more accurate.

The invention relates to the technical field of power grid power supply. The method comprises the following steps of: predicting the monthly load increase condition of the existing enterprises in theindustrial park: establishing an existing enterprise maximum load prediction model in the park by adopting support vector machine regression prediction based on a machine learning algorithm, and predicting the existing enterprise maximum load increase by utilizing the maximum load prediction model; predicting the newly added enterprise load demand condition in the park; constructing an enterprisebasic information index system and a maximum load demand scale discriminant analysis model; applying a discriminant analysis model, and in combination with basic information of newly-added enterprisesin the future of the park, subjecting maximum load requirements of the newly-added enterprises to discriminant prediction; and summarizing the existing enterprise monthly load demand of the park andthe newly added enterprise load demand of the park to obtain the monthly load increment information. According to the method, load demand prediction is carried out by combining the existing enterpriseload increase demand of the park built in the future and the maximum load demand of newly-added enterprises, and the short-term power consumption increase demand of the park is effectively judged.