Patents
Literature
Hiro is an intelligent assistant for R&D personnel, combined with Patent DNA, to facilitate innovative research.
Hiro

46results about How to "Relevant" patented technology

Method and system for implementing configuration on basis of network services

ActiveCN103634177AEfficient self-configuring configurationRapid self-configuration configurationNetworks interconnectionService flowStructure of Management Information
The invention discloses a method and a system for implementing configuration on the basis of network services. The method and the system have the advantages that network service configuration databases are created for users, topological structures, identification of service flows, identification of network devices for carrying the service flows and configuration information of physical ports, VLAN (virtual local area network) selection and the like can be selected from the databases for network topological contents and specific service contents on the basis of network service requirements, accordingly, service types, source MAC (media access control) addresses and target MAC addresses can be identified according to ports for the network devices for carrying the service flows, network topology can be creased on the basis of the identification of the service flows and the configuration information of the ports in the network service configuration databases, the same services can be transmitted in selected VLANs, and the configuration can be implemented; the configuration information can be issued to the various network devices, so that the complexity and the troublesome degree of the network configuration can be lowered, the specific services can be transmitted among the configured networks in an identifiable manner, and network configuration lists are high in reliability.
Owner:KYLAND TECH CO LTD

Combined-type anti-counterfeiting mark based on same ID, generating method and device

The invention provides a combined-type anti-counterfeiting mark based on a same ID. The combined-type anti-counterfeiting mark is composed of a quick response code and a raster graph. The quick response code and the raster graph all comprise the same ID number extracted from an ID number database during a generating process. The method for generating the combined-type anti-counterfeiting mark comprises the steps that by using a quick response code generator, the ID number, parameter information of the quick response code, plaintext information and preset ciphertext information memory area are converted into the quick response code, wherein the ID number, the parameter information of the quick response code and the plaintext information are extracted from the ID number database; by using a raster graph generator, the raster graph is generated by the ID number which is same as the quick response code. The device for generating the combined-type anti-counterfeiting mark comprises the quick response code generator and the raster graph pattern generator. The quick response code and the raster graph comprise the same ID number, and therefore the quick response code and the raster graph have relevancy, due to the fact that quick response code can only be recognized and read through dedicated equipment, the raster graph cannot be recognized once again after being copied and scanned, and therefore the high level anti-counterfeiting performance is achieved.
Owner:兰芳园食品有限公司

Callicarpa nudiflora medicine, intermediate and fingerprint detection method and standard fingerprint of preparation

The invention discloses a callicarpa nudiflora medicine, an intermediate and a fingerprint detection method of a preparation. The method comprises the following steps: (a) preparation of a reference solution; (b) preparation of a test solution; (c) chromatographic condition; (d) formulation of a standard fingerprint of isoacteoside as a reference peak; and (e) quality control of the fingerprint. According to the above method, a pretreatment method of each tested object is simple, characteristic constituents are retained completely, stability is good, precision is high, and there is certain specificity. The separation effect of each characteristic peak in the fingerprint is good. The callicarpa nudiflora medicine, the intermediate and the fingerprint of the preparation have good correlation. The method can be used for identifying quality of the callicarpa nudiflora medicine and increasing production process controllability of the callicarpa nudiflora preparation, and is helpful for guaranteeing quality stability and clinical efficacy of the callicarpa nudiflora preparation. The invention also discloses the callicarpa nudiflora medicine obtained according to the above fingerprint detection method, the intermediate and the standard fingerprint of the preparation.
Owner:JIUZHITANG +2

Transfer learning-based Chinese-Vietnamese neural machine translation method

The invention relates to a transfer learning-based Chinese-Vietnamese neural machine translation method, and belongs to the technical field of natural language processing. The method comprises the following steps of: corpus collection and preprocessing: collecting and preprocessing parallel corpora of Chinese-Vietnamese sentence pairs, English-Vietnamese sentence pairs and Chinese-English sentencepairs; generating Chinese-English-Vietnamese trilingual parallel corpora by using the Chinese-English and English-Vietnamese parallel corpora; training a Chinese-English neural machine translation model and an English-Vietnamese neural machine translation model, and initializing parameters of the Chinese-Vietnamese neural machine translation model by using parameters of the pre-trained model; andcarrying out fine adjustment training on the initialized Chinese-Vietnamese neural machine translation model by using the Chinese-Vietnamese parallel corpus to obtain the Chinese-Vietnamese neural machine translation model to carry out Chinese-Vietnamese neural machine translation. According to the method, the performance of the Chinese-Vietnamese neural machine translation can be effectively improved.
Owner:KUNMING UNIV OF SCI & TECH

Iterative search optimization and satisfaction degree promotion method and system based on user click

The invention provides an iterative search optimization and satisfaction degree promotion method and system based on user clicking and belongs to the searching optimization field. The iterative search optimization and satisfaction degree promotion method based on the user clicking comprises step A, extracting clicked pages of a keyword from a query log according to a search keyword, extracting and clicking keywords of the pages from the query log, repeating the above process until convergence, building the keywords and corresponding clicked pages into a bipartite graph, performing the bipartite graph as input and performing a similarity iterative calculation according to an IA-CLICK (Iterative Algorithm) method; step B, reordering search results according to an obtained relevance of query keywords provided by a user and the pages through a relevance R (q,d) measurement formula based on a page similarity assemble obtained from the step A; and step C, performing a search evaluation according to a clicking situation of reordered search results of users and a user satisfaction US measurement formula. The iterative web search optimization and satisfaction degree promotion method and system based on the user clicking has the advantages of enabling user burden to be small, being strong in relevance and simple and efficient in evaluation frame and enabling reality and flexibility of method effects to be improved.
Owner:CHONGQING UNIV

Combined anti-counterfeiting mark based on same ID (Identity) as well as generation method and device

The invention provides a combined anti-counterfeiting mark based on the same ID (Identity). The combined anti-counterfeiting mark is composed of a two-dimension code and a dot-matrix graphic; a generation process of each of the two-dimension code and the dot-matrix graphic is a method for extracting the same ID number from an ID number database, and generating the combined anti-counterfeiting mark, and comprises the following steps: converting the ID number extracted from the ID number database, and parameter information, clear text information and cipher text information of the two-dimension code a into the two-dimension code; and generating the dot-matrix graphic smaller than the two-dimension code by the ID number the same as the two-dimension code through a dot-matrix graphic generator. A device for generating the combined anti-counterfeiting mark comprises a two-dimension code generator and the dot-matrix graphic generator. The two-dimension code and the dot-matrix graphic, provided by the invention, have the same ID number, so that the two-dimension code and the dot-matrix graphic have the relevance; the two-dimension code only can be read by special equipment, and the dot-matrix graphic cannot be identified again after being copied and scanned, so that the combined anti-counterfeiting mark has a high-grade anti-counterfeiting performance.
Owner:LEADER TECH BEIJING DIGITAL TECH

Intelligent self-calibration method for temperature and humidity of overlapped blocks of gas sensor

The invention relates to the field of gas sensor measurement, in particular to a gas sensor overlapped block temperature and humidity intelligent self-calibration method, which comprises the following steps: S1, measuring the resistance of a gas sensor under different temperature, humidity and concentration conditions; s2, acquiring a change rule of response resistance and each variable; s3, roughing abnormal points; s4, overlapping and blocking the data according to the change rule of the data along with temperature, humidity and concentration; s5, performing multiple linear regression on each data block; s6, repeatedly and finely selecting abnormal points and iteratively correcting the abnormal points; s7, performing multiple linear regression on the corrected data block; s8, establishing a temperature and humidity compensation expression of the concentration through inverse operation; s9, importing a plurality of compensation expressions of the same kind of sensors into the microcontroller, obtaining a plurality of predicted values, and taking the average value of all the predicted values as the calibration result of the gas sensor. According to the method, the temperature and humidity real-time intelligent self-calibration of the gas sensor is realized.
Owner:UNIV OF ELECTRONICS SCI & TECH OF CHINA

A pedestrian re-identification method based on deep learning and overlapped image inter-block measurement

The invention relates to a pedestrian re-identification method based on deep learning and measurement between overlapped image blocks. The pedestrian re-identification method comprises the following steps of 1) constructing a deep learning neural network model based on a plurality of overlapped feature images; 2) constructing a model training loss function based on the inter-block measurement of the overlapped image; 3) training the model by using the training sample; 4) inputting the pedestrian image to be identified and all comparison images into the model to obtain image features; 5) obtaining a final distance between the pedestrian image to be identified and the comparison image by using a Euclidean distance calculation formula, and 6) sorting the distance to obtain a comparison imagelibrary matching sequence corresponding to the pedestrian to be identified. Compared with the prior art, the method has the advantages of high accuracy, good robustness and the like.
Owner:TONGJI UNIV

Extremely weak optical signal detecting device and method

The invention belongs to the field of signal detection, and particularly relates to a high-precision extremely weak optical signal detecting device and method based on a DSP. The device comprises a mechanical shutter, a photomultiplier, a low-noise preamplifier, a high-pass filter circuit, a high-resolution A / D conversion circuit, a DSP processing circuit, a shutter drive circuit and a computer. Processing of modulation, background noise elimination and digital sampling integration is carried out on extremely weak optical signals through the device, and precise detection on extremely weak incident light is achieved. Meanwhile, the extremely weak optical signal detecting device is low in production cost, simple in structure and convenient to use.
Owner:XI'AN INST OF OPTICS & FINE MECHANICS - CHINESE ACAD OF SCI

Spacecraft three-dimensional process display and production data acquisition system based on portable terminal

The invention discloses a spacecraft three-dimensional process display and production data acquisition system based on a portable terminal. The spacecraft three-dimensional process display and production data acquisition system mainly comprises a manufacturing execution system module, an offline processing module and a practical digital prototype, and the offline processing module comprises an instrument process file and process model association display module, an instrument installation execution record signing module, an instrument state association type photographing module and an instrument state data and picture return module. According to the invention, the problems of associated viewing of process models and process files in a production field and discrete storage and application of photos are solved.
Owner:BEIJING INST OF SPACECRAFT ENVIRONMENT ENG

Power grid dispatching comprehensive monitoring MIS platform

The invention discloses a power grid dispatching comprehensive monitoring MIS platform which comprises a running section and a dispatching section. The dispatching section comprises a first exchanger, an MIS database and a dispatching operation end. The running section comprises a second exchanger, operation and monitoring devices, an SCADA system host, a background control end and a foreground monitoring end. The first exchanger is connected with the second exchanger through the internet, the number of the operation and monitoring devices is a lot, and the operation and monitoring devices are connected with the second exchanger. The SCADA system host, the background control end and the foreground monitoring end are connected with the second exchanger, and the MIS database and the dispatching operation end are connected with the first exchanger. According to the power grid dispatching comprehensive monitoring MIS platform, the integration degree of a power generation MIS platform can be improved, and accordingly a rational and simplified hardware integration system is provided for achieving power grid dispatching comprehensive monitoring.
Owner:STATE GRID CORP OF CHINA +1

Life cycle evaluation and fault early warning method for high-voltage circuit breaker

The invention relates to the technical field of power equipment maintenance, in particular to a high-voltage circuit breaker life cycle evaluation and fault early warning method which comprises the following steps: A) acquiring detection data; b, setting a fault source, enabling the high-voltage circuit breaker to be continuously switched on and off until a fault occurs, enabling the switching-onand switching-off action times to be N, and carrying out the periodic detection; c) obtaining a life cycle evaluation result; d) training a fault early warning neural network model; and E) inputting the detection data of the high-voltage circuit breaker to be evaluated and early warned into the fault early warning neural network model, if the fault early warning neural network model outputs a fault type, sending out fault early warning, and the corresponding fault is the fault type output by the fault early warning neural network model. The method has the substantive effects that the representation of the life cycle of the high-voltage circuit breaker is obtained, so that the maintenance of the high-voltage circuit breaker is more targeted, and the working stability and reliability of thehigh-voltage circuit breaker are improved.
Owner:STATE GRID ZHEJIANG ELECTRIC POWER CO LTD JIAXING POWER SUPPLY CO

Visualization enhancement method based on Monte Carlo particle transport simulation physical value characteristics

The invention discloses a visualization enhancement method based on Monte Carlo particle transport simulation physical value characteristics. The method comprises the following steps of carrying out differentiation characteristic visualization different from grids where other physical values are positioned on the minimum value and the maximum value of physical values in a Monte Carlo particle transport simulation result to realize the visualization enhancement on the grids where extreme values are positioned; setting the transparency of the grid where a calculation error interval is positioned and which is appointed by a user to be 0 according to calculation errors in all grids in the Monte Carlo particle transport simulation result, and setting the transparency of the grid beyond the interval to be 1 to realize virtual enhancement on the calculation grid result of the appointed error interval. The visualization enhancement method can help the user to quickly position calculation grid areas concerned in neutronics analysis, so that the efficiency of the neutronics analysis is promoted.
Owner:HEFEI INSTITUTES OF PHYSICAL SCIENCE - CHINESE ACAD OF SCI

Time series correlation detection method and device, and storage medium

The invention provides a time series correlation detection method and device and a storage medium. According to the invention, the method includes using a sliding window for carrying out moving average processing on two time sequences; obtaining two smooth time sequences; and then respectively carrying out loop iteration on the two smooth time sequences to execute an interleaving translation correlation calculation processing step, and then determining the maximum value in the interleaving correlation coefficients calculated in the interleaving translation correlation calculation processing step as two time sequence correlation coefficients so as to determine the correlation of the two sequences. According to the invention, a sliding window method is used for being matched with the maximuminterleaving correlation coefficient obtained in the interleaving translation correlation calculation processing step, and the correlation between time sequences with peak shifting correlation can bedetected. And the influence of the noise of the time sequence on the maximum cross mutual information coefficient is greatly reduced by using a sliding average method.
Owner:NEW H3C TECH CO LTD

Mobile terminal, mobile terminal contact information arrangement processing method, and storage device

The present invention discloses a mobile terminal, a mobile terminal contact information arrangement processing method, and a storage device. In the method, a relation between contacts is established and is shown by a two-dimensional array; the two-dimensional array is shown as K[x][y], wherein [x][y] shows the number of times of a contact with a number x starting conversation with a contact with number y as the object of the conversation in the first time quantum after the conversation is end so far; and based on the two-dimensional array K[x][y], when it is detected that the contact with the number x finishes the conversation, contacts with large number y in the K[x][y] are displayed in order on the screen of the mobile terminal to allow the information of the contacts of the mobile terminal to have a certain relevance according to the actual call to add a new function for a current mobile terminal: the contacts of the mobile terminal can always put continuously calling contacts together when calling, so that convenience is provided for users.
Owner:TCL COMM TECH NINGBO CO LTD

Feature point elimination method based on correlation

The invention discloses a correlation-based feature point elimination method. The method is used for visually guiding an assembling / grabbing process and pre-acquiring standard coordinates and theoretical coordinates; acquiring measurement coordinates, and utilizing a first projection error; judging whether wrong feature points need to be removed or not, wherein the method for searching the wrong feature points comprises the following steps: sorting the plurality of feature points; removing the ith feature point in sequence according to the sequence, calculating a standard reprojection error array corresponding to the ith feature point, and measuring the reprojection error array; calculating a correlation coefficient between the standard reprojection error array corresponding to the ith feature point and the measurement reprojection error array; finding an abnormal point from a plurality of correlation coefficients, and then finding an error feature point from the abnormal point; The method is advantaged in that two arrays obtained through the method are high in relevancy, and amplitudes of all elements in the set are close; therefore, accuracy of the remaining feature points is guaranteed, and assembly / grabbing accuracy is improved.
Owner:易思维(杭州)科技有限公司

Grid document processing method and processing apparatus thereof

The invention provides a method used to process files in a grid file system, which comprises the steps that: the grid file system acquires metadata used to represent information NStruc of a host user file mount point according to identity information GUInfo of a grid user; the grid file system is subject to the initialization operation, and a metadata table is constructed; a file and the host user file mount point of the file are confirmed through virtual route transformation according to a file operating command; according to information in the metadata table, the grid user is mapped to a host user nU on the host user file mount point of the file; a host searches a corresponding host file according to a file operating command submitted by the host user nU; and the host file is processed according to the file operating command. The method shields the discrepancy among different hosts and / or different file systems through the virtualized representation to the grid file and the grid user-oriented design, and provides convenient and unified grid file processing operation under the condition of not modifying a host file system.
Owner:INST OF COMPUTING TECH CHINESE ACAD OF SCI

Network live broadcast intelligent guiding method and system, and computer storage medium

PendingCN114630186AAchieve the expected live broadcast effectAchieve live broadcast effectSpeech recognitionSelective content distributionComputer networkEngineering
The invention discloses a network live broadcast intelligent guiding method and system and a computer storage medium, and the method comprises the steps: obtaining each preset process step in a target network live broadcast room and the real-time voice content of an anchor, and analyzing a matching degree coefficient between the voice content of the anchor in the target network live broadcast room and each preset process step; according to the method, the matching process step corresponding to the anchor voice content in the target network live broadcast room is screened, the matching process step is compared with the target process step being executed in the target network live broadcast room, and the corresponding operation instruction is sent according to the comparison result, so that the live broadcast process of the target network live broadcast room is intelligently guided. According to the method, the function limitation of the existing method is broken to a great extent, so that the live broadcast content of the network live broadcast room has higher relevance and logicality, the interest and expectation of live broadcast audiences are further improved, the user viscosity of a network live broadcast platform is increased, and the development and progress of the network live broadcast platform are facilitated.
Owner:武汉叶子互娱文化传媒有限公司

Illegal account identification method and device, readable medium and electronic equipment

PendingCN112182520ARelevantRelevance fully consideredDigital data authenticationEngineeringData mining
The embodiment of the invention provides an illegal account identification method and device, a computer readable medium and electronic equipment. The illegal account identification method comprises the steps of obtaining a suspicious account; grouping the suspicious accounts based on the account information of the suspicious accounts to obtain a suspicious account set; calculating a similarity value between at least two suspicious accounts in the suspicious account set based on the account information of the suspicious accounts in the suspicious account set; and calculating a score of the suspicious account set according to a similarity value between at least two suspicious accounts in the suspicious account set, so as to identify an illegal account set according to the score of the suspicious account set. According to the technical scheme of the embodiment of the invention, grouping is carried out according to the account information of the suspicious accounts to obtain the suspicious account set, the score of the suspicious account set is obtained by calculating the similarity value between the suspicious accounts, illegal account identification is carried out, and the identification accuracy and the processing efficiency are improved.
Owner:TENCENT TECH (SHENZHEN) CO LTD

Combined anti-counterfeiting mark based on the same id, and generation method and device

The invention provides a combined anti-counterfeiting mark based on the same ID (Identity). The combined anti-counterfeiting mark is composed of a two-dimension code and a dot-matrix graphic; a generation process of each of the two-dimension code and the dot-matrix graphic is a method for extracting the same ID number from an ID number database, and generating the combined anti-counterfeiting mark, and comprises the following steps: converting the ID number extracted from the ID number database, and parameter information, clear text information and cipher text information of the two-dimension code a into the two-dimension code; and generating the dot-matrix graphic smaller than the two-dimension code by the ID number the same as the two-dimension code through a dot-matrix graphic generator. A device for generating the combined anti-counterfeiting mark comprises a two-dimension code generator and the dot-matrix graphic generator. The two-dimension code and the dot-matrix graphic, provided by the invention, have the same ID number, so that the two-dimension code and the dot-matrix graphic have the relevance; the two-dimension code only can be read by special equipment, and the dot-matrix graphic cannot be identified again after being copied and scanned, so that the combined anti-counterfeiting mark has a high-grade anti-counterfeiting performance.
Owner:LEADER TECH BEIJING DIGITAL TECH

Combined anti-counterfeiting mark based on the same id, and generation method and device

The invention provides a combined-type anti-counterfeiting mark based on a same ID. The combined-type anti-counterfeiting mark is composed of a quick response code and a raster graph. The quick response code and the raster graph all comprise the same ID number extracted from an ID number database during a generating process. The method for generating the combined-type anti-counterfeiting mark comprises the steps that by using a quick response code generator, the ID number, parameter information of the quick response code, plaintext information and preset ciphertext information memory area are converted into the quick response code, wherein the ID number, the parameter information of the quick response code and the plaintext information are extracted from the ID number database; by using a raster graph generator, the raster graph is generated by the ID number which is same as the quick response code. The device for generating the combined-type anti-counterfeiting mark comprises the quick response code generator and the raster graph pattern generator. The quick response code and the raster graph comprise the same ID number, and therefore the quick response code and the raster graph have relevancy, due to the fact that quick response code can only be recognized and read through dedicated equipment, the raster graph cannot be recognized once again after being copied and scanned, and therefore the high level anti-counterfeiting performance is achieved.
Owner:兰芳园食品有限公司

A phase screen simulation method for light propagating in short distance in underwater turbulent flow

The invention relates to a phase screen simulation method of short distance transmission of light in underwater turbulent flow, and belongs to the field of underwater target detection. The phase screen simulation method of short distance transmission of light in underwater turbulent flow is provided for solving the shortcomings that a phase screen model cannot really simulate the properties of turbulent flow as the correlation between phase screens in multiple layers of phase screens is not considered in the prior art. The method comprises the following steps: inputting an initial image; calculating a pixel value of a current reference point by using a reference point with a known value; calculating a relative pixel difference; calculating an absolute value of the reference point; repeating the above process until left and right reference points of all phase screens are calculated; generating multiple processes, wherein each process contains one framework; simultaneously calculating the absolute value of each point in the framework by using a parallel calculation method; repeating the process of calculating the absolute value of each point of each process; and repeating the above steps until all points are calculated. The phase screen simulation method provided by the invention is suitable for the analogue simulation of a transmission model of light in underwater turbulent flow.
Owner:HARBIN INST OF TECH

Zero-knowledge proof circuit optimization method and device, terminal equipment and storage medium

The invention provides a zero-knowledge proof circuit optimization method and device, terminal equipment and a storage medium. The method comprises the following steps: constructing an abstract syntax tree of an initial zero-knowledge proof circuit; optimizing an initial circuit constraint of the initial zero-knowledge proof circuit according to the abstract syntax tree; and carrying out association classification on the optimized initial circuit constraints to obtain circuit constraints of different classifications of the zero-knowledge proof circuit, and running the circuit constraints of different classifications in parallel in the zero-knowledge proof circuit. According to the zero-knowledge proof circuit optimization method provided by the invention, original circuit constraints in the zero-knowledge proof circuit can be optimized, some unnecessary circuit constraints in the zero-knowledge proof circuit are eliminated, and the possibility of circuit constraint errors is reduced. Meanwhile, the optimized circuit constraints are correspondingly associated, so that the associated circuit constraints have context relevance, and the operation efficiency is improved.
Owner:HANGZHOU QULIAN TECH CO LTD

Action determination method and device, readable medium and electronic equipment

The invention relates to an action determination method and device, a readable medium and electronic equipment, and relates to the technical field of machine learning, and the method comprises the steps: obtaining an environment image of an environment where a target object is located and an object feature vector of each environment object in the environment image, determining the position information of each environment object in the environment image, and for each environment object, splicing the object feature vector of the environment object to a position corresponding to the position information of the environment object, and taking the spliced environment image as input of a pre-trained action model to obtain a target action to be executed by the target object output by the action model. According to the method, the object feature vectors are spliced to the positions corresponding to the position information, so that the object feature vectors and the position information have relevance, the action model can quickly extract important features from the spliced environment image, and the accuracy of the target action output by the action model is improved.
Owner:BEIJING BYTEDANCE NETWORK TECH CO LTD

Fire-fighting gas cylinder monitoring system and equipment matching method

The invention discloses a fire-fighting gas cylinder monitoring system and an equipment matching method. The fire-fighting gas cylinder monitoring system is characterized by comprising fire-fighting gas cylinders, pressure gauges, mobile matching equipment and a monitoring terminal, the fire-fighting gas cylinders are arranged in a distributed manner, the pressure gauges and the fire-fighting gascylinders are correspondingly arranged, two-dimensional codes are attached to the equipment, data information such as the pressure state is transmitted to the monitoring terminal through bus connection with control cabinets or base station wireless transmission, and the mobile matching device is used for scanning the two-dimensional codes on the equipment to associate and pair the fire-fighting gas cylinders and the pressure gauges and transmitting updated data to the monitoring terminal. According to the fire-fighting gas cylinder monitoring system, the monitoring and management of large-scale fire-fighting gas cylinders are realized, the maintained or replaced equipment is matched by scanning the two-dimensional codes, the data of the paired equipment is updated, the working intensity ofrelated personnel is reduced, and the management efficiency is improved.
Owner:VITALONG FIRE SAFETY GRP

Face quality evaluation method and system based on random embedding stability

The invention discloses a face quality evaluation method and system based on random embedding stability, and the method comprises the steps: carrying out the embedding in a face recognition neural network, obtaining a face quality evaluation score during the face recognition process, and specifically includes: 1, carrying out the frame extraction of each image frame in a video stream according to a preset rule; 2, performing face detection on any image frame, and performing cutting alignment on the detected face to obtain a face image; 3, inputting the face image into a first neural network to obtain a corresponding face feature; 4, inputting the face features into a second neural network to obtain n random features; 5, combining the n random features in pairs to obtain different random feature pairs, and calculating a face quality evaluation score according to the feature pairs; and 6, inputting the face features in the step 3 into a third neural network to obtain face recognition features for a subsequent face recognition or comparison process.
Owner:北京睿芯高通量科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Eureka Blog
Learn More
PatSnap group products