2340results about How to "Resolve accuracy" patented technology

A multi-temporal, multi-angle, automated target exploitation method is provided for processing a large number of images. The system geo-rectifies the images to a three-dimensional surface topography, co-registers groups of the images with fractional pixel accuracy, automates change detection, evaluates the significance of change between the images, and massively compresses imagery sets based on the statistical significance of change. The method improves the resolution, accuracy, and quality of information extracted beyond the capabilities of any single image, and creates registered six-dimensional image datasets appropriate for mathematical treatment using standard multi-variable analysis techniques from vector calculus and linear algebra such as time-series analysis and eigenvector decomposition.

The invention discloses a hybrid machine learning credit scoring model building method. The method comprises the steps of 1, determining client risk classification standards based on a loan client historical data set; 2, based on the loan client historical data set, obtaining a loan client data feature set through feature extraction; 3, selecting at least two model algorithms from an alternative model library, building corresponding models based on the selected algorithms, performing model performance check on the built models by adopting a K-fold cross check method, performing standard check on the models about to pass the model performance check based on model check standards, obtaining evaluation index statistical quantity values, and selecting a model type used by final modeling according to the evaluation index statistical quantity values returned by the standard check of the models; and 4, based on the algorithm corresponding to the selected model type, building a credit scoring model. The technical effect of efficiently and accurately finishing user credit evaluation through the built hybrid machine learning credit scoring model is achieved.

The invention discloses a man-machine cooperation device and method based on a Kinect video camera. The man-machine cooperation device and method based on the Kinect video camera is mainly based on a Kinect video camera which is launched by the Microsoft Corporation recently so as to achieve detecting and tracking of a human body target. The man-machine cooperation device and method based on the Kinect video camera can accurately judge the position of a target object and the intention of a tester by utilizing Kinect human skeleton detecting technology, and therefore the risks which occur in the process of handover are avoided. Besides, a rail rapid transit (RRT) algorithm based on working space is provided to achieve route planning. Through repeated experimental verification, the system can conduct interpersonal handover safely.

A gravity gradiometer is combined with a two-stage actively controlled isolation system. The gravity gradiometer and two stage isolation system may then be mounted within (or on) a mobile vehicle such as, for example, an aircraft. It has been recognized by the inventors herein that the accelerations imparted to an aircraft during normal operations can be separated through system design into two relatively distinct regimes within the frequency domain. The invention provides a first isolation mount, which forms part of the isolation system, to isolate accelerations (and resulting translations) falling within a first of the two frequency regimes. The second isolation mount, which is mounted to the first isolation mount, isolates accelerations falling within the second of the two frequency regimes. A gravity gradiometer can then be mounted to the second isolation mount. As a result of housing the gravity gradiometer within the nested isolation system (a combination of the first and second isolation mounts), the gravity gradiometer is substantially isolated from the accelerations experienced by the mobile vehicle. Consequently, gravity gradients measured by the gravity gradiometer are relatively noise free and provide heretofore-unobtainable accuracy.

The present invention discloses a deep learning-based SAR remote sensing image water surface target detection method, which mainly solves the problems of the slow detection speed and low positioning accuracy in the existing SAR image water surface detection method. The specific implementation is: designing and pre-training an image classification model; then designing and training a target detection model based on the model, wherein the target detection model comprises an RPN region suggestion network and a Fast R-CNN target detector; and finally using the trained model to detect the SAR imagewater level. The method disclosed by the present invention has the advantages of a high detection speed and high detection accuracy, and can be used for water surface target detection of large-area SAR images.

The invention discloses a voice wake-up method, device, terminal and storage medium, and belongs to the technical field of voice recognition. The method comprises the steps that a wake-up voice used for waking up an intelligent terminal is acquired; a network is recognized by means of a prebuilt acoustic model and a preset wake-up word of the intelligent terminal, analysis is conducted on the acoustic characteristics of the wake-up voice, and the confidence of the acoustic characteristics of the wake-up voice relative to the preset wake-up word is acquired; if the confidence is in the range of preset medium confidence, the wake-up voice is uploaded to a remote server; if the language characteristics acquired by analysis of the wake-up voice by means of a language model in the remote server are matched with the language characteristics of the preset wake-up word, an intelligent terminal wake-up instruction generated by the remote server is received. According to the voice wake-up method, device, terminal and storage medium provided, a custom of the wake-up word is achieved.

The invention discloses a method and user equipment (UE) for measuring and feeding back channel quality, which is used for solving the problem that a cell edge user (CEU) and a radio link failure (RLF) can not be distinguished accurately caused by incomplete reference signal receiving power (RSRP) and / or reference signal receiving quality (RSRQ) information and the problem of certain influence on link adaptive performance caused by only feeding back one channel quality indication (CQI) in the prior art. According to the technical scheme of the invention, the UE measures and feeds back the RSRP and / or RSRQ information and channel quality indication (CQI) information of subframes which are almost fully empty subframes corresponding to strong interference neighbor cells and the RSRP and / or RSRQ and channel quality indication (CQI) information of subframes which are normally transmitted subframes corresponding to the strong interference neighbor cells, thereby solving the problem of accurately distinguishing a cell center user (CCU) and the CEU in a heterogeneous network and preventing the RLF phenomenon. In addition, the invention can enable users to obtain better link adaptability according to the CQI of different types of subframes when data is transmitted on different subframes, thereby improving the system throughput.

The invention relates to a PDA (personal digital assistant)-based realization method for intelligent patrol of electric equipment. The method is characterized by comprising the following steps: downloading patrol task information to a handheld PDA from an office PC (personal computer) by a patrol staff, patrolling predefined equipment patrol items one by one through an RFID (radio-frequency identification) electronic label by the handheld PDA during outdoor patrolling, and recording data content; after the patrol task is finished, submitting a patrol result to a server through a patrol management host. According to the method, the patrol quality of the electric field equipment can be improved, the problem of non in-place patrolling, missing patrolling, non on-time patrolling or the like is solved, the patrolling efficiency is improved, and the situations of low efficiency, high possibility of missing items and errors during manual reporting in the past are avoided.

Embodiments of the present application provide a method and an apparatus for labeling image data, wherein the method comprises the following steps: obtaining image data to be labeled; determining a target region in the image data to be labeled by using a preset detection network model, and marking a position of the target region; using the preset classification network model to recognize the imagedata at the position of the target area, and labeling the recognition result. The pre-trained detection network model is adopted to find and label the target area, and then the classification networkmodel trained in advance is used to identify and label whether the image data in the target area is the target object or not, thus solving the technical problems of poor accuracy and low efficiency in the existing data labeling methods, and achieving the technical effect of automatically labeling a large amount of image data efficiently and accurately.

The invention discloses a method for highly sensitive quantitative detection of quantum dot fluorescence immunochromatographic assay. The method includes: building a fluorescence immunochromatographic assay test strip on the basis of optimizing the structure of the test strip and components by the aid of excellent fluorescent characteristics of quantum dots and by means of combining quantum dot fluorescence labeling technology and immunochromatographic assay; detecting fluorescence signal strength of a quantitative belt and a quality control belt by the aid of a fluorescence quantometer and correcting the fluorescence strength of the quantitative belt by the aid of the quality control belt after immunochromatographic assay of the test strip; and further quantitatively detecting analyte according to a standard curve obtained by the fluorescence quantometer. The method is simple, rapid, accurate, low in cost and quite high in sensitivity. Compared with a conventional colloidal gold immunochromatographic assay method, the method has the advantages of fine labeling stability, low non-specificity, high sensitivity, wide linear range and accuracy in quantization. The method is applicable to samples such as blood samples, urine samples, spittle, excrement and the like, and can be applied to detection of critical illness, poison, food safety and the like.

The invention discloses a network hot event detection method based on text classification and clustering analysis. The method solves the problem that the efficiency and accuracy rate of the existing network hot event detection method based on clustering analysis need to be improved. The method comprises the steps that feature words are respectively selected for various classes of files through feature extraction and feature selection by utilizing a training corpus; each training text and test text are represented as vectors in all of the feature spaces by utilizing a vector space model method, and the weight of each dimension of the vectors is determined by utilizing a TF-IDF (term frequency-inverse document frequency) method, and then each test text is classified; the classified test texts in different classes are respectively subjected to clustering analysis, so the hot cluster of each class is obtained, the feature word representing the hot event is obtained through further analysis, and then the word property and other aspects of each feature word are analyzed; the description of each hot event is generated by utilizing relevant language knowledge and necessary linguistic organization. With the network hot event detection method based on text classification and clustering analysis, the detection efficiency and accuracy rate of hot events can be effectively improved.

The invention discloses a system and method for recommending domain information based on a microblog platform. The system comprises a data obtaining and preprocessing module, a domain key word extracting module, a user defined key word extending module, a linear combining module, a similarity calculating and personality recommending module and a subject obtaining module; the method for recommending the domain information is designed and achieved aimed at the character of the microblog platform. According to the method, seamless combination is conducted between key word extracting and key word expanding, thus the extracting of domain character is guaranteed, the dynamism of a recommending result is also guaranteed, and through the experiment of the corresponding system based on Sina microblog, the effectiveness of the method is verified; the method can assist enterprises in microblog marketing, and thus the efficiency in microblog marketing of the enterprises can be effectively improved.

A power system fault diagnostic method belongs to the power system dispatch and fault analysis domain, comprising: step 1 of downloading the fault data base of the power system to process to form a rule base; step 2 of collecting the voltage of the power system, current, the on off state signal of the breaker and the action signal of the relaying protection using the collecting device; step 3 of performing on-line fault diagnosis; step 4 of uploading the fault information to the host computer of the device and on-line adding the fault. The prior data collecting device, the mature technique and the outlying theoretical knowledge are combined together. The modified general rough intensivism contracted calculation can solve the problem that it is incapable for diagnosing or accurately diagnosing due to information deletion when the large scale chain fault generates. Especially the method can solves the fault diagnosis under 'avalanche data' which can not be solved by conventional method and ensure the system safely and stably run under each condition.

The invention discloses a picture blurring method and device and electronic equipment, and belongs to the field of picture processing. The method comprises the steps of obtaining a first picture and a second picture, with the same framing range, taken by the same camera; obtaining depth information according to the first picture and the second picture; determining the background region of the first picture according to the depth information; and blurring the content in the background region of the first picture. Thus, the problem that a user needs to open a picture through software and select the background region in the prior art is solved, the blurring effect is improved, and the user operation is simplified.

The invention discloses a doctor advice intelligent auditing system and method based on drug rule combination. The system comprises doctor advice basic rule base and rule combination library maintenance module, a doctor advice extraction module, a doctor advice checking module, a label printing (two-dimensional code generation) module and a departmental scanning module. The doctor advice extraction module is used for extracting information of doctor advices from electronic medical records; the doctor advice checking module performs checking in the mode of repeated dosage checking, combination rule joint checking and incompatibility checking to find checking alarm information to send the information to a clinical pharmacist for confirmation, and drugs with problems are turned down to a doctor; the doctor advise basic rule base and rule combination library maintenance module is used for maintaining and updating a combination library; the label printing module is used for labeling checked drugs to generate two-dimensional codes; the departmental scanning module performs departmental scanning on the two-dimensional codes of the drugs when the drugs depart. The doctor advice intelligent auditing system is used for recording the departmental time of the checked drugs and tracking the using process of drugs conveniently.

The invention discloses a battery power saving method for a portable electronic product. The method comprises the following steps: detecting current battery power of a system; sending a low power prompt signal when the current battery power of the system is lower than a low power triggering threshold; and setting a power saving mode expected by a user according to the low power triggering threshold, wherein the power saving mode is associated with a plurality of sets of operation parameters and / or on-off states related to power consumption of the system. The invention further discloses a battery power saving apparatus for the portable electronic product. The battery power saving apparatus comprises a battery power detector, a low power prompting device and a power saving mode setting device. When it is detected that current battery power of the system is lower than the low power triggering threshold, the low power prompt signal is sent, and the power saving mode expected by the user is set according to the low power triggering threshold. On the basis of the apparatus, the invention also discloses a mobile terminal of the portable electronic product using the battery power saving apparatus. The mobile terminal guarantees that the user sets a correct power saving mode at the right time, so the purpose of battery power saving is realized in reality.

The invention discloses a video duplicate removal method and device. The video duplicate removal method includes the first step of dividing a first video set into a plurality of video subsets according to video text data, the second step of calculating the similarity of two videos in the video subsets through a spatial vector model, and the third step of carrying out duplicate removal processing on the videos in the video subsets to obtain a second video set according to the similarity. According to the video duplicate removal method and device, the problems that in the prior art, accuracy of duplicate removal conducted on the videos is low and calculation complexity is high are solved, and accurate and quick duplicate removal on massive videos can be achieved.

The invention discloses a method and device for processing a brilliant video replay clip. The method comprises the following steps of: detecting and acquiring all identifier occurring time points in a video in an identifier template set, wherein the identifier template set is used for storing a preset identifier template; matching the time points to obtain brilliant video replay clip information of the video and storing the brilliant video replay clip information in a brilliant replay database; and when a user orders the video, inquiring in the brilliant replay database to obtain brilliant video replay clip information, and playing a brilliant video replay clip corresponding to the brilliant video replay clip information. According to the invention, the brilliant video replay detection accuracy and flexibility can be enhanced, and the user experience is improved simultaneously.

The invention relates to a finite element modeling and bearing method for a line-tower coupling system of a power transmission tower. In the method, a three dimensional CAD (Computer Aided Design) model of the whole structure component of the power transmission tower is actively simplified; a frame characteristic structure is extracted, and a finite element model base is established according to the frame characteristic structure; a single plane is gradually transited to multiple planes by transforming work planes according to the actual shape of the power transmission tower structure, and the single plane structure is gradually 'spliced' to form a multi-plane structure dynamically; a lead is subjected to parameter modeling by adopting an analysis and numerical value comprehensive method,and is combined with the power transmission tower model, so that the finite element model of the line-tower coupling system of the power transmission tower is finally determined; and wind speed, temperature and other parameters are input in a loader program to perform uniformly distributed load, and after the calculation is completed, a required result is analyzed and extracted automatically by virtue of the program. By utilizing the method, the finite element analyzing circle is effectively shortened, the computational mechanic theory of the line-tower coupling system of the power transmission tower is represented, and the finite element model is more accurate than the CAD model geometrically.

The invention discloses a method for estimating service life of an MOS (Metal Oxide Semiconductor) device. The method comprises the steps of: analyzing the relevance of a physical failure mechanism and a device macroparameter degradation, constructing a device macroparameter degradation model based on the physical failure mechanism; carrying out reliability analog simulation on a device, analyzing the sensitivity of the device macroparameter degradation to the physical failure mechanism; selecting a stress applying mode, carrying out a parameter degradation service life-based test on the device and obtaining reliability test data; analyzing the reliability test data, establishing a parameter degradation-based service life estimating computing model; and obtaining a service life estimating result of the device according to the service life estimating computing model. By using the method, the problems of accuracy and maneuverability of the traditional device service life predicting method under the novel technical conditions are solved, the quantity of test samples is reduced, the testing cost is lowered, and the accuracy of predicting the testing result is improved.

The invention belongs to the technical field of computer vision and artificial intelligence, and particularly relates to a large-scale face recognition method based on a depth convolution neural network model. The method comprises steps of putting forward a residual error learning depth network model facing large-scale face recognition, wherein the residual error learning depth network model is formed by a convolution layer, a residual error layer and a full connection layer, and the residual error layer is formed by adding one path of multiple convolution layer cascade data and one path of original data to calculate the sum; and carrying out normalization operation in batch after each convolution layer in the model. According to the invention, by use of the characteristics of strong learning ability and good residual error learning convergence of the depth convolution neural network, layers of the model are increased in the aspect of the layer number of the network model; and in the aspect of residual error layer structure, the invention provides a highly efficient residual error layer structure. In the field facing the large-scale face recognition, the accuracy of the provided method is greatly improved compared with a base line model, and the accuracy of face retrieval in a million-class face recognition database can reach 74.25%.

The invention discloses a message push method and apparatus, and belongs to the field of computer and Internet technologies. The method comprises the following steps: obtaining user attribute data and / or historical behavior data of a user, wherein the user attribute data is used for reflecting personal basic information of the user, and the historical behavior data is used for reflecting personal habits and application preferences of the user on a network; detecting whether the user attribute data and / or the historical behavior data meets push conditions of messages; and if at least one from the user attribute data and the historical behavior data meets the push conditions of the messages, pushing the messages to a client used by the user. According to the invention, the problems of low message push accuracy and waste of network resources in the prior art are solved; and messages of interest are pushed to the user, the message push accuracy is improved, and at the same time, the network resources are saved.

The invention discloses a semi-supervised intrusion detection method based on a depth generation model. The method comprises the steps of: 1, preprocessing data: converting symbol attributes in a dataset into numerical attributes, and then normalizing all the numerical attributes; 2, converting high-dimensional feature representations of labeled and unlabeled data into low-dimensional representations of a new feature space by using the variational self-encoding technology in the generation model, adding a constraint to low-dimensional feature vectors to obey Gaussian positive distribution soas to obtain a hidden variable z, and training a classifier by using the hidden variable z in combination with a labeled sample; 3, reconstructing labeled sample data: jointly generating a new labeledsample by using the hidden variable z in combination with label class information; 4, reconstructing an unlabeled sample: predicting the probability of each class of an unlabeled sample by using thehidden variable z, and then generating a new unlabeled sample in combination with the hidden variable z; and 5, calculating a reconstruction error of the model with the newly generated labeled and unlabeled samples, and training and optimizing model parameters in combination with a classification error till convergence.

The invention discloses a space interaction method of mobile equipment with the human-computer interaction capacity. The method includes the steps of rebuilding a depth map of an image according to the two-dimensional image shot by the mobile equipment in real time; performing region segmentation on the depth map to obtain a human body position region in the depth map; mapping the human body position region in the depth map to a virtual scene; and detecting whether the human body position in the virtual scene collides with other objects in the virtual scene, if the collision happens, determining a body language according to the temporal space change of the human body position in the virtual scene, and the virtual scene making response according to the body. According to the method, accuracy and timeliness of space interaction can be improved.

The invention relates to a cross-time multi-dimensional abnormal data monitoring method and system. The method includes: collecting Web logs and OLA log queries to obtain website information, delivery advertisement information and user information; cleaning obtained heterogeneous data by employing Python in a Spark environment, generating an electronic table corresponding to each dimension, and preparing modeling analysis; establishing a normal advertisement click behavior model by employing a GBDT algorithm; conducting a modeling analysis on an abnormal data sample by employing a support vector machine, and establishing an abnormal advertisement click behavior model; and generating a cheating click monitoring model, and deploying an online anti-cheating model. According to the method and system, the problems of fusion analysis of various heterogeneous data and small sample classification accuracy are solved by employing machine learning and cloud computing technologies through monitoring of cheating Internet advertisement traffic, subsequent Internet advertisement click data can be recorded and analyzed in a whole machine learning chain, the anti-cheating model is improved, and the accuracy of advertisement delivery can be better achieved.

The invention belongs to the technical field of image segmentation, in particular to an image segmentation method, a system and an electronic device based on depth learning. The image segmentation method based on depth learning comprises the following steps: step A: normalizing the original image; B, inputting the normalized image into a ResUNet network model, extracting a feature map containing global semantic information from the input image by the ResUNet network model, upsampling the feature map and stacking the feature map to obtain a final feature map; step c: classifying pixel by pixelthe feature map after upsampling and stacking processing, and outputting the image segmentation result. The present application solves the common problem of gradient disappearance in a convolution neural network, and also enables the network to be more easily trained and converged to a better segmentation result.

The invention discloses a lithium-battery variable fractional order and equivalent circuit model and an identification method thereof. The lithium-battery variable fractional order and equivalent circuit comprises a run time circuit and a battery I-V characteristic circuit, wherein a capacitor in the battery I-V characteristic circuit is a variable fractional order capacitor. A second order RC circuit model is generalized to a non-integer order, and the model parameters and the number of fractional order of different SOC are identified based on a least square method, so that the fractional order and equivalent circuit varying order according to the SOC is obtained. The instruction of fractional order realizes the continuous change of the order number of the model, so that the model is relatively stable, good in dynamic property and high in precision. The variation of fractional order realizes more freedom and more flexibility and innovation of the model. As the number of RC networks is not increased, the fractional order model effectively solves the contradiction between the accuracy and practicality of the model, is suitable for various working conditions of batteries, and has high practical value. The invention provides a precise battery model easy to realize for precise estimation of SOC.

This application discloses a business data processing method, an account identification method, a business data processing device, an account identification device, and a computer terminal. The methodincludes the following steps that: a variety of kinds of information of an account to be identified are obtained; a risk value corresponding to each kind of information in the various kinds of information is obtained, wherein the risk value is used to indicate that the reliability degree of the account to be identified; the account type of the account to be identified is determined according to the plurality of risk values corresponding to the plurality of kinds of information; and the business data of the account to be identified are processed through adopting a business processing mode corresponding to the account type of the account to be identified.

The invention discloses an operation control method and device of a virtual object in a game and a mobile terminal. The method is applied to the mobile terminal capable of presenting a graphical userinterface on a screen, and comprises the steps of determining second virtual objects satisfying target conditions as target second virtual objects in response to a first touch operation acting on thegraphical user interface, wherein the distance between each second virtual object and a first virtual object is in accordance with a preset threshold value, and the first virtual objects execute a corresponding action according to an operation instruction received by the mobile terminal; generating an object selection control according to the target second virtual objects, wherein the object selection control comprises an object identifier; determining at least one object identifier according to the second touch operation and selecting the target second virtual objects corresponding to the object identifiers in response to the second touch operation acting on the object selection control. The method solves the technical problems of low selection efficiency and poor accuracy of game objectsin existing multi-player team games.