133results about How to "Fix extraction issues" patented technology

The invention relates to a man-machine multi-round dialogue method oriented to the travel field, which is used in the information technology field. The invention solves the problem that the current multi-round conversation system has difficulty in extracting the intention information and the slot information of the user question. The invention normalizes the short text question, extracts the intention information of the normalized short text question by using DAN, CNN or BLSTM model, and obtains Micro. F1 value was 93.47%. Using the BLSTM-CRF model to extract the slot information of the normalized short text questions, and the model achieves the ideal effect with F1 value of 89.47%. Taking the historical slot information and the slot information of the current question as inputs, the current conversation state information is determined, and the next reply strategy is determined according to the intention information of the current question. According to the determined reply strategy, the corresponding template is selected to reply to the user. The invention can be applied to the field of information technology.

The invention relates to a workpiece recognition method based on a geometric shape feature and a device thereof. The method comprises the steps of (1) taking a two-dimensional image of a workpiece on a conveyor belt, carrying out difference operation on the two-dimensional image of the workpiece on the conveyor belt and a conveyor belt image, and obtaining a foreground area comprising the workpiece and the shadow of the workpiece, (2) using a shadow detection method to remove the shadow in the foreground area comprising the workpiece and the shadow of the workpiece to obtain the accurate area contour of the workpiece, (3) extracting the geometric feature for the accurate area contour of the workpiece so as to obtain the feature vector of the workpiece area contour, wherein the feature vector comprises a Hu moment and a Fourier operator, and (4) using a support vector machine SVM to train and classify the feature vector of the workpiece area contour. Through employing the above steps, the method and the device can be widely applied to the fields of workpiece grasp and transportation in a factory production line, circumferential welding manipulator, industrial painting and equipment assembly.

The invention provides a de-noising method of a transient electromagnetic detecting receiving signal, which comprises the following steps of: decomposing a transient electromagnetic detecting receiving signal by means of empirical mode, i.e. 'screening'; leading all extreme points of signal data x (t) to form into an upper envelope and a lower envelope through a cubic spline interpolation function; obtaining a mean value m1, wherein a first component h1=x (t)-m1, and the h1 is regarded as data to be processed to repeat the screening with k times; obtaining a first IMF when the standard deviation SD of the continues twice screening results meets the requirement, wherein c1=h1k=h1(k-1)-m1k and a residual signal r1; repeatedly 'screening' the r1; decomposing signal data to be n IMFs and one residual quantity rn; judging signal dominant and noise dominant component according to the diminishing condition of each IMF energy; and accumulating again after filtering each IMF of the signal dominant to obtain the de-noised signal. The method adopts empirical mode decomposition of adaptive time-frequency analysis, does not need priori information, respectively processes the different natural frequencies of the signal, removes high-frequency noise, filters high-frequency noise, is applied to removing the noise of the non-linear and non-stationary transient electromagnetic detecting echo signal, and is patricianly applied to removing the noise of a deep echo signal.

The invention discloses a high resolution image semantic information extraction method and system. According to the method, a dual multi-scale manifold sorting convolutional neural network is employed to solve a high resolution image semantic information extraction problem, multi-scale information characteristics of a pooling layer of the convolutional neural network are utilized, a multi-scale convolutional-expansion convolutional dual network layer is constructed, a continuous domain inner manifold sorting optimization method is further employed, and multi-scale space results are optimized. The method is advantaged in that the multi-scale optimization results are utilized, fusion processing on the extracted semantic information is carried out in a mean value fusion mode, and thereby the final semantic segmentation result of high resolution images is acquired.

The invention discloses a convolutional neural network feature extraction method based on principal component analysis. The convolutional neural network feature extraction method based on principal component analysis includes the steps: selecting the convolutional neural network which has been trained on an Imagenet data set, taking the convolutional neural network as the feature extractor of an image, extracting the feature mapping graph from the output of each pooling layer of the convolutional neural network, taking all the extracted feature mapping graphs of each layer as the deep featuresof the image, utilizing principal component analysis to perform dimensionality reduction on the deep features, and utilizing bilinear interpolation to reset the final result feature mapping graph tothe original image size to obtain the efficient image deep features. The deep features obtained through convolutional neural network feature extraction method based on principal component analysis have rich semantic information of the image, are low in the feature dimension and small in data size, and can be used for various tasks of identification and classification of images.

The invention relates to a transient travelling wave amplitude integral type superspeed bus protection system which has the advantages of simple structure, rapid detection speed, high sensitivity and the like and is convenient in use and a method thereof. The method is implemented as follows: installing N numbered protection units at the outlet of each outgoing line if a bus is connected with N numbered outgoing lines; realizing mutual communication between every two protection units via a signal bus; collecting the current and voltage signals of the line by each protection unit at a high speed, and independently judging the failure direction; if a certain protection unit judges that the line has a forward direction failure, sending out a lockout action message, transmitting the judgment result to other protection units by signal buses, receiving judgment results about respective line failure direction from other protection units, and comprehensively analyzing the judgment results of all protection units by each protection unit so as to determine the bus state; if all the protection units judge that the failure comes from the reverse direction, determining that the bus has a failure so as to promptly give out an instruction to trip off respective side switch to isolate the failed bus; and otherwise, determining that the bus has no failure and requires no isolation.

The invention discloses an arc light grounding fault distinguishing method based on harmonic wave energy and waveform distortion characteristics. Firstly, waveforms and their harmonic amplitude valuesof three-phase voltage and zero-sequence current of a system are obtained through a power distribution network miniature PMU; the harmonic amplitude values of the waveforms of the three-phase voltageand zero-sequence current are utilized to calculate dimension-unified harmonic energy of each period, and whether the system has suspected arc light grounding faults or not is judged; randomness analysis is conducted on the harmonic energy of each period in a detection window to obtain randomness indexes of the harmonic energy, and it is determined whether identified suspected arc light groundingfaults are mistakenly judged or not; waveform distortion characteristics of stable arcing are extracted by conducting time-frequency domain analysis on the waveforms of the three-phase voltage and zero-sequence current, the influence of the stable arcing process on randomness analysis is avoided, and accordingly mistaken judgement is avoided. The arc light grounding fault distinguishing method accurately extracts arc light grounding fault characteristics and achieves the technical basis of a reliable, safe and quick fault distinguishing algorithm.

The invention discloses a process for preparing hirudin by using blood- suction leeches, as well as its use, the preparing process consists of pulping leeches, formulating suspension and drying. The substance can be prepared into capsules, soft capsules, granules and tablets. The obtained medicament can be used for treating cardiovascular and cerebrovascular diseases including cerebral apoplexy, hypertension, hyperlipemia and coronary disease.

The invention belongs to the field of measurement of the diameter and the space coordinate geometric dimensioning of a spatial curved surface thin-wall micro-hole, and relates to a method for measuring the diameter and the coordinate position of the spatial curved surface micro-hole in a non-contact mode, in particular to the method for measuring the diameter and the coordinate position of the spatial curved surface micro-hole in a blade body of a blade by compositely designing and applying a telecentric optical lens and a five-axis coordinate measurement machine. A contact-type probe of the five-axis optical composite coordinate measurement machine is used for setting up a measuring basis of the diameter and the coordinate position of the spatial curved surface micro-hole of a workpiece, a two-dimensional rotating table is used for determining a spatial posture according to the direction of the axis of the spatial curved surface micro-hole, and the telecentric optical lens is used for extracting an image of the spatial curved surface micro-hole. The method can solve the difficult technical problem that data extraction and evaluation can not be conduced due to the fact that the image of the micro-hole is blurred, digitalized precise measurement of the diameter and the coordinate position of the spatial curved surface micro-hole is achieved, and the feasible measurement method is provided for effectively controlling machining quality of a gas film hole of the blade of an aviation engine and the like.

The invention relates to an aerial target classification method based on wind field disturbance characteristics. The method includes steps of resolving wind field disturbance characteristics; accurately extracting the wind field disturbance characteristic parameters of trailing vortex core position, vortex core radius, vortex core spacing and vortex circulation by an algorithm for resolving wind field disturbance characteristics of an aerial target; inversing type identification characteristics; inversing to obtain type identification characteristic parameters of target track characteristics, physical characteristics and motion characteristics by an algorithm of inversing the type identification characteristics of the aerial target; identifying attribute of the aerial target; and outputting the type attribute of a plane target to be identified by inputting the type identification characteristic parameters of the plane target to be identified. Extraction of wind field disturbance characteristics of the aerial target is achieved, inversion of the wind field disturbance characteristics to target characteristics is achieved, and the aerial plane target is classified and identified on the basis of laser radar detection for specific airspace atmospheric wind fields.

The invention discloses a detonator shell dot character identifying method and belongs to the field of image processing and mode identifying. The detonator shell dot character identifying method includes that 1, carrying out graying and binaryzation pre-treatment on gathered character images; 2, carrying out morphological filter and near neighborhood search correction on discontinuous character images; 3, obtaining single character region images through a connected domain segmentation method; 4, using the distance between the character boundary and the zone boundary as a recognition feature value, and using a minimum distance classifier and a multi-stage identification method to obtain the final identified result. The detonator shell dot character identifying method is suitable for identifying fracture characters, and the detonator shell dot character identifying method solves the problem of bad detonator shell character identifying effect. The detonator shell dot character identifying method can also be used for identifying product labels, license number, print documents and the like, and the detonator shell dot character identifying method has advantages of broad application prospect and convenience in implementation on system modules.

The invention relates to a method and a system for extracting database error message, in particular to a method and a system for extracting the error message returned from a server side during the operation of six frequently-used databases and carrying out related operation audit used in network business audit product. The system comprises a database type identification locator, an error message return package locator, an error message locating and extracting device, an error code library and an audit device. The invention solves the problem that the audit of the error message returned from the server side in the database operation process is lacked in the prior audit product, and overcomes the defects that the error message audit is difficult because different database types use different protocols. The invention has very good expansibility. For new database types, the system and the method have flexible expansibility to expand audit range and very high extraction efficiency and accuracy for the error message of the different database types, and can be widely applied to network business audit products.

The invention discloses a method for extracting sweet buckwheat dietary fibers, which comprises an enzyme method and an alkaline method for extracting the dietary fibers in kernels and shells of the sweet buckwheat and makes the visual color of the obtained dietary fibers more acceptable by people and improves the application value of the dietary fibers by the decoloring and modification treatments. Currently, most processing residues of the buckwheat are generally abandoned or put away cheaply so as to cause the waste of a large amount of buckwheat resource. The invention adopts different methods to extract and carry out decoloring treatment to prepare the sweet buckwheat dietary fibers with higher activity and few hazardous substance residuals, applies the prepared buckwheat dietary fibers to the food industry, and applies the prepared buckwheat dietary fibers into other fields by modification. The invention can solve the technical problems of extraction, decoloring, modification and the like of the sweet buckwheat dietary fibers, can improve the competitive advantage of the economy of the buckwheat production place, realizes the industrialization development and utilization of the sweet buckwheat dietary fibers, and promotes the economy development of the buckwheat production place.

Embodiments of the invention disclose an enterprise relationship graph construction method and apparatus, a computer device and a storage medium. The method comprises the steps of crawling enterpriseinformation from a national enterprise credit information publicity system / enterprise annual report / by using a web crawler algorithm; based on a named entity recognition algorithm, extracting enterprise entities and entity relationships from the crawled enterprise information; determining an attribute value of the enterprise entity, and storing the attribute value of the enterprise entity and theentity relationship into a preset graphic database; and generating an enterprise relationship graph according to the preset graph database. The enterprise relationship graph construction method basedon a knowledge graph aims to solve the problem that enterprise information data cannot be massively extracted and analyzed in the prior art by establishing a relationship graph and acquiring a large amount of enterprise information data.

The invention provides a feature selection and weight calculation method of an imbalance text set, and belongs to the field of text information processing. In order to solve the classification problem of imbalance text data, a feature selection and weight calculation method and system are provided. The category discrimination degree and the average word frequency factor are combined, the chi-squared statistic method is improved so as to conduct feature selection, meanwhile, a commonly-used feature weight calculation method is improved, and on the basis of the improvement, the TF-IDF weight calculation method is provided. The effect of the method on solving the imbalance data set problem is superior to that of a traditional feature selection method, and the method is effective and feasible in effectively improving the classification accuracy.

The invention discloses a CAD drawing-based building indoor cross-floor space extraction method. The method comprises the following steps of: obtaining a building single-floor space extraction result; extracting a building cross-floor space identifier; judging a building cross-floor space communication relationship; and carrying building cross-floor space modeling. According to the method disclosed by the invention, cross-floor spaces in buildings can be directly extracted according to CAD drawings, so that the deficiency, caused by the traditional indoor space extraction algorithm, to the cross-floor spaces in the buildings is made up, the building cross-floor indoor space extraction problem can be effectively solved, the indoor space modeling errors caused by neglecting the cross-floor spaces are avoided, and the algorithm correctness and practicability of building indoor space extraction are improved.

The invention discloses a method for segmenting aorta in a CT image combining edge and area characteristics. The method combines edge and area characteristics, by means of level set theory, can obtain clear and accurate object contour, implements segmentation of the aorta of the CT image, has very rapid operation and provides accurate segmentation result, better addresses the extraction of the aorta in the CT image, and provides a reference to medical diagnosis.

The invention relates a method for extracting binuclear lymphocyte region accurately, quickly and automatically in CB method micronucleated cell image in the field of radiation biodosimetry; the method comprises the following steps: performing algebraic subtraction (component R-component B) to a collected micronucleated cell image MN_rgb in RGB format with a size of 1500*1000pixel to obtain an image MN_rb, performing median filtering (5*5) and morphologic filtering (disk, r=6) to the image MN_rb and eliminating the region with adhered border to obtain an image MN_m, then using a roundness threshold and a grey level threshold to screen independent lymphocyte in a target area that Area is a member of [1000pixel, 3500pixel], then using the roundness threshold, the grey level threshold and theextension degree to screen cells in series connection in a target area that Area is a member of [3500pixel, 8000pixel] and separating, and judging whether the independent or separated lymphocyte is binuclear lymphocyte. The method designed by the invention realizes the automatic extraction of binuclear lymphocyte in CB method micronucleated cell image, has the advantages of accurate and rapid extraction and the like and takes a key step for researching the automatical analysis system of CB method micronucleated cell image.

The invention relates to a vector point spatial data full-blind watermarking method based on grid dividing. The transform domain watermarking algorithms have the characteristic of being high in robustness. At present, most transform domain watermarking algorithms specific to linear vector space data and planer vector space data are characterized in that a watermark is embedded with a geometric object as a unit, but geometric objects of point spatial data are all independent points, and the situation that the transform domain watermarking algorithms are directly applied to the vector point spatial data watermarking can be difficultly achieved. If the whole of the point spatial data serves as a geometric object and a watermark is embedded in the geometric object, extraction of the watermark can be influenced by adding and deleting spatial points. The vector point spatial data full-blind watermarking method based on grid dividing achieves the watermark embedding and watermark extraction aims of the vector point spatial data under transform domains. The vector point spatial data full-blind watermarking method is advanced and scientific, the watermark is high in robustness and good in nonvisibility, and meanwhile the data accuracy after the watermark information is embedded can be guaranteed. An experiment shows that the vector point spatial data full-blind watermarking method has good robustness on attacks such as adding and deleting point operations, cutting, data object sequence scrambling and geometric transformation operations of the vector point spatial data, and has the good use value.

The invention discloses a proxy application and system for a host system with a service container. The proxy application is arranged in the host system, the proxy application monitors the service container in the host system, when finding that a start interface or a restart interface of the service container is called, the proxy application registers the corresponding service information and address information of the service container in a service registration center; and when monitoring that a stop interface of the service container is called, the proxy application deletes the corresponding service information and the address information of the service container from the service registration center. Correspondingly, the invention further discloses a system comprising the proxy application. The proxy application and the system disclosed by the invention can register the corresponding information of the service container under container deployment. The proxy application and the system disclosed by the invention can be applied to Internet insurance business, and can realize a container level virtualization deployment scheme based on a Docker container so as to improve the application development efficiency and the resource utilization rate.

The invention belongs to the ship target ISAR imaging simulation technology field, and especially relates to extraction of structure features of length, width and the like of a deck of a ship target. Firstly, a ship target is simulated by utilization of a scattering point model, then ISAR imaging of the ship target in a sea clutter environment is simulated, finally, the ISAR image is pretreated, and the structure features ( the deck length and the deck width) of the ship target are obtained. The extraction method solves the problem of extraction of structure features of a ship target, and has good effects on extraction of structure features of ship targets at different angles.

The invention relates to a volatile oil extracting, collecting and separating device which is a device for collecting and separating volatile oil, and applicable to the fields of chemical engineering, medicines, foods and the like. The volatile oil extracting, collecting and separating device comprises an extracting tank, a condenser and a collecting and separating facility, wherein the top of the extracting tank is collected with the upper end of the condenser by virtue of a collecting pipeline, the lower end of the condenser enters the inner layer of the collecting and separating facility by virtue of an inlet pipeline, and the outer layer of the collecting and separating facility is connected with the extracting tank through an outlet pipeline; the collecting and separating facility is provided with an air filter; the air pressure balance between the inner layer and the outer layer is kept so as to prevent pollution. The device disclosed by the invention is simple, strong in maneuverability, and applicable to the extraction and separation of volatile oil in the fields of chemical engineering, foods, medicines, and the like.

The invention relates to a high-precision global vision measurement method for a three-dimensional thin-wall structure weld joint. The method comprises the following steps: (1) preparing a workpiece and a measurement system; (2) calibrating the measurement system and optimizing parameters; (3) measuring a space curve weld joint; (4) carrying out data splicing and welding seam extraction; (5) splicing a skin and blanking; and (6) calibrating a robot welding track. The method is innovatively applied to the field of measurement of the three-dimensional contour weld joint; on-site automatic measurement of the three-dimensional contour of the actual space curve weld joint is achieved, the weld joint position and the contour characteristics of the weld joint are recognized and extracted, on thisbasis, weld joint track calibration in the robot off-line programming process is achieved, and accurate discharging of the skin in the weld joint assembling process can be achieved.

The invention provides a webpage information analysis method including the steps of matching a URL(uniform resource locator) of a to-be-analyzed webpage to a domain name of an analyzable domain name set, positioning to a corresponding URL feature set according to the domain name successful in matching, matching the URL of the to-be-analyzed webpage to URL feature of the URL feature set, positioning to a corresponding goods template set according to the URL feature successful in matching, matching the to-be-analyzed webpage to the goods template of the goods template set, analyzing the to-be-analyzed webpage according to the goods template successful in matching, and then feeding back an analysis result. The invention further provides a webpage information analysis device correspondingly. By the webpage information analysis method and device, various types of goods URLs of a web can be identified accurately, different type of goods templates are used to perform matching and identifying to different type of goods URLs, and thus goods information on the webpage can be analyzed as much as possible.

The invention belongs to the technical field of partial discharge detection of high-voltage direct-current transmission equipment, discloses a method for extracting partial discharge waveform features of a converter transformer, and particularly relates to the method for extracting the partial discharge waveform features of the converter transformer on the basis of a multi-element feature extraction package. The method includes the steps that a file is read, screened and processed in advance, the multi-element feature extraction package is called, discharge waveform time-frequency spatial features are obtained, and the extracted partial discharge waveform spatial features are integrated and stored. The method is applied to partial discharge detection and insulation defect diagnosis of the converter transformer in a high-voltage direct-current transmission system. According to the purpose of partial discharge waveform feature extraction of the converter transformer, the concept that the multi-element feature extraction package is called is adopted, and the effective achieving method is provided. The method overcomes the defects of an existing method for extracting the partial discharge waveform features of the converter transformer and has the advantages of being rapid and accurate and features are comprehensive and detailed.

The invention discloses a target detection and identification method based on structural similarity measurement, and the method comprises the steps: predefining a template image to form a data set, reading a to-be-detected image, carrying out gray transformation, balancing the brightness of the to-be-detected image, obtaining a preprocessed to-be-detected image, solving an optimal segmentation threshold value of the preprocessed to-be-detected image, performing binarization processing on the to-be-detected image preprocessed in the step 2 by taking the optimal segmentation threshold as a global threshold to obtain a binary image of the preprocessed to-be-detected image; 3, performing morphological opening operation on the binary image in the step 3 to eliminate noisy points in the binary image, filling holes in the binary image to obtain an expansion result image, and marking candidate connected regions in the expansion result image as candidate targets and extracting all candidate targets; and matching the target template image in the step 1 with the candidate targets in the step 5 to obtain the structural similarity measurement of the candidate target, and taking the candidate target corresponding to the maximum structural similarity measurement as a final target.

According to a long-text structured text abstract extraction method provided by the invention, word vectors can be dynamically obtained according to surrounding words by adopting a dynamic word embedding method, so that the problem of polysemy in the text is solved. Text structure analysis is adopted so that paragraphs are reasonably divided according to relation recognition result between sentences, and a computer is made to understand the text from the global perspective. Abstract extraction based on a model and a rule is adopted, and abstract extraction is carried out on each section on thebasis of chapter structure analysis, so that the problem of direct interception of a traditional long text abstract is solved; and the problem of multi-field text abstract extraction is solved.

The invention relates to a plant classification method based on sparse expression dictionary learning. The method is technically characterized by comprising the following steps: performing parameter initialization, wherein the parameter initialization comprises arranging the size, a sparse limitation factor and an error tolerance parameter of each plant type dictionary; for a training sample of each type of plant blade images, obtaining an over-complete dictionary of each type of blade images by use of a K-SVD algorithm; splicing the over-complete dictionary of each type of blade images after training to form a redundancy dictionary, and performing normalization processing on each column of the redundancy dictionary; obtaining a sparse coefficient through solving a minimum norm; and calculating residual errors and selecting a corresponding sample type with a minimum difference as a final identification result of a sample to be identified. According to the invention, the over-complete dictionaries are solved by use of type-based dictionary learning and sparse representation of an image to be identified is calculated, such that the calculation time of an algorithm is reduced, the requirement for real-time performance is met, the obtained identification rate is quite high, and the average identification rate is as high as more than 95%.

The invention discloses a robust remote sensing image feature point pair extraction method for nonlinear radiation distortion. The method comprises the following specific steps: carrying out image preprocessing on two acquired multi-modal remote sensing images; respectively extracting accurate angular points and edge feature points of the master image and the slave image, and obtaining feature points of each image; generating feature vectors of the feature points; registering the feature vectors of the feature points of the multi-modal image; further screening all candidate matching point pairs by using an RANSAC algorithm; and finally, obtaining accurate and uniform feature matching point pairs which are robust to nonlinear radiation distortion. A phase consistency measurement moment diagram robust to nonlinear radiation distortion is used as an object for feature point extraction, a large amount of accurate and robust feature point information can be obtained, and a mode of extracting Harris corner points by adopting image blocks is adopted, so that excessive concentration of image feature point distribution caused by non-uniform illumination radiation is effectively avoided, and the accuracy of image feature point extraction is improved. And the feature extraction efficiency is low.

The invention discloses a method for detecting mycotoxin in fermented tea. The method comprises dissolving a fermented tea sample in a polar solution, carrying out salting-out, carrying out centrifugation, carrying out filtration, putting the filtrate in 1290 / 6460 ultra-high performance liquid chromatography-triple quadrupole mass spectrometer, carrying out analysis through a mobile phase (0.1% formic acid-5 mmol / L ammonium acetate-water)-methanol, wherein the mobile phase A is 0.1% formic acid-5 mmol / L ammonium acetate-water and the mobile phase B is methanol, and carrying out separation through a gradient elution method comprising improving a concentration of the mobile phase B step by step so that the mycotoxin in the fermented tea sample reaches to the base line in short time and then is separated. The method realizes effective extraction and determination of various mycotoxins in fermented tea, can detect a variety of mycotoxins at the same time and has the characteristics of high detection efficiency, low environmental pollution, simple operation and promotion and use easiness.