164results about How to "Improve separability" patented technology

The invention discloses a method for reconstructing an outer outline polygon of a building based on multivariate data. The method comprises the following steps of: respectively dividing DSM (Design Standards Manual) data and image data so as to obtain a mask image of an interest region of the building and an image dividing object; combining the mask image with the image dividing object so as to obtain a complete building object; carrying out boundary tracing on the building object so as to obtain curves of the building; using points corresponding to local maximum curvature values of the curves as angular points; connecting the angular points in sequence so as to obtain the outline polygon of the building; dividing the building object into regions by using a hierarchical clustering method and calculating the main direction of the building; establishing a linear model of the polygon of the building and correcting and regularizing the linear model of the outline of the building with the combination of the main direction of the building and gradient information of the image data; calculating an intersection point of every two adjacent straight line sections by using the linear model of each line section of the polygon; and by taking the intersection points as the angular points, connecting the angular points in sequence so as to form the final polygon of the building. According to the method, the DSM data is organically combined with the image data, the data are complementary to each other in the whole process, so that the problem of reconstructing the polygon of the outline of the building is solved well, and the method has very strong robustness in the two-dimensional outline modeling aspect of the building.

The invention discloses a new rice and old rice quality detection device and detection method based on a visual light, near-infrared and terahertz integrated spectrum technology. The detection device is composed of a halogen light source, a visual light-near-infrared hyperspectrum sensor, a transmission-type terahertz time-domain spectrum system, a rotation-type sample carrying device, a groove-type sample table, an electronic control detection rocker arm, electronic control sample table rocker arms, a computer, a data acquisition card, a control module, a mode switching button, a shading hood, an instrument rack and the like. When detection is conducted, visual light-near-infrared hyperspectra and terahertz spectra of a sample are collected through the mode switching button. An optimal characteristic wave band combination of the related visual light, near-infrared and terahertz integrated spectra which are stored in experiments is extracted and substituted into a related quantitative detection model, and qualitative results of the related new rice and old rice quality of the rice sample and related protein and amino acid are obtained and output. Accordingly, limitation generated when the new rice and old rice quality is detected through a single spectrum technique is broken through, and the detection device and detection method can be used for real-time storage quality detection on warehouse rice.

The invention provides a GIS automatic imaging system and an automatic imaging method, which are characterized in that: WEB technology, mobile calculation technology, microelectronics technology and GIS technology are utilized to perform the automatic processing, checking and generating of GIS information data from GPS information data, multimedia information data and business information data. The system comprises three parts, namely, a GIS geographical information marking system, a data acquisition safeguard system, and a GIS data analyzing system; wherein, the GIS geographical information marking system is used to provide the approaches and methods for the acquisition of geographical object information, and guarantee the validity of the acquisition of the geographical information by controlling the authority and scope of agents; the data acquisition safeguard system is used to acquire the multimedia information and geographical information through various devices, and safeguard the accuracy of the geographical information by use of a range finding system, a self-stabilizing platform system, a GPS positioning module as well as a vehicle-mounted computer system; the GIS data analyzing system is intended for integrating and synthesizing the multimedia information and geographical information, and automatically checking the integrated information data. The invention has the advantages of more convenient operations, higher precision and shorter acquisition time.

The invention discloses a method for classifying multi-spectral remote sensing data land use based on semi-supervisor manifold learning, relating to a land use classification method. The method comprises the following steps of: taking the multi-spectral remote sensing data as a sample data set according to a wave band generator matrix of the data; selecting a part of sample data from the sample data set, marking sample class labels according to priori knowledge, and randomly selecting a part of sample data as unmarked data from the sample data set; establishing a similarity graph and a difference graph to measure the similarity and the difference of data points, and calculating a weight matrix; calculating according to an optimal target function to obtain a projection matrix; projecting the whole multi-spectral remote sensing data; and executing the land use classification by using a K-adjacent classification algorithm. The invention adds the randomly selected unmarked sample data by utilizing a semi-supervisor manifold learning method, calculates the projection matrix by the optimal target function so as to increase the precision of the land use classification and effectively saves the cost of marking the training sample classes at the same time.

The present invention discloses a self-adaption fault diagnosis method based on permutation entropy (PE) and manifold-based dynamic time warping (MDTW), enabling a bearing fault diagnosis process to be systematic and raising handleability and real-time performance of the diagnosis method. Firstly, a nonlinear and nonstationary bearing vibration signal is decomposed into a plurality of single-package components by applying an adaptive time-frequency analysis method; the adaptive time-frequency analysis method may be selected from empirical mode decomposition, local mean decomposition and local characteristic-scale decomposition methods; and then, extracting the PE of each single-package component as a fault signature. The PE can reflect complexity of the signal and has high robustness and rapidity. The MDTW method is provided by the present invention so as to rapidly and accurately measure distance test data and training data, thereby determining the current fault state and realizing bearing fault diagnosis, and the method has excellent practical engineering application values.

The invention discloses a semantic role labeling method based on a synergetic neural network, and relates to the fields of semantic role labeling, mode identification and synergetic neural networks, in particular to a method for introducing the principle of the synergetic neural network into shallow semantic analysis. The semantic role labeling method comprises the following steps: extracting characteristics from training language material and testing language material and constructing corresponding semantic characteristic vectors; performing kernel transformation on the semantic characteristic vectors and constructing a prototype pattern and a mode to be tested on the basis; constructing an order parameter and calculating a plurality of candidate roles for each dependent component; constructing a predicate base and combining the candidate roles of all the dependent components corresponding each predicate to get role chains of all the predicates; and optimizing a network parameter, performing dynamic evolution on the synergetic neural network to get an optimal role chain, and outputting the labeling mode. The principle of the synergetic neural network is firstly introduced into the semantic role labeling, and the method can be widely applicable to various natural language processing tasks. The semantic role labeling method has better application prospects and application value.

The invention discloses a statistical dictionary learning-based radar high range resolution profile (HRRP) target identification method. The method comprises the steps of obtaining continuous HRRP original signals of T targets, and preprocessing the signals; dividing a main frame into two sub-frames according to a maximum probability difference algorithm, and obtaining an initialized statistical dictionary; configuring related parameters before training of the statistical dictionary; performing training to obtain an optimal dictionary and a transposed matrix; and performing test identificationclassification on the to-be-tested HRRP original signals by utilizing the transposed matrix. The method is especially applied to HRRP target identification under a low-signal-noise-ratio condition; and compared with a single statistical modeling and dictionary learning method, the method provided by the invention has better identification performance.

The invention relates to a bearing fault mixing intelligent diagnostic method based on second generation wavelet transform and a BP neural network. The bearing fault diagnostic method based on the second generation wavelet transform and the BP neural network includes steps: firstly, using the second generation wavelet transform to resolve a bearing original vibration signal measured by a sensor; secondly, extracting time domain statistical features and frequency domain statistical features from the resolved signal so as to form a combined feature set, and then performing feature evaluation on the extracted feature set so as to obtain a sensitive feature set; using the sensitive feature set as input of the BP neural network for network training, and building a fault diagnostic model based on the BP neural network so as to achieve classification and diagnosis of faults. The bearing fault diagnostic method based on the second generation wavelet transform and the BP neural network and the fault diagnostic model based on the BP neural network are used in the classification and the diagnosis of the bearing faults, and results indicate that the bearing fault diagnostic method based on the second generation wavelet transform and the BP neural network is high in classification and diagnosis accuracy, high in speed and high in efficiency, effectively improves bearing fault diagnostic effects, and is conveniently used in engineering practice.

A transferable liquid crystal laminate comprises at least a releasable substrate, a protective layer and a cholesteric liquid crystal layer, characterized in that the protective layer side of the releasable substrate has been subjected to an easily separable adhesive treatment. The transferable liquid crystal laminate is excellent in separability and the appearance of a transferred portion after being transferred and is reduced in the formation of burrs, resulting in the decrease of yield.

The invention relates to a multispectral face identification method, and a system thereof. The system is characterized by comprising a multispectral imaging system, a color camera, a face identification module, a data storage module, a central control module, and a spectrometer. The multispectral imaging system outputs filmed face image data to the face identification module. The face identification module identifies the face image data according to information in a standard face database in the data storage module, and then outputs an identification result. The central control module controls image filming of the multispectral imaging system and identification of the face identification module. The multispectral imaging system comprises an objective lens, a liquid crystal adjustable optical filter, and a CCD camera. The liquid crystal adjustable optical filter is disposed in front of a CCD camera lens of the CCD camera, and the objective lens is disposed on a front end of the liquid crystal adjustable optical filter. In the method, a plurality of characteristics of the face image is extracted, thereby making a between-class distance in an identification process more obvious and separability of identification algorithm better which help to improve identification effects.

This invention relates to an image process and mode identification technology, especially to a binary method for file images, which carries out initial rating of front background pixel to an image to analyze stroke adjacent domain information including grayness information, grads information and geometrical information then to reinforce the image to character strokes based on the stroke adjacent domain information and finally carries out binary-state on the reinforced image. This invention also puts forward a quick rating method and an improved method for getting a binary threshold value based on the Niblack.

The invention provides a method for reducing dimensions of high spectroscopic data based on a factorial analysis model, comprising the following steps: (1) reading in the high spectroscopic data; (2) establishing the factorial analysis model for dimension reduction of the high spectroscopic data; (3) calculating average of data, covariance matrix and correlation matrix; (4) calculating proper value and standard eigen vector of the correlation matrix of the data; (5) carrying out solving on factor loading matrix by a parametric estimation method; (6) calculating covariance of special factors and communality of data variables in the factorial analysis model; (7) calculating the biggest factor loading spin matrix based on variance; (8) calculating factor scores by a least square method based on variance; and (9) obtaining eigen dimensionality characterizing high spectroscopic data, thereby realizing dimension reduction of the high spectroscopic data. The method is an automatic method for dimension reduction of the high spectroscopic data, which can effectively remove relativity of wave bands of the high spectroscopic data and improve separability of different categories of ground objects.

The invention discloses a hyperspectral image semi-supervised classification method based on space-spectral information. The hyperspectral image semi-supervised classification method combines spectral information and spatial information in a hyperspectral image to act on a support vector machine classifier, adopts a self-training semi-supervised classification framework, utilizes an active learning method as a sample selecting strategy of semi-supervised classification, decomposes initial classification results obtained through semi-supervised classification according to classes so as to obtain various classes of binary images as input images of an edge preserving filter, regards a first principal component content as a reference image of the filter, utilizes the edge preserving filter to perform local smoothing, eliminates noise, and classifies image elements according to a class with maximum probability, thus the classification process is completed. The hyperspectral image semi-supervised classification method combines the spectral information and the spatial information to improve the classifiability of classes, utilizes the self-training semi-supervised classification framework to solve the classification problem of hyperspectral image small samples, can effectively eliminate spot-like errors in the initial classification results, and increases classification precision.

The invention relates to electroencephalographic signal extraction, in particular to an electroencephalographic signal processing method based on isolated component automatic-clustering process. In order to improve signal-to-noise ratio for inducing an electroencephalographic signal, to remarkably improve the separability of the electroencephalographic signal within an action period under the stimulation of different tasks, and to facilitate the extraction and analysis of signal characteristics and the recognition of the task mode, the invention adopts a technical proposal which is in particular as below: the electroencephalographic signal processing method based on isolated component automatic-clustering process comprises the following steps: firstly, adopting an online maximum information algorithm (Informax algorithm) based on an information maximization criterion to sequentially carry out isolated component analysis (ICA) on stimulated and induced multi-channel signals, constructing a large component sample set Y with all obtained components, calculating the mutual information of the components, and finally adopting a total intra-class distance minimization criterion to carry out clustering process on a mutual information distance matrix so as to obtain class tags of all the components.

In a method for operating a radar sensor, the unambiguousness range of the radar sensor is increased with respect to the range and / or the relative velocity by: transmitting multiple ramp sets by the radar sensor, the frequency ramps of the individual ramp sets each differing in one system parameter; adapting the sampling frequency during the detection of the radar echoes in such a way that a constant number of samples always results for each frequency ramp; and, to evaluate the radar signals, the spectra are periodically continued and compared to each other.

The invention discloses a hyperspectral remote sensing image target detecting method based on variable end members, comprising the following steps of: selecting a remote sensing image to be processed by target detection; acquiring prior information required for detection, wherein the prior information comprises spectral information of target end members and spectral information of background end members; traversing the remote sensing image to be detected by utilizing a cross correlation matching technique to determine the types of background end members in each pixel in the remote sensing image to be detected; carrying out spectral decomposition on the remote sensing image to be detected in a completely restricted least square way to acquire the component information of target end members and various background end members in each pixel in the remote sensing image to be detected; establishing a detector based on the GLRT (Generalized Likelihood Ratio Test); and traversing the remote sensing image to be detected by adopting the detector to acquire the detection function value of each pixel in the remote sensing image to be detected, thereby judging whether targets exist in each pixel in the remote sensing image to be detected or not. The method of the invention has the characteristics of strong structuration, high adaptability, self-organization and self-learning.

The invention discloses a method for weighting multiple example studying features based on a master space classifying criterion. The realization scheme of the method comprises three steps of initializing a positive package representative example and a negative package representative example, building a problem to be optimized, and updating three kinds of unknown variables of the problem to be optimized. A representative example which can right express the category mark of a package in a positive package is found by adopting a heuristic search method, so that the problem of the fuzzification of the category mark of the example in the positive package is solved; repeated iteration is performed by adopting a coordinate rising method, so that the problem to be optimized can be converged into a local optimum solution; a relative weight is given according to the size of the contribution of each feature to recognition, and compared with the method of using original data for recognition, when the data which are weighted by features are used for recognition, higher recognition precision can be obtained.

The invention provides a multi-direction text detection method of a natural scene. The multi-direction text detection method comprises the following specific steps: 1) extracting a boundary ascension MSER (Maximally Stable Extreme Region), and according to a boundary fit goodness formula, recursively ejecting one area with small boundary fit goodness from two areas which have a father-only child relationship on a stable extreme value area ingredient tree obtained by an original MSER algorithm, wherein the area change [Delta] S of the two areas does not exceed a first threshold value; 2) sorting a character sorting tree area; and 3) carrying out character multi-layer fusion to form text lines, carrying out multi-layer fusion on a sorted character area set which is finally obtained in the step 2), and finally generating the text lines, wherein multiple layers are successively an expansion fusion layer, a free growth layer, a bijection growth layer and a competition layer.

The invention provides a feature selection method based on a hierarchical deep network. The invention discloses a feature selection algorithm for performing selective orthogonality on depth features of different levels of a tree classifier. the features extracted by each layer of classifier better meet the requirements of respective classification tasks; the characteristic separability is improved; the influence of similarity characteristics among the categories on the network image recognition capability is effectively inhibited; and during back propagation, the knowledge graph is utilized toguide the network to update the feature selection parameters, so that the network can pay more attention to the similarity among the categories in the coarse category during rough classification, andcan pay more attention to the difference among the similar categories during fine classification. According to the method, the effectiveness and separability of the features are improved, the recognition capability of the whole network structure is improved, and the classification accuracy is improved. According to the method, a better classification effect is achieved on each data set.

The invention discloses a radar radiating source signal sorting method for adaptively setting and automatically adjusting parameters. The method comprises the following steps: introducing intra-pulse features, namely symmetry Holder indexes for sorting while combining with a pulse arrival angle during selection of a feature vector; setting an expected class number by utilizing a statistic method for searching a histogram peak value according to features of the feature vector based on an iterative self-organizing data analysis technique algorithm (ISODATA for short); preliminarily setting the upper limit of standard deviation of each data distribution in the ISODATA class and the lower limit of each class of inter-central minimum distances; and finally, controlling and setting an adaptive criterion for the upper limit of the standard deviation of each data distribution in the class and the lower limit of each class of inter-central minimum distances during operation of the ISODATA, thereby achieving the accurate final sorting aim. According to the manner, the signal sorting problem is solved.

The invention discloses a 2-methyl-3-methoxybenzoyl chloride synthesizing process. According to the present invention, low-cost o-xylene is used as a starting raw material, the product is synthesizedby using a conventional synthesis method comprising nitrification, esterification, reduction, diazotization, methylation, acyl chlorination and other steps, and the total yield is controlled at more than 65%; the esterification of the intermediate product improves the separation degree of the intermediate; the reaction solvent is added in the diazotization step, such that the process parameters are relatively easy to control, and the purity of the intermediate in the diazotization step is more than 96% so as to provide the guarantee for the quality of the subsequent product; and with the synthesis process, the product quality is stable and reliable, and the cost is low.

The invention discloses a method and device for predicting user churn. The method comprises the following steps: extracting basic characteristic information of a user from historical user data; according to the basic characteristic information of the user, modeling parameters that can characterize the use characteristics of the user, and obtaining a secondary characteristic group of the parameters; according to the loss state of the user, training the preset classifier by using the secondary feature group to obtain a trained classifier model; and using the classifier model to predict user churn. The invention introduces a quadratic feature extraction mechanism, models the extracted quadratic features respectively to obtain a new feature group, and takes the new feature group as training data of a classifier model, thereby improving the separability of the loss and non-loss user experience, and further improving the prediction accuracy of the loss user.

The embodiment of the invention discloses a method for tracking a video target. The method comprises the following steps: a color characteristic space of an appointed frame image in an image to be tracked is constructed; the specific color characteristic is selected; according to the specific color characteristic, the specific color combination characteristic is selected; according to the color combination characteristic, a trust image of each other frame image in the image to be tracked is constructed; the trust image is optimized; and the position of a target to be tracked is searched in the optimized trust image. The method provided by the embodiment can accurately select the color characteristic so as to well express the target to be tracked so that separability between the target to be tracked and background is enhanced; and through the optimization of the trust image, partial background objects in the trust image are removed at the same time, thereby improving accuracy of a tracking result.

The invention relates to a shallow underwater topography construction method integrating hyper-spectral data and sparse sonar data and belongs to the technical field of underwater topography reconnaissance. According to the shallow underwater topography construction method, dimensions of a hyper-spectral remote sensing image are reduced by aid of a clustering center of sonar data, area division is performed on a low dimensional remote sensing image after dimension reducing, and interpolation is performed on sonar data inside each of the areas to obtain the underwater topography. According to the shallow underwater topography construction method, the hyper-spectral remote sensing image and sparse sonar data are organically combined, in the whole process, and two kinds of data compensate to solve the problem of underwater topography construction well. The remote sensing image and the sonar data are provided with coordinate information after geometric correction, and a certain fuzzy corresponding relation exists between a grayscale of the remote sensing image and the water depth, so that water depth in each depth homogeneous area changes slightly in area-divided remote sensing image, and interpolation results of sonar data are authentic.

The invention provides a process fault identification method based on a big data intelligent kernel independent component analysis, and relates to the technical field of fault diagnosis in the processindustry. The method includes: constructing a semi-supervised kernel independent component analysis algorithm through sample data, obtaining a spatial conversion matrix and a state projection matrixof the sample data, and constructing a production operation state library of each operation state category; and preprocessing newly acquired data, performing preliminary fault diagnosis through the obtained spatial conversion matrix and the state projection matrix, calculating a scoring factor of real-time condition data through an obtained confidence interval in each projection direction, calculating a FICD statistic, and performing accurate fault identification. According to the method, with the combination of a semi-supervised classification learning method based on class membership and thekernel independent component analysis, fault diagnosis and accurate fault identification of the operation state of an industrial process are performed according to the state projection matrix and theconstruction of the corresponding statistic, and the identification degree and the accuracy for identifying the smelting operation state of an electrical smelting furnace for magnesia can be effectively improved.

The invention discloses an image significance object detection method based on multiscale low-rank decomposition and with sensitive structural information. The method comprises the following steps of: a three-dimensional volume data generation stage: performing superpixel decomposition and DT (Delaunay Triangulation) on a two-dimensional image to form three-dimensional volume data corresponding to the two-dimensional image, a Biharmonic distribution calculation stage: obtaining a Biharmonic diffusion result of each superpixel point, a subgeneration description stage: performing histogram statistics on L2 distances between sampling points on Biharmonic isolines of the superpixel points to form shape description of the isolines, and a multiscale low-rank decomposition stage: obtaining a final significance object detection result by differencing sparse matrixes obtained by the low-rank decomposition under different scales and performing residual error matrix summation based on the shape description of the Biharmonic isoline of each superpixel point. The method is based on GPU (Graphics Processing) parallel implementation, can detect one or more significance objects in the image, and has the characteristics of high detection precision of the significance object, complete detection of the significance object, good noise resistance and the like.

Provided is a toner for electrostatic image development, having both sufficient low-temperature fixability and good post-fixing separability simultaneously. The toner for electrostatic image development includes: a second amorphous polyester resin including a structural unit represented by any of the following general formulas (1) to (3); and a first amorphous polyester resin including no structural units represented by the general formulas (1) to (3).[In the general formulas, R1, R2, R5, R9 and R10 are each independently an alkyl group having 4 to 15 carbon atoms or an alkenyl group having 4 to 15 carbon atoms, R3, R4, R7, R8 and R11 are each independently an alkylene group having 4 to 14 carbon atoms or an alkenylene group having 4 to 14 carbon atoms, R6 is a saturated or unsaturated divalent aliphatic hydrocarbon group having 4 to 15 carbon atoms, and R12 is a saturated or unsaturated trivalent aliphatic hydrocarbon group having 4 to 14 carbon atoms, and X is an aromatic ring, a carbocyclic ring or a group represented by the following formula (A).]

The invention relates to a mechanical vibration signal feature extraction method based on combination of stochastic resonance and kernel principal component analysis. The method comprises the steps that firstly, the stochastic resonance method is applied for conducting pretreatment on rotor oscillation original signals measured by a sensor, the signal periodicity is improved, and the oscillation signal to noise ratio is improved; then, a time domain feature set is extracted for the pretreated output signals; then the kernel principal component analysis method is adopted for conducting nonlinear feature transformation for the extracted time domain feature set, and therefore the final needed feature set is obtained. The method is applied to feature extraction and failure diagnosis of simulated failure of an engine rotor, the result shows that the feature set extracted through the method is of linear independence, the number of dimensions is smaller, the separability is higher, the precision and efficiency of the failure diagnosis can be effectively improved, and application in engineering practice is facilitated.

The invention provides a hyperspectral remote sensing image end member extraction method on the basis of a revised extended morphological operator, which is characterized in that an original image is filtered by the opening-and-closing operation and the closing-and-opening operation which are defined by the revised extended morphological operator so as to fulfill the aim of combining spectral information with space information to carry out end member extraction. The method mainly aims to solve the problem of limitation in the extended morphological operator; the revise is carried out by leading in a reference vector; for the provided revised extended morphological operator, the alternately replacing phenomenon is eliminated; the correct replacing direction is ensured; and a separating effect of pure pixels and mixed pixels is reinforced. An experimental result shows that the condition that the method provided by the invention can optimize the end element extraction effect is proved onthe aspects of the spectral curve similarity and a corresponding ground feature distribution map and the like; and the method has a moderate calculated amount and lays a foundation for application ofsubsequent analysis and classification of hyperspectral remote sensing image mixed end members and the like.