518results about How to "Accurate extraction" patented technology

The present invention relates to an entity relationship recognition method and apparatus. The method comprises obtaining a statement sequence from a target text in a corpus, and performing named entity recognition and dependency grammar marker on the statement sequence to obtain a marked text sentence; matching and retrieving the marked text sentence on basis of an entity relationship seed to obtain a training example; replacing the entity relationship seed word in the training example with predetermined identification, processing the training example after replacement combined with the named entity recognition and the dependency grammar marker, and generating a candidate rule; fuzzifying the candidate rule to obtain fuzzy rules; determining whether the fuzzy rules comprise a new rule; and retrieving the corpus according to the fuzzy rules to obtain a seed set when the fuzzy rules comprise the new rule, and using the obtained seed set as an entity relationship recognition result. Manual participation can be effectively reduced, dependence on the calibrated corpus is reduced, a new entity relationship can be found timely, and the entity relationship recognition method and apparatus are self-adaptive to entity relationship mining in different fields.

Disclosed is a multicast network system including a multicast network which provides a plurality of relay nodes, and a monitoring server and a plurality of acknowledgment agents connected to the multicast network. The monitoring server includes an acknowledgment request message generating and sending unit that generates and sends an acknowledgment request message storing an acknowledgment probability to the plurality of acknowledgment agents; a collecting unit that collects acknowledgment messages from acknowledgment agents; and a collection number comparing unit that compares the number of acknowledgment messages collected by the acknowledgment message collecting unit with the expected number of acknowledgment messages obtained from the acknowledgment probability and the total number of acknowledgment agents. The acknowledgment agent has an acknowledgment message sending unit that determines whether an acknowledgment is needed for the acknowledgment request message based on the acknowledgment probability stored in the acknowledgment request message to send an acknowledgment message.

There are provided a watermark information embedding device and a watermark information detection device which can correctly extract classified information. In the watermark information embedding device, a plurality of dot patterns the wave directions and/or wavelengths of which are changed depending on dot arrangements are prepared, one symbol is given to one of the dot patterns, and the dot patterns are combined and arranged, so that the classified information is given. The watermark information detection device includes two-dimensional wavelet filters the number of which is equal to the number of dot patterns and which have the same wave directions and the same wavelengths as those of the dot patterns. Convolutions between an arbitrary region in a watermarked image and the plurality of two-dimensional wavelet filters are calculated, and it is determined that the dot pattern corresponding to the two-dimensional wavelet filter having the maximum convolution is embedded in the region.

The automatic fingerprint distinguishing method based on template learning includes the following steps: registering fingerprint figure and recording the fingerprint template and characteristic information of the fingerprint figure to be matched; distinguishing fingerprint figure to judge whether the characteristic information of the fingerprint figure is the same as that of some template information in the system data base; feeding back detail information in fingerprint figure via knowledge rule to perfect corresponding reliable detail points in templates. The present invention classifies detail points of fingerprint templates, records the kernel of classes, interval and other information and has one coarse matching link before common matching. The present invention can extract and repair the characteristic information of multiple templates, raise the distinction effect and performance of automatic fingerprint distinguishing system, and has important application in biological distinction.

An average face-image is created from images belonging to a face-image group and sample points are specified on the average face-image. Sample points are detected on images in the face-image group corresponding to sample points on the average face-image thereby creating recognition subspaces of feature vectors. Sample points are detected on registered images and an input image corresponding to the sample points on the average face-image by using the recognition subspaces. A degree of similarity between the feature vectors in the recognition subspaces of the registered images and the input image is calculated at the detected sample points. Finally, a person is verified based on the degree of similarity.

The invention discloses an extracting system of net page, which is characterized by the following: comprising document analyzer, classified word module and definition element treating module; extracting net page header and text with different projecting display attribute from the net page source file with the document analyzer; proceeding classified vocabulary for net page text, header and classified information with the classified module; getting the first vocabulary meter; converting the vocabulary of the first word meter to definition element with the definition element treating module; calculating weight of the definition element; proceeding reversal deacidize for the definition element; getting theme word collection. This invention also discloses an extracting method of net page theme. This invention avoids the puzzle of word rarefaction and information lost.

The invention relates to a table recognition method and device, a computer device and a storage medium. The method comprises the steps of obtaining a target document of which the document format is aPDF format; determining a table area where table content in the target document is located through a pre-trained table positioning model; cutting table content in the table area from the target document, and generating a corresponding table picture according to the table content; performing image recognition on the table picture, and determining characters in the table picture and position information of the characters; and generating a corresponding table file in a preset format according to the character and the position information. By adopting the method based on the image detection technology, the table data can be accurately extracted from the PDF document.

The invention discloses a method and a device for extracting webpage content and relates to the technical field of information. The method and the device for extracting webpage content can accurately extract webpage title and various elements in webpages when extracting the webpage content. The method includes converting HTML (hypertext markup language) source codes into corresponding document tree structures, and determining webpage titles according to TITLE tags of the document tree structures; determining webpage elements in webpages according to the webpage titles; determining attributes of various text blocks according to the webpage titles and density and word number of the text blocks of the document tree structures; and extracting the webpage titles, the webpage elements and the text blocks taking the webpage content as text, and acquiring full texts of the webpages, wherein the webpage elements at least include website LOGO, page navigation, news publishing time and news sources. The method and the device for extracting the webpage content are applicable to webpage content extraction.

A precise measuring method for the static image of cutting tool includes such steps as locating the precise coordinate position of the margin of said static image by digital image processing method, calculating the precise geometric parameters of cutting tool, and obtaining the envelope of said cutting tool. Its advantages are high percision and anti-interference power, and high locating correctness.

A SAW mode sensor for sensing parameters such as temperature, pressure, and strain. The sensor is made of a piezoelectric crystal cut at selected angles, with an attached electrode layer with a signal receiver and signal transmitter. The signal receiver initiates a wave in the substrate which propagates in the substrate and the speed of the wave and amplitude of the wave is interpreted as the parameter being sensed.

An image of a tissue site including a blood vessel is captured while broadband light and narrow band light are emitted to a body cavity. Thereby, broadband image data corresponding to the broadband light and narrow band image data corresponding to the narrow band light are obtained. Picture elements in the same positions are identified between the broadband image and the narrow band image to obtain a brightness ratio LM therebetween. Based on depth correlation information between the brightness ratio and blood vessel depth, a blood vessel depth D corresponding to the brightness ratio is obtained to determine whether each picture element includes a blood vessel and whether the blood vessel depth D is at the surface. Based on the determination, a surface blood vessel region is extracted. A broadband image having the surface blood vessel region with the reduced contrast is generated.

The invention relates to method for counting the pedestrian flow, in particular to a method for counting the pedestrian flow from multiple views under the complex scene of a traffic junction, comprising the following steps: carrying out background modeling to videos by adopting the hybrid code book algorithm, and detecting and segmenting the moving targets; calculating the movement vector by adopting the optical flow method to obtain the speed and the direction of the moving targets, distinguishing vehicles and pedestrians according to the speed and the direction of the moving targets, eliminating the vehicle targets, extracting the pedestrian targets, and calculating the direction of the pedestrians; setting a virtual door, and counting the pedestrian targets which enter and exit the virtual door according to the direction of the pedestrians and the predetermined direction of the pedestrians entering and exiting the virtual door. The method is suitable for counting the pedestrian flow from multiple views under the complex scene of the traffic junction by detecting the moving targets, extracting the pedestrians and counting the numbers of the pedestrians entering and exiting the virtual door in real time by adopting the virtual gate technology, has the counting accuracy of 90% and can adapt to the complex situation that the illumination of the outdoor environment is changeable, the shooting is carried out under changeable inclined angles, the human body resolution is low and the vehicles and the pedestrians run together.

The invention provides a foreground segmentation method applied to mobile communication network transmission and belongs to the field of image processing and communication network transmission. Five background templates with the shooting time intervals to be consistent with image acquisition intervals in the case of operation are pre-stored, and the former four background templates are used for fine detection in the case when a target is detected coarsely. The difference region area between the monitored image and the background image is calculated to judge whether a target enters, the image does not need to be transmitted when the target does not exist, the target region is segmented when the target exists, and the target region is only transmitted. The received target region is fused in the pre-stored background image by a server for video playing or image presentation. In a shooting mode, in order to prevent a passerby from being judged to be an obstacle, the target is set to be transmitted to the server only when the target is extracted by a third party according to the scene length, the human walking speed and the monitoring shooting frequency. Background replacement can be timely carried out after each time of segmentation cycle acquisition is finished, and interference caused by light and other environmental changes can be overcome in a slow transition way.

The invention discloses a discrete picture file information extraction system and method based on deep learning, an end-to-end OCR model combines text detection with text recognition, the sharing of an image processing layer of a convolutional neural network is achieved, and the calculation speed and recognition effect are improved. Two-way LSTM is used in the aspect of character detection, a textarea is accurately segmented according to file layout when characters are recognized, and the problems that noise can be recognized as the characters through an existing OCR technology, messy codes can be generated for a chart, and wrong lines can be generated can be solved. In the aspect of text recognition, a ResNet is combined with a Transformer codec, so that written bodies and handwritten bodies can be accurately recognized; by combining an end-to-end OCR model with an information extraction model, information loss can be reduced, and the accuracy of information extraction can be improved; and the Bayesian optimization algorithm is used to realize training and parameter adjustment automation of the information extraction model, so that a user can customize the desired information extraction model even if the user does not understand machine learning.

The invention provides an oil gas prospecting method and system, and the method comprises the steps of: using ground spectrometer to measure ground object spectrum data in study region; using digital camera to collect ground object description data in the study region; collecting ground surface sample data and measuring condition data; using portable GPS to collect measured ground object coordinate data in the study region; according to the ground object spectrum data, ground object description data, ground surface sample data, measuring condition data and measured ground object coordinate data to generate ground object spectrum database in the study region; using satellite or airplane-carried high spectrum imager to collect ground object high spectrum remotely sensed image data in the study region; processing the data in the ground object spectrum database and ground object high spectrum remotely sensed image data to obtain oil gas prospecting target region data.

The invention relates to the technical field of mode identification and intelligent vehicles, in particular to a vehicle identification method based on depth information. The method is used for an automobile auxiliary driving system. The vehicle identification method particularly comprises the steps of consecutively acquiring the depth information of objects in a scanning range through a distance measuring sensor; clustering each frame of data through the method based on the distance, wherein the threshold of the distance can be automatically adjusted through the distance of the targets and the vehicle and relative positions so that accuracy of clustering is improved, and good base is provided for subsequent feature extraction; respectively conducting straight line fitting on reserved clusters through the iterative adaptation point IEPF algorithm and the least square method, and extracting the eigenvector of the targets; and judging whether the targets are vehicle targets according to the eigenvector. The method identifies the vehicle targets through a plurality of characteristics so as to be capable of accurately identifying the vehicle targets in a road environment.

The invention discloses a hyperspectral abnormal object detection method based on structure sparse representation and internal cluster filtering, aiming at addressing the technical problem of low object detection effciency of current hyperspectral abnormal object detection methods. The technical solution involves: after selecting an initial background pixel, using the dictionary learning method which is based on principal component analysis to study a background dictionary which obtains rebustness, in the course of sparse vector resolution and image reconstruction, introducing re-weighted laplacian prior to increase the solution precision of sparse vector, computing the errors betwen an original image and a reconstructed image to obtain a sparse representation error, using the internal cluster filtering to represent space spectrum characteristics of hyperspectral data, obtaining the internal cluster error by computing the error between a to-be-tested pixel and other pixel linear representation result, and finally combining the sparse representation error and the linear weighting of the internal cluster error and implementing precise extraction of an abnormal object. According to the invention, the method increases 10-15% of detection rate with the proviso of a constant false alarm rate compared with prior art.

The invention relates to an online monitoring device, a detection circuit and a testing method for a junction temperature of an electronic power switch. The online monitoring device comprises a master control unit, a to-be-tested power electronic device, a current detection unit, a Vce(on) detection unit and an AD collection unit, wherein the master control unit comprises a device driving unit and a junction temperature calculation unit power electronic switch device, the current detection unit is used for collecting a current Ic flowing through the device in a conducting state; the Vce(on) detection unit is used for collecting a conducting pressure drop Vce of an IGBT in a certain rated time period in a conducting state and obtaining an equivalent conducting pressure drop signal Vce(on) by virtue of an analogue circuit in the Vce(on) detection unit; the AD collection unit is used for finishing the conversion from simulation current and voltage to digital signals; the device driving unit is used for providing a gate electrode driving signal Vg of a to-be-tested device and a switch signal Vp(mos) of the Vce(on) detection unit; and an IGBT driving unit is used for regulating a level fluctuation sequential relationship between a gate electrode switch control signal Vg of an IGBT module and the switch signal Vp(mos) of the Vce(on) detection unit in the IGBT driving unit.

A finger/palm print image processing system includes a frequency component analysis unit (11) and a frequency component judgment unit (12). The frequency component analysis unit (11) performs a frequency analysis for each of the small regions into which the finger/palm print image is divided and acquires frequency components representing the respective small regions. According to frequency component, the frequency component judgment unit (12) judges the clarity of the small region corresponding to the frequency component. The finger/palm print image indicates at least one of the finger print and the palm print. The Fourier transform is used for the frequency analysis. According to the frequency component and the result of the Fourier transform of a clear two-dimensional sinusoidal wave, the frequency component judge unit (12) judges the clarity of the small region corresponding to the frequency component. Thus, even when the background has a blocky portion or the finger/palm print region has an unclear portion due to a scratch, it is possible to judge a region where a clear ridgeline exists.

The invention discloses a method and a system for wireless channel measurement based on wavelet decomposition threshold de-nosing. The method includes the following steps: (1) using a spread spectrum sliding correlation method to finish wireless channel measurement so as to obtain channel impulse response containing noise, (2) conducting multi-resolution wavelet decomposition for the channel impulse response and leading the channel impulse response to be converted from time domain to wavelet domain, (3) reasonably selecting a threshold function and a threshold and processing wavelet coefficients corresponding to the noise according to the threshold function; (4) conducting wavelet reconstruction and leading the de-noised channel impulse response to be converted from the wavelet domain back to the time domain, and (5) storing the de-noised channel impulse response for researching channel characteristics. Through steps of wavelet decomposition, threshold de-nosing, wavelet reconstruction and the like, the method and the system can reduce interference on the channel impulse response caused by the noise, improve accuracy and effectiveness of channel measurement results, and provide reliable data basis for follow-up researching of channel characteristics.

The invention relates to a method for extracting a face characteristic based on a local three-value mode, belonging to the technical field of image processing. The method comprises the following steps of: firstly carrying out down sampling on an original face image to construct a face image pyramid, then calculating LTP (long-term potentiation) characteristics of all the pixel points in an image at each layer, and then dividing the LTP characteristics of all the pixel points into positive LTP sub characteristics and negative LTP sub characteristics, respectively carrying out statistics to obtain a positive LTP sub characteristic value histogram and a negative LTP sub characteristic value histogram of the image at each layer in the face image pyramid, and finally connecting vectors H+ and H- corresponding to the positive LTP sub characteristic value histogram and the negative LTP sub characteristic value histogram as the final characteristic of the original face image I (x, y). In the invention, the local characteristic of the LTP characteristic and the statistics characteristics of all the local LTP characteristic histograms are utilized to realize uniformity of the local characteristic and global characteristic, and the LTP characteristic has a better inhibition effect on noise compared with the LBP (local binary pattern) characteristic which is frequently adopted. The extracted characteristic has the characteristics of constant rotation and constant grey scale, and face characteristic can be accurately extracted under the influences that illumination condition is changed, face expression is changed and a gesture is changed.

The invention discloses a method for extracting webpage subject contents. The method comprises the following steps of: selecting latest RSS (Really Simple Syndication) information and a corresponding webpage from an RSS file; searching the position of the RSS information in a Dom Tree of the corresponding webpage, wherein the information at the position is taken as a webpage template; and extracting the subject contents of a plurality of target webpages by using the webpage template. The method further comprises the following step of: after the subject contents of a predetermined quantity of target webpages in the plurality of target webpages are extracted, regenerating the webpage template and continually extracting the subject contents of the plurality of target webpages.

The invention discloses an image processing method and electronic equipment. The image processing method is used for the electronic equipment with a first camera and a second camera; the first camera and the second camera are used for shooting the same scene; the image processing method comprises the following steps: controlling the first camera to shoot a first image, and controlling the second camera to shoot a second image; synthesizing the first image and the second image as a third image; computing depth information of pixels on the third image; determining a foreground image and a background image on the third image according to the depth information of the pixel; replacing the background image of the third image by use of a preset background image so as to obtain the third image with the replaced background. By use of the image processing method disclosed by the invention, the extraction of each of the foreground image and the background is more accurate, and the shot background image can be directly replaced by use of the preset background after obtaining the image through shooting, the post processing is unnecessary to perform, and the replacement of the background image is quickly realized.

The invention discloses a method for backfilling a number in signaling monitoring and a number backfill server. The method comprises the steps of: A, establishing a correspondence among an IMSI (International Mobile Subscriber Identification Number, a TMSI (Temporary Mobile Subscriber Identification and an MSISDN (Mobile Station International ISDN (Integrated Services Digital Network) Number) with an MSC (Mobile Switch Unit) as a unit by using a logic relationship among numbers in a plurality of signaling messages transmitted by interfaces at the periphery of the MSC, generation time of the signaling messages and all signaling messages, and recording updating time of the correspondence; and B, backfilling the number of the IMSI or MSISDN for the received signaling messages according to the correspondence and the updating time thereof, and storing the signaling messages after being subjected to number backfilling. According to the method and the number backfill server disclosed by the invention, the IMSI or MSISDN can be backfilled into the signaling messages to ensure that a signaling monitoring system can completely and accurately extract the signaling messages of a subscriber/a subscriber group through an actual number, i.e. the IMSI or MSISDN of a mobile subscriber.

The invention discloses a method and device of detecting human vital signs based on ultra-wideband radar. The method comprises the steps of acquiring multiple first echo signals of normal human breathand/or heartbeat and multiple second echo signals of abnormal human breath and/or heartbeat through ultra-wideband radar, preprocessing the signals, and extracting to obtain a normal detection sequence and an abnormal detection sequence; setting a convolutional neural network; training the convolutional neural network based on the normal detection sequence and the abnormal detection sequence; detecting breath and/or heartbeat of a current detection subject through the trained convolutional neural network, and outputting the results of whether the heartbeat is normal. The device herein comprises a signal processing module, a convolutional neural network setting module, a training module and a detection module. The method and device have the advantages that deep learning helps improve the accuracy of detecting heartbeat frequency, the capacity to distinguish normal heartbeat and abnormal heartbeat is improved, and real-time detection is distinguished accurately.

The invention relates to a clastic rock reservoir modeling method and device based on a digital geological outcrop. The clastic rock reservoir modeling method comprises the steps that the geological outcrop is selected according to known data of a target exploratory area; geological information of the geological outcrop is collected, and geological outcrop data bodies are obtained; the pictures of the outcrop in all the directions and at all the positions are obtained; actual measurement is conducted on the geological section of the geological outcrop, and a rock sample of the geological outcrop is obtained; analysis, assaying and measurement are conducted on the rock sample, and a rock physical property analysis graph is obtained; the geological outcrop data bodies are spliced, and the three-dimension data body, with uniform coordinates, of the geological outcrop is obtained; through the pictures of the geological outcrop in all the directions and at all the positions, the geological phenomena of the three-dimension data body of the geological outcrop are described; a three-dimension geological model of the geological outcrop is established according to the described geological phenomena of the geological outcrop and the actual measurement information of the geological section; according to the rock physical property analysis graph, a clastic rock reservoir physical property model is established through the three-dimensional geological model of the geological outcrop.

A sound separation apparatus for separating a target signal from a mixed input signal, wherein the mixed input signal includes the target signal and one or more sound signals emitted from different sound sources. The sound separation apparatus according comprises a frequency analyzer for performing a frequency analysis on the mixed input signal and calculating spectrum and frequency component candidate points at each time. The apparatus further comprises feature extraction means for extracting feature parameters which are estimated to correspond with the target signal, comprising a local layer for analyzing local feature parameters using the spectrum and the frequency component candidate points and one or more global layers for analyzing global feature parameters using the feature parameters extracted by the local layer. The apparatus further comprises a signal regenerator for regenerating a waveform of the target signal using the feature parameters extracted by the feature extraction means.

Since both of local feature parameters and global feature parameters can be processed together in the feature extraction means, the separation accuracy of the target signal is improved without depending on the accuracy for extracting feature parameters from the input signal. Feature parameters to be extracted include frequencies and amplitudes and their variation rates for the frequency component candidate points, harmonic structure, pitch consistency, intonation, on-set/off-set information and/or sound source direction. The number of the layers provided in the feature extraction means may be changed according to the types of the feature parameters to be extracted.

The invention is suitable for the field of medical image segmentation and provides a method and system for automatically segmenting bones from abdomen image data. The method comprises the steps of automatically extracting a first segmentation result comprising an abdomen aorta or part of an abdomen aorta area in the abdomen image data according to shape features of the abdomen aorta; according to the first segmentation result, modifying data comprising the abdomen aorta or part of the abdomen aorta area in the abdomen image data to obtain abdomen image modified data; automatically selecting at least one seed point of the bones from the abdomen image data; and according to the seed point and the abdomen image modified data, obtaining a second segmentation result of the bones by first threshold connection. Accordingly, bone areas can be segmented from an abdomen image automatically and accurately.

The invention discloses a human body motion video recognition method based on non-negative matrix factorization. The human body motion video recognition solves the problems that due to the fact that in the prior art, extraction of characteristics of a motion vehicle is influenced by the background environment, the recognition rate is lowered, and due to the fact that the number of the dimensions of the extracted characteristics is excessively high, the calculation amount is excessively large. The method specifically includes the steps of firstly, preprocessing a data set; secondly, detecting space interest points; thirdly, constructing a cube; fourthly, constructing a characteristic matrix; fifthly, training a dictionary; sixthly, conducting classification. According to the method, the influences of the background environment on characteristic extraction can be effectively eliminated in the human body motion video recognition process, the recognition rate of the human body motion video can be improved, the number of the dimensions of the extracted characteristics is low, and the calculation amount and the complexity in the human body motion video recognition process are reduced.

The invention relates to the technical field of copyright image protection, and provides a decentralized image copyright protection system and method with infringement real-time detection, and the method comprises the following steps of (1), enabling a registered user to upload an image through a client, and enabling an SIFT image feature extraction algorithm in the system to carry out the featuredata collection on the uploaded image; (2) extracting the invariant characteristics, such as visual angle, translation, affine, brightness, rotation, etc., of an existing copyright image by a user through an SIFT algorithm to form a local feature vector set of the image, and taking the local feature vector set as a registered image feature data set; (3) carrying out distributed storage on the feature data sets in the steps (1) and (2) by utilizing an IPFS (Internet Protocol File System), and constructing a network transmission protocol for persistent and distributed storage and file sharing;and (4) performing comparative analysis and calculation, and judging that the image does not accord with the condition of forming a new work. The system and the method provided by the invention have the characteristics of being tamper-proofing, realizing the high safety, permanent preservation and decentralization, being able to trace, being low in cost, etc.